Search

Childhood obesity is a global public health concern. Previous research, mainly conducted in developed countries, suggests that marketing and media exposure is associated with unhealthy eating behav...

This paper proposes a vehicle logo recognition (VLR) system centered on front-view cars, which has been largely neglected by vision community in comparison to other object recognition tasks. The study focuses on local features that describe structural characteristics by locating the logo of a car using a coarse-to-fine (CTF) strategy that first detects the bounding box of a car then the grille and at last, the logo. The detected logo is then used to recognize the make of a car in a reduced time. Our system starts to progress in detecting the bounding box of a car by means of a vocabulary voting and scale-adaptive mean-shift searching strategy. The system continues to process in locating the bounding box of an air-intake grille using a scale-adaptive sliding window searching technique. In the next level, the bounding box of a logo is located by means of cascaded classifiers and circular region detection techniques. The classification of vehicle logos is carried out on the patch-level as occurrences of similar visual words from a visual vocabulary, instead of representing the patchbased descriptors as bag-of-features and classifying them using a standard classifier. The proposed system was tested on 25 distinctive elliptical shapes of vehicle logos with 10 images per class. The system offers the advantage of accurate logo recognition of 86.3% in the presence of significant background clutter. The proposed scheme could be independently used for part recognition of grille detection and logo detection.

As a part of the vehicle identification system, the logo recognition, while matching with the license plate recognition, can be used to define the identity of the vehicle more accurately and provide reliable evidence for the deck car investigation, illegal escape and vehicle tracking. However, it is a difficult problem for the research to position the logos of different vehicles and the identification of the vehicles under low illumination conditions. This paper firstly uses the features of the color of the license plate to locate the license plate, and carries out the rough location of the logo according to the prior knowledge. Then, uses gray level, contrast enhancement, smoothing de-noising, edge detection and background suppression methods to deal with the coarse location of logo and realize the positioning of logo accurately. Next, extracts features of Vehicle-logo according seven HU invariant, considering the influence of low illumination conditions, this paper adds three HU invariant distances and establishes the characteristic library of the logo image. Thirdly, uses the support vector machine(SVM) to identify the logo and Cross validation(CV) methods to optimize the parameter C and g of SVM at the same time. In order to improve the recognition accuracy of the algorithm under low illumination conditions, the Grey Wolf Optimize (GWO) is used to further optimize the kernel function. Finally, takes 9 kinds of common Vehicle-logo as the logo to be identified, uses SVM to train 80% of the samples and test 20% of the samples. The results of experiments show that the increase of the invariant moments feature can obviously improve the accuracy of the logo, GWO is better than CV to improve the accuracy, and the average recognition rate is more than 92%, which effectively solve the problem of Vehicle-logo identification under low illumination conditions.

Logo recognition system deals with matching of the input trademark or logo with stored trademark images in database. This application, under CBIR umbrella, focuses on optimizing search through database by extracting minimum features from set of the images and using relevance feedback mechanism to identify the relevant images. Obtaining higher accuracy in retrieval process is the main challenge of the work. The retrieval results of CBIR system can be enhanced by using machine learning mechanisms with relevance feedback for Short Term Learning (STL) and Long-Term Learning (LTL). This paper proposes the relevance feedback system embedded with machine learning and optimization technique for logo recognition. Relevance feedback technique is used as baseline model for logo recognition. Feature set is optimized using particle swarm optimization (PSO) and search process is made intelligent by incorporating self-organizing map (SOM). These techniques improve the basic model as depicted in the results.

The Intelligent Transport System (ITS) not only extracts the features of vehicles via recognizing video signals, but also reduces the traffic jams, thereby, improve the vehicle pass ability of the highway entrance, and the efficiency of transportation. License plate recognition (LPR) is an important part of ITS. However, the license plate may be fake or be intentionally sheltered or smeared, which increases the difficulty of LPR. Considering that the logo information is difficult to change, we combine the logo information and license plate information to improve the reliability of vehicle identification. Based on this idea, a highway entrance vehicle logo recognition system is designed by using convolutional neural network(CNN) to detect and classify the vehicle logo in real time. We design a dataset named VL to train CNN and test the performance of our system.

