Search

Handwriting style of an individual will be usually varying from other individuals. The degree of complexity involved in recognition of a handwritten character of south Indian languages is high due to the presence of compound characters. Furthermore the complexity of interpreting handwritten text aggravates due to irregular handwriting styles. Building of datasets for training the database in such environments is tremendous and requires a lot of manual data processing and other preliminary tasks. The major contribution focuses on creation of morphable feature space for addressing handwritten recognition problem. In this paper, techniques useful for simplifying dataset preparation procedures along with a hybrid recognition approach for improvising recognition accuracies of Telugu handwritten characters are proposed. The techniques include text document segmentation, automatic cropping and pincushion and distance transformation based recognition approach. A feature vector is generated from the normalized Gabor features are extracted from pincushion and distance transform models of a character image and classified using Ada boost classifier with a recognition accuracy of 89%.

In this work, the evaluation of segmentation techniques towards the detection tumours or gliomas on BraTs 2018 data sets on T2 weighted 3D MRI images is carried out. Techniques experimented include pixel wise Linear Binary Pattern (LBP), region growing and Otsu thresholding to compare the efficiency of the same with proposed approach. In this work, 3D MRI images are initially subject to slicing from which about 10% from the top and bottom are normalized as those slices does not possess any significant details (empty slices). Further, the enhancement of tumour/gliomas regions is performed on normalized slices using high boost ‘convolution kernel processing in various trials with varying kernel sizes of high boost convolution kernel. The glioma regions enhanced using kernel are subject to local adaptive thresholding technique for extraction of gliomas from normalized slices. The detected gliomas are compared with the results of ground truth segmentation to determine the effectiveness of the proposed technique. It is observed in 80% of slices are successful in detection of gliomas both in case of Low Grade Glioma (LGG) and High Grade Glioma (HGG). By analyzing the results, it is observed that proposed method has outperformed with average Jaccard and Dice similarity recorded as 87.20% and 89.18% for HGG and where as in case of LGG it is found to be 83.77% and 87.74%.

Automated reading of handwritten Kannada documents is highly challenging due to the presence of vowels, consonants and its modifiers. The variable nature of handwriting styles aggravates the complexity of machine based reading of handwritten vowels and consonants. In this paper, our investigation is inclined towards design of a deep convolution network with capsule and routing layers to efficiently recognize Kannada handwritten characters. Capsule network architecture is built of an input layer, two convolution layers, primary capsule, routing capsule layers followed by tri-level dense convolution layer and an output layer. For experimentation, datasets are collected from more than 100 users for creation of training data samples of about 7769 comprising of 49 classes. Test samples of all the 49 classes are again collected separately from 3 to 5 users creating a total of 245 samples for novel patterns. It is inferred from performance evaluation; a loss of 0.66% is obtained in the classification process and for 43 classes precision of 100% is achieved with an accuracy of 99%. An average accuracy of 95% is achieved for all remaining 6 classes with an average precision of 89%.

The proposed research aims to restore deteriorated text sections that are affected by stain markings, ink seepages and document ageing in ancient document photographs, as these challenges confront document enhancement. A tri-level semi-adaptive thresholding technique is developed in this paper to overcome the issues. The primary focus, however, is on removing deteriorations that obscure text sections. The proposed algorithm includes three levels of degradation removal as well as pre- and post-enhancement processes. In level-wise degradation removal, a global thresholding approach is used, whereas, pseudo-colouring uses local thresholding procedures. Experiments on palm leaf and DIBCO document photos reveal a decent performance in removing ink/oil stains whilst retaining obscured text sections. In DIBCO and palm leaf datasets, our system also showed its efficacy in removing common deteriorations such as uneven illumination, show throughs, discolouration and writing marks. The proposed technique directly correlates to other thresholding-based benchmark techniques producing average F-measure and precision of 65.73 and 93% towards DIBCO datasets and 55.24 and 94% towards palm leaf datasets. Subjective analysis shows the robustness of proposed model towards the removal of stains degradations with a qualitative score of 3 towards 45% of samples indicating degradation removal with fairly readable text.

