Search

Humanitarian demining is an indispensable part of any post-conflict reconstruction and rehabilitation efforts, as landmines and other unexploded ordnance (UXO) continue to pose a significant threat to civilians and infrastructure. Traditional non- technical survey demining methods are typically time-consuming, labor-intensive, and require professional familiarity with a number of different complex tools and involved procedures. The emergence of the latest artificial intelligence (AI) technologies for deep learning (DL) and data analysis techniques presents an opportunity to transform humanitarian demining by significantly improving its efficiency, speed, accuracy, and safety. One such approach is the development of a data observatory that implements data ingestion pipelines into a data lake using data transformation to improve the usability of unstructured data about a mine scene. This paper addresses the key characteristics and benefits of such a system and its potential to improve humanitarian demining.

Logical representation of knowledge enables reasoning based on formal logic, integration of heterogeneous data sources, augmentation of existing knowledge, and discovery of hidden knowledge that is hidden or unknown even to experienced professionals. Moreover, many high- performance software tools and services for logical reasoning are readily available in the academic community. Computational ontologies, or ontologies for short, are a comprehensive, formal, scalable, and adaptable methods for knowledge representation and automated reasoning. They have been successfully used in many diverse application domains, from medicine and natural sciences to engineering and applied sciences. However, the use of ontologies in the context of demining is novel and has not been exploited to the full advantage. In this context, a new core ontology called MINEONT (an acronym for "MINE-action ONTology") has been proposed. The MINEONT is written in the OWL-DL 2 formalism and allows an expressive and formal representation of concepts in mine action, high- level semantics, geospatial metadata, and information obtained through method of remote sensing, non-technical surveys, and other mine action survey types. The vocabulary of this ontology supports, among others, the following corpora: multisensory aerial and satellite imagery, derived indicators of mine presence and absence, contextual data, terrain analysis information, formalized knowledge of the humanitarian demining specialists in the form of declarations and procedures. The designed ontology has many advantages in describing data for non- technical surveys. However, it is also relevant for building a specialized recommender artificial intelligence (AI) expert system. Such an ontology- based decision support system (DSS) would integrate a large amount of collected data and facilitate interpretation and decision-making processes. In the paper, details about the Web Ontology Language (OWL) are given and its advantages for application in the aerial non- technical survey for humanitarian demining are explained. The MINEONT model is introduced and described in detail. The advantages of MINEONT compared to other existing methods in the aerial non-technical survey system are planned to be tested and demonstrated as part of scientific projects in which domestic and foreign partners, including CROMAC, will participate.

In mountainous regions such as mountainous Croatia, natural and semi-natural landscapes are most common and are changing due to climate change and more intense economic activities. Due to these changes, there is a need of inventarisation and valorisation of mountainous landscapes in order to manage and preserve them. In Croatia, there is no inventory that would provide us with a zero status of mountain landscapes. On the other hand, the mountainous environment and the size of the area ​ present an obstacle for landscape mapping. The use of Sentinel-2a imagery and a digital relief model allow ​a fast, inexpensive, and accurate mapping of landscapes in large areas such as mountainous Croatia. Orthophoto images were used to increase the accuracy of sampling and thus the accuracy of the final classification. Remote sensing techniques, GIS analysis and field validation were used to create an inventory of the landscapes types. This type of landscape mapping, which includes geological, geomorphological and land cover data, provides a better understanding of the composition and configuration of mountain landscapes, leading to knowledge-based management and landscape-level protection.

The results of this analysis show that the trends towards open and unrestricted access to data, software, standards, and processing capabilities of the existing cloud computing platforms generally support basic scientific principles such as openness, accessibility and reusability. However, the challenges have been identified that need to be resolved in order to move from open Earth observation data to fully open analyses. For example, fair and impartial access for a heterogeneous group of users’ needs to be ensured, open access to the source code as well as the documentation needs to be provided, and non- restrictive licences that allow the creation of derivative works need to be applied. Looking ahead, we hope that this comparative analysis of the five operational platforms could inspire interested experts and non-expert users to explore the new paradigm of cloud-based platforms for Earth observation data access and processing.

The industry of sensors and unmanned aerial systems (UAV) enables the advancement of detection of explosive objects (landmines, unexploded ordnance, improvised explosive devices) and the non-technical survey of areas contaminated by explosive threats. In the last ten years, dominates research and development of the direct detection of explosive devices versus detection of indirect indicators of the explosive threats on large areas in frame of civilian Non-Technical Survey (NTS). The impact of the mentioned phenomena is not used to its full potentials in the civilian domain of NTS. Although UAV technology can provide a large amount of data, interpretation of the data and the decision-making as an outcome of the interpretation are not developed yet. Between 1998 and 2020 many research and technology development projects on aerial NTS are realized. Some of them applied Decision Support System (DSS) paradigms: GIS-based DSS paradigm, Multi-Criteria-based DSS paradigm, Automatic Target Recognition (ATR) paradigm. The Scientific Council of HCR Center for testing, development, and training, initiated in December 2019 research and development of the Aerial Non- Technical Survey, based on UAV data acquisition, and a recommender system in its DSS functions. The proposed recommender system utilizes deep learning algorithms and collaborative filtering (CF) that rely on datasets containing selection preferences for different objects. DSS knowledge database will be based on ontologies to facilitate formal description of complex semantic structures, knowledge transfer, and reuse. The objects and preferences will be derived by fusion of features of secondary indicators, ranked by eigenvalues. The novel DSS will be implemented in the new Aerial Non-Technical Survey System which will be realized through projects by several partners. The operational validation will be done by CROMAC in mine suspected areas of Croatia.

On 19 June 2020, Ivan Racetin defended his doctoral thesis Anomaly detection on hyperspectral and multispectral digital images of a mine suspected area at the Faculty of Geodesy, University of Zagreb., Ivan Racetin obranio je 19. lipnja 2020. na Geodetskom fakultetu Sveučilišta u Zagrebu doktorski rad Detekcija anomalija na hiper­spek­tra­lnim i višespektralnim digitalnim snimkama minski sumnjivog područja.