A large amount of non-processed visual information is uploaded in social networks every day. Different features can be analyzed from the images such as objects, scenes, sentiments, people’s mood, color, etc. In this paper, we propose a novel method to detect, locate and classify logos in images, based on consensus. First, we present a basic logo recognition method. Second, an incremental learning algorithm is proposed to detect logos of any class by just using a synthetic image template, without the need of annotating a training set. Then, a crowdsourced solution (collaborative network) is generated within a VisualAD platform to carry out the consensus between several executions of the incremental learning method. The predictions will be the result of individual predictions from several users that improve the recognition. Finally, the principles enabling its Blockchain implementation are set and considerations on their extension to visual identity

Nowadays, information is provided through diverse network channels and, above all, its diffusion occurs in an always faster and pervasive manner. Social Media (SM) plays a crucial role in distributing, in an uncontrolled way, news, opinions, media contents and so on, and can basically contribute to spread information that sometimes are untrue and misleading. An integrated assessment of the trustworthiness of the information that is delivered is claimed from different sides: the Secure! project strictly fits in such a context. The project has been studying and developing a service oriented infrastructure which, by resorting at diverse technological tools based on image forensics, source reputation analysis, Twitter message trend analysis, web source retrieval and crawling, and so on, provides an integrated event assessment especially regarding crisis management. The aim of this paper is to present an interesting case-study which demonstrates the potentiality of the developed system to achieve a new integrated knowledge.

Most of vehicle have the similar structures and designs. It is extremely complicated and difficult to identify and classify vehicle brands based on their structure and shape. As we require a quick and reliable response, so vehicle logos are an alternative method of determining the type of a vehicle. In this paper, we propose a method for vehicle logo recognition based on feature selection method in a hybrid way. Vehicle logo images are first characterized by Histograms of Oriented Gradient descriptors and the final features vector are then applied feature selection method to reduce the irrelevant information. Moreover, we release a new benchmark dataset for vehicle logo recognition and retrieval task namely, VLR-40. The experimental results are evaluated on this database which show the efficiency of the proposed approach.

A logo is a special symbol or symbol that represents a company or organization. A logo can be a name, symbol or other graphic element that is displayed visually [1]. All institutions have a logo, especially an educational institution, namely the University, but the university logo is rarely remembered by lecturers, students, or the public. To be able to remind and provide information about the university, the university logo scan application which will later be directed to the university link is considered by the author suitable to be made. This kind of research has been tested with high accuracy but does not use the University logo. This research will utilize Augmented Reality program design with slight modifications and apply the Template Matching method that will be made with the Unity Game Engine and Vuforia SDK. This research was successful as evidenced by the smooth installation of APK and good program performance. In addition, this study obtained 84% accuracy in the detection of markers (logos) which were known by the distance and focus of the camera. This research also gets 100% accuracy in the direction to the intended university link. Keywords: Logo, University, Matching Templates, Vuforia SDK, Unity Game Engine

Vehicle logo recognition plays an important role in manufacturer identification and vehicle recognition. This paper proposes a new vehicle logo recognition algorithm. It has a hierarchical framework, which consists of two fusion levels. At the first level, a feature fusion model is employed to map the original features to a higher dimension feature space. In this space, the vehicle logos become more recognizable. At the second level, a weighted voting strategy is proposed to promote the accuracy and the robustness of the recognition results. To evaluate the performance of the proposed algorithm, extensive experiments are performed, which demonstrate that the proposed algorithm can achieve high recognition accuracy and work robustly.

We propose a method for localization and classification of brand logos in natural images. The system has to overcome multiple challenges such as perspective deformations, warping, variations of the shape and colors, occlusions, background variations. To deal with perspective variation, we rely on homography matching between the SIFT keypoints of logo instances of the same class. To address the changes in color, we construct a weighted graph of logo interconnections that is further analyzed to extract potentially multiple instances of the class. The main instance is built by grouping the keypoints of the graph connected logos onto the central image. The secondary instance is needed for color inverted logos and is obtained by inverting the orientation of the main instance. The constructed logo recognition system is tested on two databases (FlickrLogos-32 and BelgaLogos), outperforming state of the art with more than 10 % accuracy.

The paper presents a vehicle logo recognition system based on novel combination of tensor based feature extraction and ensemble of tensor subspace classifiers. Each originally two-dimensional vehicle logotype is transformed to a three-dimensional feature tensor applying the extended structural tensor method. All such exemplary logo-tensors which correspond to a single class are stacked to form a 4D logo-class-tensor. Decomposing each 4D logo-class-tensor into the orthogonal tensor subspace allows classification of unknown logotypes. The proposed system allows reliable vehicle logo recognition in real conditions as shown by experiments.