Object segmentation and classification in the video sequence is a classical and critical problem that is constantly addressed due to vast real-time applications such as autonomous vehicles, smart surveillance, etc. Segmentation and classification of moving objects in video sequences captured from real-time non-constraint sequences is a challenging task. This paper presents a moving object segmentation and classification method for video sequences captured in a real-time environment. The key contributions of the paper include a method to detect and segment motion regions by applying the non-parametric Kolmogorov–Smirnov statistical test in the Spatio-temporal domain and a probabilistic neural network-based classification method to classify the moving objects into various classes. Promising results are obtained by experimentation using challenging PETS and Change Detection datasets. To corroborate the efficacy, a comparative analysis with contemporary method is also performed.

Achieving high accuracies in recognition of handwritten text is a challenging research problem and never exhausting. The factors that instill challenges in handwritten character recognition include high degree of variability in writing, script type and the type of documents etc. In this paper, we focus on recognition of handwritten Telugu text commonly found in document images. The character set include all the vowels, consonants and single level vowel consonant clusters chosen in accordance with the commonly used terminology employed in composition pre-printed documents such as admission forms. In this paper, an algorithm is devised that performs zone-based feature extraction of the segmented character images. In the proposed work, the Gabor features are extracted from the character image at zone level and its efficiency is evaluated individually on two different zone representations and entire image at various scales and orientations. The classification and recognition performance is analyzed using nearest neighbor classifier, Naive Bayesian, multi-class SVM and probabilistic neural networks classifiers. The efficiency of the classifiers are also tested with statistical, Histogram of Gradients and Hu moments‟ feature extraction methods and the best accuracy of the system is found to be 84.8% for Gabor features with zone representation 2 and with multi-class SVM classifier.

Enhancement of document images is an interesting research challenge in the process of character recognition. It is quite significant to have a document with uniform illumination gradient to achieve higher recognition accuracies through a document processing system like Optical Character Recognition (OCR). Complex document images are one of the varied image categories that are difficult to process compared to other types of images. It is the quality of document that decides the precision of a character recognition system. Hence transforming the complex document images to a uniform illumination gradient is foreseen. In the proposed research, ancient document images of UMIACS Tobacco 800 database are considered for removal of marginal noise. The proposed technique carries out the block wise interpretation of document contents to remove the marginal noise that is present usually at the borders of images. Further, Hu moment’s features are computed for the detection of marginal noise in every block. An empirical analysis is carried out for classification of blocks into noisy or non-noisy and the outcomes produced by algorithm are satisfactory and feasible for subsequent analysis.

Starch, composed of amylose and amylopectin, greatly influences rice cooking and textural quality, which in turn is controlled by various isoforms of several enzymes. Activity of one or more isoforms of starch-synthesizing enzymes results in various forms of starch structure based on the amylopectin chain length and average external, internal and core chain length distribution and hence results in varying physicochemical and cooking quality. Since the synthesis of starch is highly complex, it is crucial but essential to understand its biosynthetic pathway, starch structure and effects on the physicochemical properties that control eating and cooking quality, and alongside conduct research on gene/QTL mapping for use in marker-assisted selection (MAS) with a view to improve and select cultivars with most desirable range and class of rice starch properties. This article presents the updates on current understanding of the coordination among various enzymes/isoforms towards rice starch synthesis in endosperm and their effect on rice grain physicochemical, cooking and eating qualities. The efforts in identifying regions responsible for these enzymes by mapping the gene/QTLs have provided a glimpse on their association with physicochemical and cooking properties of rice and, hence, improvement is possible by modifying the allelic pattern, resulting in down or nil regulation of a particular enzyme. The clear understanding of the tissue specific coordination between enzyme isoforms and their subsequent effect in controlling eating and cooking properties will enhance the chances to manipulate them for getting desired range of amylose content (AC) and gelatinization temperature (GT) in improved cultivars through combining desired alleles through MAS.