More than 50% of the world’s populations live in urban areas, and they emit almost 80% of global carbon dioxide. They are also responsible for 75% of energy consumption in the world. That increases pressure on food, water, and especially energy. These facts imply that there is an urgent need to provide low carbon dioxide level in cities with efficient and renewable energy to foster the growth of the green economy. Therefore, attention should be paid to renewable energy sources such as the Sun and be aware of the potential of solar power of some areas. In favor of this assertion, recently published studies have suggested that solar power will generate 20% of global electricity by 2027. The above-mentioned facts and predictions were the motivation for conducting preliminary research on the possibility for predicting solar potential. The predictions of solar power can be provided by means of remote sensing methods. Digitalized analog VNIR aerial images (UMK camera) with very high spatial resolution and Worldview2 channels are used as input data. For the purpose of discriminating the roofs, the classification was made by controlling both sets of images. The average number of sun hours in one day and the average consumption of an ordinary household were calculated. Initial results obtained in this preliminary research are demonstrated in this paper for an urban neighborhood in Zagreb (Croatia).

The resulting map seek to improve the identification of areas where there is no threat so that parts of suspected hazardous areas can be proposed for mine reduction, or suspected hazardous areas can be better defined.

The key object of interest in airborne and satellite Non-technical Survey are the indicators of mine presence and absence. That is why it is of great importance to establish an adequate list of indicators that will ensure the quality use and implementation of aerial and satellite Non- technical Survey. Close cooperation of all experts in the process of airborne and satellite Non- technical Survey is needed. Experts in the Mine Action Centers conduct an analytical assessment of existing Mine Information System data and establish a list of special and general requests for collecting additional information from the depth of Suspected Hazardous Area. Experts in remote sensing methods help them in forming the list of indicators for a specific geographic area. After multi-sensor airborne data collection and procurement of satellite images of a particular Suspected Hazardous Area, mine scene interpreters are locating and extracting the indicators from these images. The level of confidence is then assigned to the extracted indicators. In locating the indicators in the images, mine scene interpreters use expert and contextual knowledge of conducting war doctrines in particular geographic areas and (optional) simpler image processing methods. Geocoded indicators and their confidences are input for the production of Mine Danger Maps. This paper describes the role of indicators in Non-technical Survey processes in Croatia and Bosnia and Herzegovina within TIRAMISU – Advanced Intelligent Decision Support System.

Landmines affect the lives of people living in war-affected countries. The real challenge in humanitarian demining nowadays, within Non-technical Survey, is finding new approaches to SHA reduction. The establishment and use of mine information systems in humanitarian demining can improve the land release process by efficient mapping of suspected hazardous areas and visualizing of mine risk on it. Croatian Mine Action Center established such system and uses it for marking all suspected hazardous areas in Croatia. Nevertheless, data in this system comes from different sources and contains some inherent uncertainty. The level of ambiguity is being attempted to diminish by conducting analytical assessments of mine information system data and by collecting additional information about suspected hazardous areas. The landmine risk mapping requires defining the input data (indicators) from mine information systems and additional data that most suitably capture the association between the spatial objects of interest in the surrounding geographic area. The confidence level for each indicator is determined, which is, in fact, an estimate of the level of uncertainty. The fusion of all existing data is used after that for the purposes of creating of mine danger maps. These maps show the overall uncertainty of all input data and need further action for their use in reduction according to the Croatian Mine Action Center Standard Operational Procedures.

In the mine action community, Non-Technical Survey is implemented to provide information for better spatial definition of mine or explosive remnants of war contamination. Airborne and spaceborne surveys of wider Suspected Hazardous Area (SHA) regions are often executed to collect information about physical changes and circumstances that may be employed in the subsequent technical interventions. Simple electro-optical sensors which acquire data from the visible part of electromagnetic spectrum, like DSLR or SLR cameras, can be used for this kind of surveys. These aerial RGB images are then inspected by trained human photo (mine scene) interpreter who is familiar with the military doctrine and specific characteristics of the conflict. Mine scene interpreter seeks for the indicators of mine presence and absence. Novel image processing techniques could provide benefits to the standard photointerpretation workflow by making it more efficient. Geographic object-based image analysis (GEOBIA) is one of them. The idea of GEOBIA is to process georeferenced images at the object level, as the entities found in the environment are usually objects (or grouped pixels) instead of single pixels. This approach enables analysis of spatial relations and geometrical object parameters. The main step influencing the GEOBIA results is the quality of image segmentation. In this paper, results of more image segmentation algorithms are tested on the color aerial images of SHA and images of exploded ammunition depot in the Croatia. Acquired results are discussed, while potential and constraints of image segmentation for application in the Non-technical survey are underlined.

Although calibration during geocoding process shows drastic improvements in achieved geocoding accuracy, the internal orientation error remains. In order to properly determine these errors, we must first of all know the mathematical model of mapping of the line sensor. The mathematical model of mapping of the line sensor has been expanded to account for the distortion of the lens in order to eliminate all systematic errors induced by displacement of the real sensor mapping plane from the ideal mathematical model of the central projection. After a mathematical model was established, a calibration of the hyperspectral scanner was performed in the laboratory, observing test samples from 5 different recording sites. Calibration elements are determined by airborne beam equation (BBA method).