Logos sometimes also known as trademark have high importance in today’s marketing world. Logo or trademark is of high importance because it carries the goodwill of the company and the product. Logo matching and recognition is important to discover either improper or unauthorized use of logos. Query images may come with different types of scale, rotation, affine distortion, illumination noise, highly occluded noise. Sift descriptor, surf descriptor and hog descriptor are very good features to use among the existing techniques to recognize the logo images from such difficulties more accurately. General Terms Logo Recognition, invariant to scale, rotation, invariant to illumination noise, occluded objects .

In the field of deep learning and image recognition, to improve the accuracy of recognition, the neural model with a complex structure is usually selected as the training model. However, the model with a complex structure has the disadvantages of a large amount of calculation and time-consuming, which limits the ability of deep CNN to deploy on resource-limited devices like mobile phones. This paper presented a new logo recognition approach that is based on knowledge distillation, improving the recognition accuracy of a small model by knowledge transfer. At the same time, a bias neural network is introduced to increase the recognition accuracy of the target class. In this paper, we select ResNet-50 as the cumbersome network, ResNet-18 and VGG16 as small networks respectively. With only knowledge distillation, the average recognition accuracy of ResNet-18 and VGG16 have increased by 8% and 11% respectively. With the proposed bias neural network, the recognition accuracy of ResNet-18 and VGG16 further increased by 2%–10%. The recognition accuracy of the target class is within 5% of that of ResNet-50, which means the bias neural network with fewer layers and parameters is able to reach nearly the same recognition performance as the cumbersome network on target logo classes. And the experiments validate that the bias neural network can improve the accuracy of bias classes.

The increasing popularity of Social Networks makes change the way people interact. These interactions produce a huge amount of data and it opens the door to new strategies and marketing analysis. According to Instagram (https://instagram.com/press/) and Tumblr (https://www.tumblr.com/press), an average of 80 and 59 million photos respectively are published every day, and those pictures contain several implicit or explicit brand logos. The analysis and detection of logos in natural images can provide information about how widespread is a brand. In this paper, we propose a real-time brand logo recognition system, that outperforms all other state-of-the-art methods for the challenging FlickrLogos-32 dataset. We experimented with 5 different approaches, all based on the Single Shot MultiBox Detector (SSD). Our best results were achieved with the SSD 512 pretrained, where we outperform by 2.5% of F-score and by 7.4% of recall the best results on this dataset. Besides the higher accuracy, this approach is also relatively fast and can process with a single Nvidia Titan X 19 images per second.

Aselsan;et al.;Huawei;IEEE Signal Processing Society;IEEE Turkey Section;Netas, 26th IEEE Signal Processing and Communications Applications Conference, SIU 2018 -- 2 May 2018 through 5 May 2018 -- -- 137780, Machine vision based logo and trademark recognition, is one of the most efficient and widely used method to measure brand awareness on internet and social media. Similarity of logos geometric structure, difference pose and lighting conditions are the leading factors that makes the recognition task tedious. For this reason, different image descriptors have been used to extract the same information under various conditions. In this work, we examine fusion of image descriptors which obtained by extracting data from spectral and spatial domains independently. thereby features extracted from various domains targeted to form non-overlapping distinctive feature vectors. As spectral and spatial features we used GIST and FHOG descriptors. Experimental results held on the latest dataset Logos-32plus. Quantitative evaluation shows that our method have higher accuracy rates against the state of the art method

The amount of data produced every day on the internet increases every day and with the increasing popularity of the social networks the number of published photos are huge, and those pictures contain several implicit or explicit brand logos. Detecting this logos in natural images can provide information about how widespread is a brand, discover unwanted copyright distribution, analyze marketing campaigns, etc. In this paper, we propose a real-time brand logo recognition system that outperforms all other state-of-the-art in two different datasets. Our approach is based on the Single Shot MultiBox Detector (SSD), we explore this tool in a different domain and also experiment the impact of training with pretrained weights and the impact of warp transformations in the input images. We conducted our experiments in two datasets, the FlickrLogos-32 (FL32) and the Logos-32Plus (L32plus), which is an extension of the training set of the FL32. On the FL32, we outperform the state-of-the-art by 2.5% the F-score and by 7.4% the recall. For the L32plus, we surpass the state-of-the-art by 1.2% the F-score and by 3.8% the recall.