Bacterial blight (BB) is a major production threat to Basmati, the aromatic rice prized for its unique quality. In order to improve the BB resistance of two elite, traditional BB-susceptible Basmati varieties (Taraori Basmati and Basmati 386), we utilized the strategy of limited marker-assisted backcrossing for introgression of two major BB resistance genes, Xa21 and xa13, coupled with phenotype-based selection for improvement of their plant type and yield. Improved Samba Mahsuri, an elite high-yielding, fine-grain-type BB-resistant rice variety served as donor for BB resistance. Backcross-derived improved Basmati lines at BC1F5 possessing a single resistance gene (i.e. either Xa21 or xa13) displayed moderate resistance to BB, while lines possessing both Xa21 and xa13 showed significantly higher levels of resistance. Two-gene pyramid lines (Xa21 + xa13) possessing good grain and cooking quality similar to their respective traditional Basmati parents, short plant stature (

The appearance and cooking quality of rice determine its acceptability and price to a large extent. Quantitative trait loci (QTLs) for 12 grain quality traits were mapped in 2 mapping populations derived from Oryza sativa cv Swarna × O. nivara. The BC(2)F(2) population of the cross Swarna × O. nivara IRGC81848 (population 1) was evaluated during 2005 and that from Swarna × O. nivara IRGC81832 (population 2) was evaluated during 2006. Linkage maps were constructed using 100 simple sequence repeat (SSR) markers in population 1 and 75 SSR markers in population 2. In all, 21 QTLs were identified in population 1 (43% from O. nivara) and 37 in population 2 (38% QTLs from O. nivara). The location of O. nivara-derived QTLs mp1.2 for milling percent, kw6.1 for kernel width, and klac12.1 for kernel length after cooking coincided in the 2 populations and appear to be useful for Marker Assisted Selection (MAS). Four QTLs for milling percent, 1 QTL each for amylose content, water uptake, elongation ratio, 2 QTLs for kernel width, and 3 QTLs for gel consistency, each explained more than 20% phenotypic variance. Three QTL clusters for grain quality traits were close to the genes/QTLs for shattering and seed dormancy. QTLs for 4 quality traits were associated with 5 of the 7 major yield QTLs reported in the same 2 mapping populations. Useful introgression lines have been developed for several agronomic traits. It emerges that 40% O. nivara alleles were trait enhancing in both populations, and QTLs for grain quality overlapped with yield meta-QTLs and QTLs for dormancy and seed shattering.

Broadening of the genetic base for identification and transfer of genes for resistance to insect pests and diseases from wild relatives of rice is an important strategy in resistance breeding programs across the world. An accession of Oryza nivara, International Rice Germplasm Collection (IRGC) accession number 105710, was identified to exhibit high level and broad-spectrum resistance to Xanthomonas oryzae pv. oryzae. In order to study the genetics of resistance and to tag and map the resistance gene or genes present in IRGC 105710, it was crossed with the bacterial blight (BB)-susceptible varieties ‘TN1’ and ‘Samba Mahsuri’ (SM) and then backcrossed to generate backcross mapping populations. Analysis of these populations and their progeny testing revealed that a single dominant gene controls resistance in IRGC 105710. The BC1F2 population derived from the cross IRGC 105710/TN1//TN1 was screened with a set of 72 polymorphic simple-sequence repeat (SSR) markers distributed across the rice genome and the resistance gene was coarse mapped on chromosome 7 between the SSR markers RM5711 and RM6728 at a genetic distance of 17.0 and 19.3 centimorgans (cM), respectively. After analysis involving 49 SSR markers located between the genomic interval spanned by RM5711 and RM6728, and BC2F2 population consisting of 2,011 individuals derived from the cross IRGC 105710/TN1//TN1, the gene was fine mapped between two SSR markers (RMWR7.1 and RMWR7.6) located at a genetic distance of 0.9 and 1.2 cM, respectively, from the gene and flanking it. The linkage distances were validated in a BC1F2 mapping population derived from the cross IRGC 105710/SM//2 × SM. The BB resistance gene present in the O. nivara accession was identified to be novel based on its unique map location on chromosome 7 and wider spectrum of BB resistance; this gene has been named Xa33. The genomic region between the two closely flanking SSR markers was in silico analyzed for putatively expressed candidate genes. In total, eight genes were identified in the region and a putative gene encoding serinethreonine kinase appears to be a candidate for the Xa33 gene.