Unplanned explosion of the ammunition storage depot in Padjene, Croatia, triggered by the forest fire, happened in September 2011. Croatian research teams engaged in EU FP7 TIRAMISU project, CROMAC – CTDT Ltd. (Croatian Mine Action Center - Centre for Testing, Development and Training) and Faculty of Geodesy, University of Zagreb, promptly reacted by deployment of data acquisition module of Advanced Intelligence Decision Support System (Bajic 2010) for the aerial survey of impacted area. Troubling remnants of the explosion were scattered unexploded ordnances found in the various forms: undamaged, lightly or heavily deformed, burned or corroded. Remaining unexploded ordnances were not varying only in the form, they varied in the size also, ranging from rifle ammunition to cluster bombs. For dealing with these situations ground teams of demining experts are activated for clearance and recovery tasks. Primary objective of aerial survey in visible spectrum is to provide ground teams overview of affected area which serves as input data for planning of terrain clearance actions. This paper explores the capabilities for implementation of anomaly detection methods for extraction of unexploded ordnances on aerial RGB images acquired by consumer grade digital camera. There are two approaches in image analysis for the needs of object detection and extraction: classification and target detection. In the classification approach, the aim is to assign every pixel to the corresponding thematic class. Target detection approach, contrarily, scans the image with the aim to identify the presence of specific object or material (target). Although target detection can be regarded as binary classification with two classes: background and target, due to the difference in the essence of these approaches they are being distinguished. Number of classes is not the only difference between these approaches, term target also states that the number of the target pixels needs to be negligible compared to the number of the background class pixels. Result of this requirement is that different quality assessment methods need to be implemented depending on the approach used. Classification approach, mostly based on Bayes criterion, tends to minimize the probability of detection errors (false alarms and misses). This can easily be achieved by classifying every pixel as background, because of very sparse population of target pixels. Popular Neyman-Pearson criterion (Kay 1998), which maximizes the probability of detection while keeping the probability of false alarm at the constant rate, is more optimal for the target detection. Anomaly detection methods are a subset of target detection methods in which no a-priori information about the target spectra is available. Reed-Xiaoli (RX) (Reed & Yu 1990) algorithm is the benchmark anomaly detector: it calculates the Mahalanobis distance (Mahalanobis 1936) between the pixel under test and the background. Mahalanobis distance uses covariance matrix and mean value which can be calculated from local background (vicinity of pixel under test) or global scene covariance matrix and mean, so we distinguish local and global RX algorithm. In this paper, local and global RX algorithms were applied to the images of exploded ammunition depot acquired by Nikon D90 camera mounted on helicopter. As no ground truth data was feasible, images were visually interpreted. Object were vectorized using the object-based image analysis to reduce the human error in the manual vectorization. Experimental receiver operating curves (Kay 1998) were then constructed for both anomaly detection algorithms used. Results showed that application of anomaly detection methods can speed up the process of visual interpretation, but it cannot be used as completely automatic method for extraction of unexploded ordnances. REFERENCES Bajic, M. (2010): The Advanced Intelligence Decision Support System for the Assessment of Mine-suspected Areas, The Journal of ERW and Mine Action, 14. Kay, S. M. (1998): Fundamentals of Statistical Signal Processing, Volume 2: Detection Theory. Signal Processing, Prentice-Hall, Upper Saddle River, NJ, USA Mahalanobis, P. C. (1936): On the generalised distance in statistics, Proceedings of the National Institute of Sciences of India, 2, 49–55. Reed, I. S., & Yu, X. (1990): Adaptive multiple-band CFAR detection of an optical pattern with unknown spectral distribution, IEEE Transactions on Acoustics, Speech, and Signal Processing, 38, 1760–1770.

Copernicus is a European Union Programme that was designed to provide service in delivering free and openly accessible data about Earth surface through satellite and in-situ observations. Copernicus provides images of a set of the Sentinel families and contributing missions. Copernicus also collects and provides information from the ground (in-situ data). Copernicus services are available worldwide today. Although the images of the satellite system are characterized by rough resolution (which limits the application of more precise analysis), they can be a good source of information and data that may provide a better observation of certain objects or phenomena of interest on the scene to experts and help them in decision making. This paper presents the analysis of the possibility to use satellite images of Sentinel system. The results of Sentinel system image analysis have been compared with the analysis results and processing of Landsat satellite images (that are also public and free). The information and data obtained as the results of interpretation and processing of Sentinel and Landsat satellite systems images can be used by experts in various fields of human activity as a support in making decision. The fields of human activity covered by the research presented in this paper are the following: forestry and agronomy ; humanitarian demining ; monitoring of natural and man-made disasters (floods, fires, environmental pollution). The results obtained by means of interpretation and processing of Sentinel and Landsat satellite images have been analysed and interpreted. The interpretation of the results has been made visually, statistically and then presented by means of confusion matrix. Some channels (images) of the Sentinel 2 satellite system have better surface resolution than all multispectral channels of the Landsat satellite system. However, the channels of Landsat satellite system have better surface resolution than some channels of the Sentinel 2 satellite system. In both cases, it is the surface resolution of 10 m and higher. Hence, it will not be possible to identify certain objects and phenomena of interest on the scene, but only to detect them with a certain level of confidence. Due to the mentioned fact, the processing has also been made that included the channels of both systems. The bands have been also mutually sharpened. Further research is planned to be carried out within the scope of future concrete project tasks in order to define the procedures more thoroughly that are intended for solving specific problems based on the conclusion obtained from performed research.

Salinization greatly affects the ecosystems, crop production and the economy. Source of salinity can be generally categorized as primary or naturally occurring and secondary or human induced salinity. Primary salinity stems from atmospheric effects on the rocks containing primary minerals for thousands of years, discharged ions are then combined in various salts which are later naturally distributed to remote geographical areas. Improper management of natural resources which results in soil and water salt build-up applies to secondary salinity. The area being analyzed in this paper is the alluvial river Neretva delta. Today, this area is used for intensive agricultural production of fruits and vegetables, from which citrus orchards are being most present. Agronomic production in the river Neretva delta poses substantial economic importance at a local and a national level as well. Lately, the sustainability of this production has become threatened by the increase in the salinity of groundwater and surface waters. Climate, hydrological and hydrogeological processes, including the engineering structures dictate the salinity fluctuation in the delta. There are two main directions of salt water penetration in the river Neretva delta: sea water intrusion through the riverbed and from deep underground layers where groundwater is already salinized because of direct connection with the sea through porous aquifer. Intensive agriculture demands high water levels, particularly in semi-arid climate, which then culminates with the rising of salty groundwater table. Furthermore, drop in quality of water used for irrigation due to reduced precipitation and extensive use of fertilizers and herbicides add to the total salinity in the river Neretva delta. Soil or water salinity is often depicted by measurements of electrical conductivity which is expressed in deciSiemens per meter (dS/m). This paper describes the interrelationship between in situ measurements of electrical conductivity in open watercourses with Landsat TM5 data in the period between May 2009 and October 2011. Water electrical conductivity was measured once per month on the 13 locations mostly placed in the melioration canals and near the pumping stations. Satellite data was firstly corrected for radiometric and atmospheric effects, afterwards the intensity, salinity, and vegetation indices were calculated. A total of 37 satellite images were analyzed in this study, where 23 of them are placed in the path 30 and row 188, while remaining 14 are placed in the same path but in the row 187 of the Worldwide Reference System. These images needed to be analyzed separately since pixel footprint of Landsat TM5 equals to 900 square meters. Pixel misalignment between these two groups of images and spectral mixing of classes at the pixel level, resulted in considerable differences in spectra at a given point for the two substantial images from different rows. In near vicinity of in situ sampling points, pixels classified as cultivated and uncultivated areas were selected. Correlation analysis of the original Landsat bands, their ratios and spectral indices with the electrical conductivity values was performed and accomplished correlation values were discussed.