In order to achieve the significant vehicle logo recognition, a novel convolutional neural network(CNN) architecture is proposed. Considering the complexity of high dimensional data, we employ inception architecture and build a deep CNN network, which reduces the data dimensions and accelerates the computation of a large number of samples. We prepare a dataset to evaluate our algorithm, and obtain an overall accuracy of 99.02%. The comparison results show that our proposed algorithm outperforms linear support vector machine (SVM), LeNet-5, ImageNet and GoogLeNet in terms of accuracy improvement and computation-time reduction.

We propose a vehicle logo recognition method that uses SIFT representation and SVM classification. At the training phase, for each training example, a region of interest (ROI) containing the vehicle logo is extracted based on the vehicle plate location, SIFT features are extracted from the ROI, and keywords as well as their counts are obtained by clustering the SIFT features. For all the training examples, their keywords as well as the corresponding counts are used as input and their categories are used as output for training an SVM classifier. At the recognition stage, by a similar procedure of the training stage, for each test example, SIFT features of the ROI are extracted, and keywords as well as their counts are generated by clustering. These keywords as well as their counts are used as input to the SVM classifier and the category of the vehicle logo is obtained. The method is dependent on processing of a ROI rather than on accurate location of the vehicle logo. It uses little prior knowledge, and is easy to use. The method provides a satisfactory recognition rate, and thus is a feasible method for fusion of multiple classifiers.

In this paper, we propose a vehicle logo recognition algorithm based on multifeature fusion using a hierarchical classification approach, which can be applied at traffic checkpoints. First, a typical database of vehicle logos is set up based on surveillance images recorded at traffic checkpoints. Next, three features, HOG, Curvature histograms, and GIST, are extracted and three corresponding first level classifiers are trained using the support vector machine (SVM) algorithm. The probability that a certain test sample belongs to a certain kind can be obtained by predicting the sample with each level-1 classifier. All these probabilities are then concatenated and used as features for training a second-level SVM classifier. The resultant new classifier is used for classifying the vehicle logos of the test set. The experimental results show that the proposed approach to hierarchically integrate multiple features provides excellent accuracy for the vehicle logo recognition task.

Although good logos are essential for creating brand awareness and brand equity, the effects of logo design features have not been tested empirically. Extending previous findings regarding the effects of design complexity and exposure in advertising to the field of brand logos, two experiments tested the effects of logo complexity and exposure on brand recognition and brand attitude. It was hypothesized that logo complexity moderates the effects of exposure on logo recognition and brand attitudes, such that exposure increases recognition and positively impacts brand attitudes in particular for complex logos. Experiment 1 (N = 68) tested the effects of six unfamiliar logos on recognition (in milliseconds) in a 2 (logo design complexity: simple vs complex) × 2 (logo exposure: one vs four) mixed design. Experiment 2 (N = 164) tested the effects of eight familiar logos on logo recognition and brand attitudes in a 2 (complexity: simple vs complex) × 2 (logo exposure: well-established vs recently established) w...

With the explosive growth of commercial logos, high quality logo images are needed for training logo detection or recognition systems, especially for famous logos or new commercial brands. This paper focuses on automatic collecting representative logo images from the Internet without any human labeling or seed images. We propose multiple dictionary invariant sparse coding to solve this problem. This work can automatically provide prototypes, representative images, or weak labeled training images for logo detection, logo recognition, trademark infringement detection, brand protection, and ad-targeting. The experiment results show that our method increases the mean average precision for 25 types of logos to 80.07% whereas the original search engine results only have 32% representative logo images. The top images collected by our method are accurate and reliable enough for practical applications in the future.

According to the low recognition rate of Hu invariant moments in the target images, this article proposes a vehicle-logo recognition research algorithm based on the modified invariant moments. At first, use the template matching to locate the vehicle-logo rough area and use the edge detection for accurate location. Then, calculate the characteristic value of the modified invariant moments of the vehicle-logo, finally, the vehicle-logo is recognized according to the minimum distance of invariant moments. The experimental results show that the modified invariant moments can improve the recognition rate of target images effectively.