In the scope of the EU FP7 TIRAMISU project a novel methodology for semi-automatic interpretation of digital multisensor images for the purpose of detection and extraction of unexploded ordnances was developed. Image interpretation is usually being executed by a professional human interpreter, which is highly reliable but simultaneously labor intensive and time consuming procedure. Experience showed that a human interpreter can’t be replaced by automatic methods of digital image processing, while some methods of digital image processing can reduce the workload of the interpreter. Methodology is based on combination of pixel and object based image analysis where lessons and rules learned on test dataset are then applied on other images of the same scene but different locations. Pixel based image analysis like transformations to various color spaces (IHS, CIELAB), Principal Component Analysis (PCA), Independent Component Analysis (ICA) and basic raster math were used. By executing these processing new datasets with different statistical properties are created. Now we have series of artificial layers which serve as input for statistical analysis and enables us better delineation of targeted objects. Object based image analysis offered additional geometrical parameters and reduced the human error in the process of manual vectorization of identified objects. Achieved results of methodology implementation on aerial images of exploded ammunition depot in Padjene (Croatia) are presented. Validation procedure was executed on randomly selected spatial subsets of 25 images and the outputs are shown in the form of confussion matrix.

Advanced Intelligence Decision Support System (AI DSS) is an operational Non-Technical Survey (NTS) system for collecting and processing (additional) data about suspected hazardous areas (SHA) ; data fusion and producing thematic (danger) maps about SHA. The existing Advanced Intelligence Decision Support System (AI DSS) was upgraded within FP7 project TIRAMISU (Toolbox Implementation for Removal of Anti-Personnel Mines, Submunitions and UXO) into TIRAMISU Advanced Intelligence Decision Support System (T–AI DSS) by implementing various new and improved sub-modules. T-AI DSS has benefited from the development of its components in TIRAMISU. The gaps identified by end-users and system operators have been filled in. The objectives include increasing its robustness, decreasing the workload of the operator, improving the semi-automatic mapping of features of interest for the detection of unexploded ordnance UXO, developing simpler imagery and data collecting units, implementing new sensors, optimizing application on different aerial platforms, verification in applications. T-AI DSS is an operational Non-Technical Survey (NTS) system which can integrate several NTS TIRAMISU tools and also use results obtained by several TIRAMISU tools which are not integrated into a system. The system consists of several sub-systems: for aerial multisensor image and data acquisition, for pre-processing and processing, for management of knowledge and contextual information, for implementing the outcomes of subjective photointerpretation, for fusion at pixel level, at features level and at decisions level. T-AI DSS is the only NTS tool within the frame of TIRAMISU that performs fusion. T-AI DSS is the operational TIRAMISU solution for NTS that is proposed to the mine action centres (MAC) worldwide because it is adaptable to specific terrain and situations. A simplified version (without remote sensing data acquisition) has also been developed that can be used in MACs for the support in the SHA assessment, reduction, re-categorisation and inclusion, indicating only mine presence and mine absence derived from MIS data. Services will be provided to realise NTS mission, to ensure transfer of know-how and capacity building.

Indicators of mine presence and absence are the main objects on the scene that need to be detected and spatialy positioned for the purpose of conducting quality aerial non- technical survey of suspected hazardous areas (SHA). Detection of these indicators on the images of the observed scene and positioning of the existing indicators from the mine information system (MIS) enables creation of danger map. Danger map is the result of the fusion of all existing data about mine indicators. Experts of warfare and mining in a particular area can bring conclusions on potential areas that are contaminated by mines on the basis of danger maps. However, the possibility of recognition and extraction of mine presence indicators on aerial and satellite images greatly depends on the image quality and corresponding ground sampling distance, as well as contextual knowledge and experience of mine scene interpreter. Therefore, due to the fact that the experts draw conclusions about SHA based on position of indicators in space and their zones of influence on the environment, it is necessary to add a level of confidence (the level of reliability with which the interpreter evaluates its decision) for each extracted indicator. This information is essential for a more realistic appearance of danger map depending on the available data on the observed scene and from MIS. The confidence of the decision that has been detected and extracted indicator of mine presence affects the level of danger of an area on the danger map. Tools used to automatically extract indicators on the mine scene can be validated in this way. Validation is done based on the true data about mine scene. Internal validation can be carried out by an experienced interpreter of mine scene, while operational validation will be performed after demining of the observed scene.

This paper describes application of methodology for semi-automatic interpretation of digital multisensor images for the purpose of detection and extraction of unexploded ordnances developed within the EU FP7 TIRAMISU project on the site of exploded ammunition depot in Padjene, Croatia. Process relies on combining the advantages of the both radiometric and object based image analysis using the statistical tools where lessons and rules learned on the test image are then applied on other images of the same scene but different location. In this case methodology has been applied on aerial images acquired by consumer DSLR camera mounted on helicopter flying at average height of 300 meters above ground level. Prior to this application perspective preliminary result was achieved for aluminum objects. During the further application, image processing for improved detection and extraction of corroded objects was defined and evaluated. Achieved results of methodology application on different scenes at same location (exploded ammunition depot). At the end, perspective of further research and analysis of this methodology is stated.

The crucial document for the success of T-AI DSS application is the analytical assessment of the suspected hazardous area (SHA) based on data and information that are available in MIS of MACs. The outcomes of this analysis are the general and specific requirements for the airborne and space borne collecting and producing new, additional data, information and evidences about the former situation in SHA. The requirement for improvement of this module implies the determination of general guidelines for the implementation of this analysis in any MAC’s of mine-affected countries. According to this requirement, within this T-AI DSS’s module, general guidelines for conducting of the analytical assessment for better definition of the SHA and the definition of general and specific requirements for the collection of additional data (with T-AI DSS) for better re-definition of the SHA have been developed. T-AI DSS is a tool to support decision about the status of SHA. The input includes data from mine information system (MIS), expert knowledge, airborne and satellite data, contextual data, etc. The outcomes are detected and proved positions of indicators of mine presence (IMP) in space, reconstruction of battlefield (in time and space), better (re)definition of SHA, proposals for exclusion from or inclusion in the SHA (thematic maps). The requirement for improvement analytical assessment of (MIS) data module of T-AI DSS refers to the need to determine the general guidelines for the implementation of this analysis in any MAC’s of mine-affected countries. Due to this requirement, general guidelines for conducting of the analytical assessment for better definition of the SHA and the definition of general and specific requirements for the collection of additional data (with T-AI DSS) for better re-definition of the SHA should be developed within this T-Al DSS’s module. Another requirement for the research within the scope of this module is the need to develop simplified version of the T-AI DSS (without the airborne multisensor acquisition and satellite images) that can be used in MAC’s for the support of the SHA assessment, reduction, re-categorisation and inclusion, only with indicators of mine presence and mine absence derived from MIS data and MAC’s standard operational procedures.

Computer-assisted image processing for extracting of IMP on the basis of object oriented analysis (using images segmentation) was carried out (by means of factor and cluster analysis) within the interactive interpretation of images. The final evaluation will be conducted by experienced interpreter and it will be also carried out on the basis of NIIRS and IQM parameters. The one aim of this research is to establish a methodology and guidelines for interactive semi-automatic interpretation of digital images in humanitarian demining for the purpose of detection and extraction of (strong) indicators of mine presence and absence which can be seen on the images. The experience gathered in working with AI DSS in undertaken projects led to the conclusion that a man (the human eye and contextual view on the scene) as interpreter (of mine scene) can hardly be replaced by automatic methods of digital image processing. However, some methods of digital image processing (implemented in certain ways) may help the interpreter in the identification of certain anomalies in the scene and easier detection of indicators. The basic idea of interactive methods of semi-automatic interpretation of digital images of SHA is to provide assistance to the interpreter in the interpretation of digital images rather than to replace him. The objective of this method is to be sensor independent and independent of object of research. The concept of this method was defined and initial research carried out in 2013.

The crucial document for the success of T-AI DSS application is the analytical assessment of the suspected hazardous area (SHA) based on data and information that are available in MIS of MACs. The outcomes of this analysis are the general and specific requirements for the airborne and space borne collecting and producing new, additional data, information and evidences about the former situation in SHA. The requirement for improvement of this module implies the determination of general guidelines for the implementation of this analysis in any MAC’s of mine-affected countries. According to this requirement, within this T-AI DSS’s module, general guidelines for conducting of the analytical assessment for better definition of the SHA and the definition of general and specific requirements for the collection of additional data (with T-AI DSS) for better re-definition of the SHA have been developed.

The module of T-AI DSS for data acquisition should provide the stability and reliability of (aerial) data acquisition on each platform. Subsystems (system configurations are different for different platforms) for aerial data acquisition will be examined on multiple platforms (Mi-8, blimp, Bell-206) and should have stable operation, without cancellation, on every platform. A new system for aerial hyperspectral survey: TIRAMISU light hyperspectral imaging system (T-LHSIS), was conceptualized by CTDT, designed, constructed and manufactured by FGUNIZ. T-LHSIS was tested on two platforms: blimp and Bell-206 helicopter by CTDT and FGUNIZ. The features of the system shown in the following text. T-LHSIS is a new tool, which can be used as stand alone equipment or in T-AIDSS. The helicopter Bell- 206 is a new platform which can be used within T-AI DSS. The continuous and stable electricity supply for the aerial acquisition multisensor subsystem is re- designed, developed and built up in T-AI DSS. It is necessary to provide the independence of the power supply for the aerial acquisition system of AI DSS. This module is used for the collection of additional data from the depths of SHA. The requirements (general and specific) for the collection of additional data on SHA-in are the result of work in the module for the analytical assessment of the data in MIS.

The consequences of explosion of ammunition storage depots are spatially scattered unexploded ordnances (UXOs). The unplanned explosion of ammunition depot, caused by extensive forest fire, happened in Pađene, Croatia. Airborne multisensor image acquisition (with Advanced Intelligence Decision Support System ‐ AI DSS) over destroyed ammunition depot in Pađene, Croatia, has been executed. In this paper we introduce a novel methodology for semi‐automatic interpretation of digital multisensor images for the purpose of detection and extraction of UXOs on images provided with AI DSS. It includes methods and procedures based on the statistical and object image analysis for improved identification and extraction of UXOs. The method is designed to assist the interpreter in the analysis of images, rather than as an automatic method for image analysis.

The general survey of the unexploded ordnance (UXO) scattered around the ammunition depot after its unplanned explosion and the ground based measurements of the hyper spectral samples (end members) of UXO have been done in order to create initial spectral library of UXO. The methods, technique and collected data are presented and evaluated. The Johnson’s criterion was used for detection, classification, recognition and identification of UXO in the airborne images (visible and near infrared wavelengths) of the surrounding of the ammunition depot. The test site near Benkovac contains 1000 landmines and 524 metal debris buried in the year 2001 in squares 1x1 m, at depths from 5 to 20 cm below the surface. The locations of 885 landmines and debris are not public and the blind tests are possible. The sparse vegetation covers some squares since the soil is meager. The hyper spectral data have been collected of the 1x1 m squares by the ground based mechanic scanner. The spectral samples of the vegetation inside and outside of the test site have been collected and total Nitrogen contents were measured.

This publication presents methodology for application of the Advanced Intelligence Decision Support System (AIDSS) for the assessment of the mine suspected area in the frame of the humanitarian mine action. Besides the methodology, the publication contains instructions, explanation and examples. Its aim is to enable humanitarian mine action experts in the Mine Action Centers (MAC) or Governmental Administration to prepare conditions and to apply AI DSS technology in their country.