For most of the vehicle logo recognition algorithms, the logo is difficult to be pinpointed, and the recognition is roughly to be done in bad environment. Even though the recognition accuracy of convolution neural network (CNN) is relatively high, it also needs a large number of samples. This paper proposes a multi-scale parallel convolution neural network (multi-scale parallel CNN) to recognize vehicle-logo and improves the existing vehicle detection method. The multi-scale convolution kernel is used to extract features from original data in a parallel way. This method can keep high accuracy in the condition of illumination change and noise pollution, and can adapt to the harsh environment. Experimental results show that the classification accuracy of the method is as high as 98.80% on our owe dataset and 99.80% on the dataset used in other paper, which demonstrates strong generalization ability of our proposed algorithm.

The information available on-line and off-line, from open as well as from private sources, is growing at an exponential rate and places an increasing demand on the limited resources of Law Enforcement Agencies (LEAs). The absence of appropriate tools and techniques to collect, process, and analyze the volumes of complex and heterogeneous data has created a severe information overload. If a solution is not found, the impact on law enforcement will be dramatic, e.g. because important evidence is missed or the investigation time is too long. Furthermore, there is an uneven level of capabilities to deal with the large volumes of complex and heterogeneous data that come from multiple open and private sources at national level across the EU, which hinders cooperation and information sharing. Consequently, there is a pertinent need to develop tools, systems and processes which expedite online investigations. In this paper, we describe a suite of analysis tools to identify and localize generic concepts, instances of objects and logos in images, which constitutes a significant portion of everyday law enforcement data. We describe how incremental learning based on only a few examples and large-scale indexing are addressed in both concept detection and instance search. Our search technology allows querying of the database by visual examples and by keywords. Our tools are packaged in a Docker container to guarantee easy deployment on a system and our tools exploit possibilities provided by open source toolboxes, contributing to the technical autonomy of LEAs.

The Convolutional Neural Network of Deep Learning has been a huge success in the field of image intelligent transportation system can effectively solve the traffic safety, congestion, vehicle management and other problems of traffic in the city. Vehicle identification is a vital part of intelligent transportation, and the effective information in vehicles is of great significance to vehicle identification. With the traffic system on the vehicle identification technology requirements are getting higher and higher, the vehicle as an important type of vehicle information, because it should not be removed, difficult to change and other features for vehicle identification provides an important method. The current vehicle identification recognition (VLR) is mostly used to extract the characteristics of the method of classification, which for complex classification of its generalization ability to be some constraints, if the use of depth learning technology, you need a lot of training samples. In this paper, the...

In this paper we propose a method for logo recognition using deep learning. Our recognition pipeline is composed of a logo region proposal followed by a Convolutional Neural Network (CNN) specifically trained for logo classification, even if they are not precisely localized. Experiments are carried out on the FlickrLogos-32 database, and we evaluate the effect on recognition performance of synthetic versus real data augmentation, and image pre-processing. Moreover, we systematically investigate the benefits of different training choices such as class-balancing, sample-weighting and explicit modeling the background class (i.e. no-logo regions). Experimental results confirm the feasibility of the proposed method, that outperforms the methods in the state of the art., Preprint accepted in Neurocomputing

The information available on-line and off-line, from open as well as from private sources, is growing at an exponential rate and places an increasing demand on the limited resources of Law Enforcement Agencies (LEAs). The absence of appropriate tools and techniques to collect, process, and analyze the volumes of complex and heterogeneous data has created a severe information overload. If a solution is not found, the impact on law enforcement will be dramatic, e.g. because important evidence is missed or the investigation time is too long. Furthermore, there is an uneven level of capabilities to deal with the large volumes of complex and heterogeneous data that come from multiple open and private sources at national level across the EU, which hinders cooperation and information sharing. Consequently, there is a pertinent need to develop tools, systems and processes which expedite online investigations. In this paper, we describe a suite of analysis tools to identify and localize generic concepts, instances of objects and logos in images, which constitutes a significant portion of everyday law enforcement data. We describe how incremental learning based on only a few examples and large-scale indexing are addressed in both concept detection and instance search. Our search technology allows querying of the database by visual examples and by keywords. Our tools are packaged in a Docker container to guarantee easy deployment on a system and our tools exploit possibilities provided by open source toolboxes, contributing to the technical autonomy of LEAs.

Traditional image features combined with different classifiers are widely used in existing vehicle logo recognition methods, which didn't take into account the rich structure information of vehicle logos. Considering both their gray and structure information, a novel method based on foreground-background pixel pair FBPP feature, in which pixels are randomly sampled from foreground-background skeleton areas, is proposed. The pixel pair feature extraction process takes full consideration of vehicle logo structure, which makes this feature distinctive and discriminative. The experiment results show that, compared with methods based on features mainly focused on gray information, the method based on the proposed feature can achieve higher recognition performance. Especially under weak illumination, our method has shown strong robustness.