Survey and definition of mine suspected area (MSA) of all territory of Republic of Croatia was conducted the by employees of Croatian mine action centre (CROMAC). However, due to the lack of information and in order to reduce the risk for the local population MSA was defined in larger dimension. In the following years, when after MSA was defined, CROMAC was updating it's mined information system (MIS) with new information and did permanently reduction/extension of MSA in accordance with the mentioned information. The multi-sensor system for reconnaissance and surveillance, and decision support system in conditions of uncertainty provide help in collecting and process new information for purposes of humanitarian demining community in Croatia. Requirements for new information on objects that are present on the scene are defined in collaboration with experts from CROMAC. Collecting and processing information are carried out according to those requirements. Collecting information with multi-sensor system is performed in order to find and position an object which was assumed by the experts that can be defended/protected with mines (indicators of mine presence - IMP) or an object for which experts assumed that can be free of mines (indicators of mine absence - IMA). In the conditions of uncertainty, the data fusion of all available data (MIS, GIS, ortho-photo, (war/military) maps, aero-photo images, multi-spectral images, satellite images, contextual data, common and expert knowledge) is used within decision support system. Result of the fusion is a thematic map which displays impacts of all indicators onto their environment and their mutual impacts onto each other. Such map is called a danger map, and it displays degree of danger around indicators within some specific MSA. This thematic image (danger map) is used as help to experts in humanitarian demining for better definition and reduction of MSA. Above mentioned systems were used in three locations in Croatia, and this article will present some results of this processing.

Minefield records on placing mines are the best source of information about presence, position, shape and contents of minefields. Experience showed that those information are not perfect, so, there is need for critical assessment of their value and applicability. In other words, type of imperfection has to be defined, and if possible, expressed quantitatively. This article will show some results of the statistical analysis of information quality in one set of minefield records within specific mined suspected area (MSA). Quality of information obtained from minefield record is important for production of danger map, since minefield records are the stronger indicator of mine presence on the scene with the highest level of danger. Because of the mentioned, it is very important to position a mine-field, and this can be done only based on good quality information from the minefield records.

A frequent and quite normal situation in the crisis management is an instance of uncertainty and a lack of amount, structure, time adequacy and quality of information and data, thus the decisions makers and reacting managers have to deal with this problem. In the technologic project TP-06/0007-01 was implemented Decision Support System (DSS) for the area reduction for mine action management on a strategic level through the use of data fusion methodological approach and the upgrade of the results from SMART project . Although the DSS was applied on the problems of the humanitarian mine, it has wider potential and could be applied in other crisis situations. Data fusion methodological approach is based on (1) the interpreted and fused multi-sensor imagery collected by aerial reconnaissance and surveillance, (2) enhanced and extracted indicators of mine presence and indicators of mine absence, (3) the formalized expert knowledge, (4) on the combined contextual data, (5) the spatial data provided by geographic information system (GIS) and (6) on the data contained in other information systems (e.g. Mine Information System (MIS) of the mine action centre (MAC)). The generic solutions from SMART have been integrated with data fusion methodological approach and implemented in the form of the DSS. SMART is intended to provide some of this additional information that would help in two ways: it can reinforce the suspicion of some places and reduce the suspected area on others. Output of the DSS at the final stage consist of three maps: (1) the map of mine danger measure, (2) confidence map of indicators of mine presence and indicators of mine absence, (3) the map of spatial stability of results. The methodological combination of three of them enables the validation or dismissal of areas proposed for reduction, thus they are suitable for practical use in mine action management of the humanitarian demining.

The key factors of the application of the airborne remote sensing technology for humanitarian demining are the ninefield indicators or the indicators of mine presence and indicator of mine absence and the analytical assessment of the contamination by the land mines. The basic simple concept of the minefield indicators was given in (1), were it was related to the ordinary physical features of the artificial and natural objects in the mine suspicious area. Similar concept was used in other works, while in (2) were introduced signatures of the minefields indicators and additional contextual features. The signatures depend on the sensors used for the remote sensing and this aspect was thoroughly considered in (2). The contextual features can be derived by the analytical assessment of the terrain contamination by the land mines, addition of the formalized knowledge enables efficient use of the indicators for the support of the decision making and for suspicious area reduction.

Some of the main problems in humanitarian demining are indeterminacy and lack of knowledge about current situation inside a mine suspected area (MSA). The borders of MSA are determined on the basis of information from mine information system (MIS: detected mines, mine incidents, interviews with population, minefield records, geodetic basis (maps, orthophoto maps), …). However, MSA in Croatia is a ‘’live matter’’ and changes in years after the conflict (people come back and use land in different ways). Because of that, the main idea of fusion of data, a priori information, contextual information and experts’ knowledge is analyze, interpret and combine all type of data about some MSA for getting information on current situation inside this particular MSA. These information are presented on danger maps [3] and support experts for humanitarian demining in the decision making about mine suspected area reduction or confirm suspicions in MSA [3]. In data fusion the images of airborne acquisition are used for detecting and enhancement of indicators of mine absence (IMA) [1] [4] and mine presence (IMP) [1] [4]. Furthermore, in data fusion a priori information, contextual information and experts’ knowledge are used for linking indicators with their impacts on environment and on each other within MSA. These impacts and levels of danger are shown on the danger maps [4]. The confidence map [4] is also an important product of data fusion, because it gives confidence (analytical assessment) of every statement on a danger map.

An airborne multisensor system was developed and realized in the frame of the project, its primary aim is to provide the airborne civil service of the reconnaissance and the surveillance in the crises' situations and the protection of environment. This is a multipurpose airborne remote sensing system. Besides its primary role it will serve as the laboratory for the research in different scientific disciplines too. The paper considers requirements, approaches and experience of the identification and the calibration of the system. There are two reasons that require the identification and the calibration of the system: first, to use full potential of applied sensors ; second, to provide set of the operational parameters that enable variety of the applications. The goal of the first case is to assess the relevant reference characteristics of used sensors in the ideal conditions (single sensor, no processing, no movement, a well defined scene, no transmission and the compression of the imagery). The available technical data for used sensors are not complete, only basic information is available, and the calibration and identification serve to provide the reference (ideal) characteristics of sensor. When implemented in the system, the characteristics of the particular sensors will be degraded ; the measure of the degradation will serve as a criterion for the evaluation of the design of the system. For the secon case (different applications of the system), it is necessary to identify and to calibrate the basic operational characteristics of the whole system and not only of the isolated sensor in the ideal conditions. When the basic operational characteristics of the airborne multisensor system were determined the estimation of their behavior is possible for changeable conditions. The considered airborne multisensor acquisition systwm form the following subsystems: a) passive digital electro optical sensor, b) acquisition hardware, c) acquisition software, d) synchronization subsystem, e) positioning and orientation subsystem, f) control of imaging angles for two sensors, g) radio transmission of the selected imagery from the aerial platform to the ground, h) subsystem for the navigation, i) moving map, j) subsystem for the automatic tracking of the aerial platform by GPS-GSM. The most important feature of the considered systemis its ability to parametrically geocode images oving to data collected by inertial measuring unit and GPS and the use of the software for the parametric geocoding (PARGE). In the system are applied following sensors: 1) hyperspectral line scaner V9 that can work in sampling or in the imaging mode ; 2) multispectral camera MS-3100 ; 3) longwave infrared (thermal) camera Photon, 4) color video camera Sony FCB-IX11AP ; 5) digital photo color camera Fuji FinePix S2. A hyperspectral scanner V9 covers up to forty-five channels and the multispectral camera MS-3100 covers four channels in the visible and near infrared wavelengths.Thermovision camera Photon works in the long-wave infrared range from 7.5 to 13.5 micrometers. Other two sensors cover red, gren and blue channels in the visible wavelengths. Calibration includes measurement of the modulation transfer function (MTF), ground resolving distance (GRD), radiometric response for used channels (except for V9), and spatial acuraccy. For the radiometric calibration is used target Spectralon SRT-MS-180 (LabSphere), for hyperspectral sensor is provided ground truth measurement by means of the FieldSpec 3 Spectroradiometer (ASD), for Photon are used data of the temperature collected by data loggers. Extensive calibration and identification was performed and the airborne multisensor system is now provided by a rich set of the characterustics.

Nastojanjem dobivanja boljeg uvida u min-sko sumnjivo područje (MSP) u Republici Hrvatskoj, u okviru projekta Europske komisije Airborne Minefield Area Reduction (ARC, EC IST-2000-25300, http://www.arc.vub.ac.be) formiran je digitalni zrakoplovni višesenzorski i višerazlučivi sustav za prikupljanje in-formacija (snimki) o minsko sumnjivoj površini. Sustav se sastoji od matrične kamere (vidljivi i blisko infra-crveni dio elektromagnetskog spektra, 0.4-1.1 µm), termalnog (termalni dio elektromagnetskog spektra, 8-12 µm) i hiperspektralnog linijskog skenera. Zbog specifične namjene sustava i objekata koje je trebalo otkriti na sceni, snimanja su provođena sa relativnih visina leta platforme (helikopter) od 130 do 900 m. Kako su relativne visine leta bile male, a MSP velika, broj snimki bio je također velik i pojavila se potreba za spajanjem snimki u veće cjeline kako bi se dobio uvid u cjelinu područja i interakciju utjecaja detek-tiranih objekata na sceni. Sustav nije raspolagao s modulom za automatsko mozaiciranje i geokodiranje pa se u tu svrhu prišlo ručnoj izradi mozaika, a nakon toga ručnom geokodiranju. Ti su postupci zatim omo-gućili klasifikaciju scene, kako bi se scena što bolje raščlanila i fuziju višesenzorskih i višerazlučivih slika kako bi se što više informacija moglo iz nje dobiti. U ovom radu pokazano je da je moguće provesti spomenute procese i uz pomoć ručnog mozaiciranja i geokodiranja., The digital airborne multisensor and multiresolution system for collection of information (images) about mine suspected area was created, within European commission project Airborne Minefield Area Reduction (ARC, EC IST-2000-25300, http://www.arc.vub.ac.be) to gain a better perspective in mine suspected areas (MSP) in the Republic of Croatia. The system consists of a matrix camera (visible and near infrared range of electromagnetic spectrum, 0.4-1.1 µm), thermal (thermal range of electromagnetic spectrum, 8-14 µm) and a hyperspectral linear scanner. Because of a specific purpose and seeking object on the scene, the flights for collecting the images took place at heights from 130 m to 900 m above the ground. The result of a small relative flight height and large MSPs was a large number of images which cover MSPs. Therefore, the need for merging images in largest parts, for a better perspective in whole MSPs and the interaction of detected object influences on the scene appeared. The mentioned system did not dispose of the module for automatic mosaicking and geocoding, so mosaicking and after that geocoding were done manually. This process made the classification of the scene (better distinguishing of objects on the scene) and fusion of multispectral and multiresolution images after that possible. Classification and image fusion can be even done by manually mosaicking and geocoding. This article demonstrated this claim.

Recent research in projects ARC and SMART (European Commission IST programme) shows that long wave thermal IR images enable detection of the objects that are otherwise not detectable. However, the important conclusion from this research is that the spatial resolution of available thermal IR cameras is not satisfactory. The possible way to compensate partially this deficiency can be data fusion. Due to the experience gained and the lessons learned in many years of the airborne remote sensing with thermal IR cameras we decided to analyze the fusion of the thermal IR images with images from visible and near IR wavelengths. The first goal was improvement of spatial resolution (sharpening) and this has been achieved. More important goal was the evaluation the radiometric quality of the fusion due to different physical background of the images. For this purpose many statistical parameters of different objects in the source thermal IR image were analysed and compared to the respective statistical parameters set for the same objects in the image obtained by fusion. The objects used in the analysis selected by use of the ground truth data that were available for the identification. The thermal IR contrast C between object “o” and background “b” used for comparison of the fused and source images too. The results of the analysis approved that the applied fusion method provides improvement of the spatial resolution and that thermal information of the analyzed objects conserved. Due to ap-proved quality of the fusion method, we recommend its application.

Airborne acquisition of the same part of terrain was done from different heights (500 and 900 meters), by digital matrix camera which collects data about landscape in four wavelengths: three visible channels (V: 0.4-0.5 µm, 0.5-0.6 µm, 0.6-0.7µm) and one near infrared channel (NIR: 0.7-0.9 µm). The set of images of the same terrain which are chosen consists of two RGB images (they are made of: green (G), red (R) and infrared (IR) = color infrared (CIR)) collected from: 500 m (higher space resolution) and 900 m (lower space resolution). With data fusion (used methods are: proncipal components, IHS and method of relative contribution) two products were made: RGB image with higher space resolution inside of the mentioned set of images (higher space resolution (from image collected from 500 m) was implemented in the image collected from 900 m) and RGB image with lower space resolution inside of the set of images (lower space resolution (from image collected from 900 m) was implemented in image collected from 500 m). This action produced two pairs of comparable images: the first one – the product of data fusion with higher spatial resolution and original image with higher spatial resolution (image collected from 500 m) ; and the second product of data fusion with lower spatial resolution and original image with lower resolution (image collected from 900 m). Namely, the main problem in the analysis of the quality of fused product is how to get reference image. In the majority cases that image needs to be produced, which means that both images are produced (result of fusion and reference image). On the contrary, in this work we provide the analysis of the qualitiy of fused products (on pixel level) on the existing original image basis. Those images are images collected from 500 and 900 m. Changes of images in both directions were monitored, regarding the improving and degradation of spatial resolution. The analysis of qualitiy of fused products was done through statistical parameters (mean, standard deviation, variance, corelation, scatter plot), contrast of images. At the end the existing proposition for protocol for assessment of the quality of fused product basis (Thomas& Wald, 2004) is discussed.