In our day today life if we want to buy any product then we first see the brand or logo of that product whether that brand or logo is original or not. Depending upon this we conclude that product is original product otherwise fake product. So in this project we detect first the original logo which is present in the image by comparing original logo image with the test image by using two algorithms first CDS-SIFT and second CDS RANSAC. And then conclude that the proposed method is more efficient than the existing method in terms of Execution time, Accuracy, FRR(False Rejection Rate), FAR (False Acceptance Rate). The proposed method of logo detection is based on Context Dependent Similarity (CDS) kernel. CDS kernel's function is dependent upon three terms first energy function, second context criterion and third entropy term. Energy function balances the fidelity term. The analysis of proposed method is done using MATLAB and comparative analysis of proposed method against the nearest neighbor of existing SIFT method is done and claim that proposed method is best method. Secondly we will evaluate the performance parameters.

Among all the various computer vision applications, automatic logo recognition has drawn great interest from industry as well as various academic institutions. In this paper, we propose an angle-distance map, which we used to develop a robust logo detection algorithm. The proposed angle-distance histogram is invariant against scale and rotation. The proposed method first used shape information and color characteristics to find the candidate regions and then applied the angle-distance histogram. Experiments show that the proposed method detected logos of various sizes and orientations.

Vehicle logo recognition (VLR) is a main issue in vehicle identification system. Logo recognition is still a challenge technique since VLR methods suffer from the large within-class variations due to the different illumination conditions, different viewpoints et al. In this paper, motivated by the excellent performance of the collaborative representation based classification (CRC), we formulate VLR problem under CRC scheme. It is noted that the performance of CRC is generally proportional to the size of the dictionary for better data representation capability. However, a large dictionary requires high computational cost. Aiming at maintaining the CRC performance but reducing the cost, the principal components analysis (PCA) is firstly employed on the class-dictionary to remove within-class information redundancy and noisy components. In addition, we introduce a new idea to code a data over a local dictionary instead of a global dictionary used in a conventional CRC, where the local dictionary is built by selecting the K-nearset neighbors of this data. As a result, a novel locally collaborative representation based classification with principal components (termed as LCRC_PC) method is systematically derived. The proposed LCRC_PC method is evaluated on two data sets. The average accuracies are 99.44% and 99.53% on a self-built data set and a public data set, respectively. Moreover, the computational cost of LCRC_PC is about a tenth of that of conventional CRC. Experimental results validate the effectiveness and robustness of our proposed LCRC_PC method.

This thesis presents a logo spotting framework applied to spotting logo images on document images and focused on document categorization and document retrieval problems. We also present three key-point matching methods: simple key-point matching with nearest neighbor, matching by 2-nearest neighbor matching rule method and matching by two local descriptors at different matching stages. The last two matching methods are improvements of the first method. In addition, using a density-based clustering method to group the matches in our proposed spotting framework can help not only segment the candidate logo region but also reject the incorrect matches as outliers. Moreover, to maximize the performance and to locate logos, an algorithm with two stages is proposed for geometric verification based on homography with RANSAC. Since key-point-based approaches assume costly approaches, we have also invested to optimize our proposed framework. The problems of text/graphics separation are studied. We propose a method for segmenting text and non-text in document images based on a set of powerful connected component features. We applied dimensionality reduction techniques to reduce the high dimensional vector of local descriptors and approximate nearest neighbor search algorithms to optimize our proposed framework. In addition, we have also conducted experiments for a document retrieval system on the text and non-text segmented documents and ANN algorithm. The results show that the computation time of the system decreases sharply by 56% while its accuracy decreases slightly by nearly 2.5%. Overall, we have proposed an effective and efficient approach for solving the problem of logo spotting in document images. We have designed our approach to be flexible for future improvements by us and by other researchers. We believe that our work could be considered as a step in the direction of solving the problem of complete analysis and understanding of document images., Cette thèse présente un framework pour le logo spotting appliqué à repérer les logos à partir de l’image des documents en se concentrant sur la catégorisation de documents et les problèmes de récupération de documents. Nous présentons également trois méthodes de matching par point clé : le point clé simple avec le plus proche voisin, le matching par règle des deux voisins les plus proches et le matching par deux descripteurs locaux à différents étapes de matching. Les deux derniers procédés sont des améliorations de la première méthode. En outre, utiliser la méthode de classification basée sur la densité pour regrouper les correspondances dans le framework proposé peut aider non seulement à segmenter la région candidate du logo mais également à rejeter les correspondances incorrectes comme des valeurs aberrantes. En outre, afin de maximiser la performance et de localiser les logos, un algorithme à deux étages a été proposé pour la vérification géométrique basée sur l’homographie avec RANSAC. Comme les approches fondées sur le point clé supposent des approches coûteuses, nous avons également investi dans l’optimisation de notre framework. Les problèmes de séparation de texte/graphique sont étudiés. Nous proposons une méthode de segmentation de texte et non-texte dans les images de documents basée sur un ensemble de fonctionnalités puissantes de composants connectés. Nous avons appliqué les techniques de réduction de dimensionnalité pour réduire le vecteur de descripteurs locaux de grande dimension et rapprocher les algorithmes de recherche du voisin le plus proche pour optimiser le framework. En outre, nous avons également mené des expériences pour un système de récupération de documents sur les documents texte et non-texte segmentés et l'algorithme ANN. Les résultats montrent que le temps de calcul du système diminue brusquement de 56% tandis que la précision diminue légèrement de près de 2,5%. Dans l'ensemble, nous avons proposé une approche efficace et efficiente pour résoudre le problème de spotting des logos dans les images de documents. Nous avons conçu notre approche pour être flexible pour des futures améliorations. Nous croyons que notre travail peut être considéré comme une étape sur la voie pour résoudre le problème de l’analyse complète et la compréhension des images de documents.