The potential contamination of the Adriatic Sea by the oil and other pollutants is a strategic threat to Croatia, due to the dependence of the national economy on the tourism. While the Adriatic is a narrow sea, available time for the early detection and the reaction is only several hours, therefore the aerial reconnaissance and surveillance is the most suitable solution. The Synthetic Aperture Radar (SAR) and ultra violet (UV) sensors are mostly used for detection of the oil spills, although they are not available at the time of the calamitous event, especially if the required reaction time is short. The hyper spectral remote sensing enables efficient detection of the oil spills, and this technology is promising and suitable for Croatia. In Croatia is available the hyper spectral line scanner, that was used in the airborne acquisition of the ground reflectivity. The airborne system was developed for the aerial sampling of the hyper spectral reflectivity data, while the platform was the helicopter Bell-206, in 2002 and 2003. The upgrading into the airborne imaging system for the detection of the oil spills and the geographical mapping is under way. The concept of this project is presented for the first time. For this project the support was required from Croatian Ministry of science, education and sports (technology program) and from the NATO (program Science for Peace). Besides the monitoring of the unexpected Adriatic Sea eco – accidents or catastrophes, the monitoring of the open fresh waters is foreseen too. In the frame of this project is foreseen the upgrading of the existing airborne hyperspectral line scanning system into the imaging system, development of the standard operating procedures (SOPs), training in office and on training on the job of the crews for the aerial reconnaissance and surveillance system based on piloted fixed wings airplane and/or the helicopter. The upgraded system will provide detection of the oil spill, the following the movement of the spill, enable approval of the cleaning, enable assessment of the damage of the beach, provide situational awareness in near real time. It is important to note that it will provide objective, accurate and reliable geographically mapped data about history of the contamination, aimed for legal procedures. In the paper we also present use of the V9 ImSpector hyper spectral line scanner for the visible and a near infrared wavelengths, that was aimed to sample reflectivity of the ground surface by the airborne remote sensing and the preliminary results of the research of the spectral reflectivity characteristics of the oil and sea water.

The current report, the Addendum to the D7.2, re - presents and clarifies parts of D7.2 and provides links to other deliverables where needed. The way how information, data, maps and facts were presented in D7.2 caused difficulties to reader and we try to improve this in the current document. In the same time we stress the fact that most of “unclear” items were given in previous deliverables

The main issue of the work was the measurement of the response of particular channels of the VNIR high resolution camera aimed for the development of the radiometric calibration. The analysed VNIR camera has several controlable parameters: gain, integration (exposition) time, bias, separately for near infrared (NIR) and red (R), channels and common for green and blue (G/B) channels. The control of these parameters is possible during the airborne acquisition and this provides very flexible imaging system that can be matched to the actual visibility above the mapped terrain. The measurements show that the channels of the VNIR camera have very different responses in NIR and R channels in comparison to G/B channels, if the gain was changed from minimum to maximum. The particulary powerfull feature of the considered digital VNIR camera is the available dynamic range, it can be 8 bits (from one to all four channels) or 10 bits (from one to three channels). Preliminary measurements and analysis show that 10 bits range enables robust mode of imaging during the changeable visibilty. The price for this is increase of the requirment to the memory of the acquisition system. The selection of the imaging parameters is not strightforward, it depends strongly on the features of the mapped scene. Recent analysis was performed with aim to enable production of the mosaics in visible (V) or in colour and the infrared bands (CIR). These mosaics serve as background for the subjective in-terpreetation of the shown scene, while outputs were landuse thematic map in CORINE methodology and specific features (vectors). Due to radiometric imperfection of the considered VNIR camera, the obtained VNIR images are radiometricaly distorted. The analysis of the influ-ence of the radiometric features of acquired images was performed on raw and on calibrated images, by analysis of the quality of the classification. The obtained results enable conclusion that the well calibrated digital VNIR camera can provide automated detection of objects in the scene and their thematic interpretation that was not possible on panchromatic digital ortho photo image map.

The information collected through the airborne remote sensing are useful for many scientific and practical works. In order to use collected information and existing knowledge of the observed object in the most efficient way for the purpose of solving specific problems, we have to examine their characteristics and if possible fuse them. This article is motivated by the existence of two operative digital sensor systems in Croatia that collect information at different wavelengths and with different methods of taking images. The first sensor is the digital matrix (staring) camera, for three visible channels (V: 0.4–0.5 μm, 0.5-0.6 μm, 0.6-0.7 μm), for near infrared channel (NIR: 0.7-1.0 μm) and the second sensor is the longwave thermal infrared (TIR: 8-14 μm) parallel scan camera. This paper will analyze the possibility of fusing information provided by above sensors, of one not very accessible area (mountainous terrain, with large differences of heights). Digital sensor systems yield multispectral images with different spatial and radiometric resolutions, geodetic and photogrammetric data (geographic maps, GPS data, ortophoto maps, geocoded mosaics), the insight in the field situation (ground truth) and other accessible sources provide inputs for fusion. It will also consider influence of the fusion on the results of classification of mosaics developed on multi-sensor, multi-spectral and multi-resolution digital images from above mentioned two sensor systems. The mosaics made on the basis of the above-mentioned sensor systems have been separately made, and later on the multi-spectral mosaics containing the information from all 5 channels will be created. Spatial and geometrical quality of multi-sensor and multi-spectral images and mosaics were analysed in earlier works, so this article will not deal with this matter, but will use the results of those earlier analyses. The work is an integral part of the research conducted in the scientific project ARC funded by the European Commission.

In the framework of SMART, danger maps are synthetic documents designed to help the end-users in their decision-making process regarding area reduction. They are created from the results of all detection and classification tools used in SMART, and other sources. They constitute the final output of the system.