In this paper, we propose a Vehicle Logo Recognition (VLR) approach which uses moment invariant for feature extraction. Moment invariants and Minimum-Mean Distance (MMD) classifier are adopted to recognize six different types of vehicle logos from a public dataset. Vehicle logos obtained from coarse and fine segmentation, are recognized using Tchebichef and Legendre moment invariants. In either coarse or fine segmented vehicle logo images, Tchebichef moment invariants perform better than the Legendre's. With the experimental accuracy results of 88.3% on the 240 dataset images of six different types of vehicle logos, it has demonstrated the effectiveness of the proposed method in recognizing the fine segmented vehicle logo, which supports the use of the system for real application.

In this paper, a novel logo recognition algorithm based on a set of invariant features, which are calculated by using Radon transform and complex moments is proposed. This set of features is invariant to Rotation, Scaling, and Translation (RST) and it is also robust to additive noise. Radon transform is powerful tool for rotation, scaling, and translation properties which make it useful for our purpose. To obtain the RST invariant features, at first Radon transform is applied to logo image and then the complex moments are calculated from the radial and angular coordinates of Radon image. Logo recognition is carried out based on a similarity-based strategy by using Normalized Cross Correlation (NCC) measure. The proposed algorithm is evaluated on the UMD logo database (including 106 classes of logos). The experimental results validate the effectiveness of our algorithm in logo recognition and its robustness to additive noise.

The objective of this research is to develop computer software for recognizing a company logo. The system is called "Logo Recognition System or LRS". There are 2 parts of the LRS, namely: a client part and server part. The client part consists of common device, which is owned by the user, such as a mobile device, tablet and smart phone. On the client part, the LRS provides an easy graphic user interface for capturing a logo image. After that the device sends the logo image to the server part. The server part is a computer server, which does a process of recognizing with several algorithms and generates a result with a link of the logo's company website. The LRS consists of 4 components, namely: 1) Image Acquisition 2) Image Preprocessing 3) Image Recognition, and 4) Result Presentation. The system uses the python program to develop the logo recognition system on both client and server. The LRS can recognize the rotation image with any angle. The precision rate of the system is around 79.6 percent.

TV logo recognition plays an important role in video content understanding. So far, Real-time TV logo recognition under complex background based on single frame is still a very challenge task. By analyzing the characteristics of TV logo, a new TV logo recognition algorithm based on multiple feature fusion via hierarchical matching was proposed in this paper. The recognition process consisted of three stages. Firstly, a coarse matching based on frame differentiation and normalized cross correlation coefficient was employed to narrow down the candidate space significantly; then a fine matching based on HOG was used to describe the contours features of candidate logos; finally, the above features were fused to get better recognition performance. Experiment results showed the algorithm proposed had a remarkable performance in TV logo recognition under complex background based on single frame.