Realization of high accuracies towards south Indian character recognition is one the truly interesting research challenge. In this paper, our investigation is focused on recognition of one of the most widely used south Indian script called Kannada. In particular, the proposed exper-iment is subject towards the recognition of degraded character images which are extracted from the ancient Kannada poetry documents and also on the handwritten character images that are collected from various unconstrained environments. The character images in the degraded documents are slightly blurry as a result of which character image is imposed by a kind of broken and messy appearances, this particular aspect leads to various conflicting behaviors of the recognition algorithm which in turn reduces the accuracy of recognition. The training of degraded patterns of character image samples are carried out by using one of the deep convolution neural networks known as Alex net.The performance evaluation of this experimentation is subject towards the handwritten datasets gathered synthetically from users of age groups between 18-21, 22-25 and 26-30 and also printed datasets which are extracted from ancient document images of Kannada poetry/literature. The datasets are comprised of around 497 classes. 428 classes include consonants, vowels, simple compound characters and complex com-pound characters. Each base character combined with consonant/vowel modifiers in handwritten text with overlapping/touching diacritics are assumed as a separate class in Kannada script for our experimentation. However, for those compound characters that are non-overlapping/touching are still considered as individual classes for which the semantic analysis is carried out during the post processing stage of OCR. It is observed that the performance of the Alex net in classification of printed character samples is reported as 91.3% and with reference to handwritten text, and accuracy of 92% is recorded.

Classification of videos based on its content is one of the challenging and significant research problems. In this paper, a simple and efficient model is proposed for classification of sports videos using deep learned convolution neural networks. In the proposed research, the gray scale variants of image frames are employed for classification process through convolution technique at varied levels of abstraction by adapting it through a sequence of hidden layers. The image frames considered for classification are obtained after the duplicate frame elimination and each frame is further rescaled to dimension 120x240. The sports videos categories used for experimentation include badminton, football, cricket and tennis which are downloaded from various sources of google and YouTube. The classification in the proposed method is performed with Deep Convolution Neural Networks (DCNN) with around 20 filters each of size 5x5 with around stride length of2 and its outcomes are compared with Local Binary Patterns (LBP), Bag of Words Features (BWF) technique. The SURF features are extracted from the BWF technique and further 80% of strongest feature points are employed for clustering the image frames using K-Means clustering technique with an average accuracy achieved of about 87% in classification. The LBF technique had produced an average accuracy of 73% in differentiating one image frame to other whereas the DCNN had shown a promising outcome with accuracy of about 91% in case of 40% training and 60% test datasets, 99% accuracy in case of 60% training an 40% test datasets. The results depict that the proposed method outperforms the image processing-based techniques LBP and BWF.

Fragrance or aroma in rice is considered as a special trait with huge economic importance that determines the premium price in global trade. With the availability of molecular maps and genome sequences, a major gene for fragrance (badh2) was identified on chromosome 8. An 8-bp deletion in the exon 7 of this gene was reported to result in truncation of betaine aldehyde dehydrogenease enzyme whose loss-of-function lead to the accumulation of a major aromatic compound, 2-acetyl 1-pyrroline (2AP) in fragrant rice. However, several studies have reported exceptions to this mutation and indicated the involvement of other genetic loci in controlling fragrance trait. These studies emphasize the need to characterize the fragrance and its underlying factors in a wide range of genetic resources available for this trait. This review summarizes the new insights gained on the genetic and molecular understanding of fragrance in rice.

Samba Mahsuri (BPT5204) is a medium slender grain indica rice variety that is very popular with farmers and consumers across India because of its high yield and excellent cooking quality. However, the variety is susceptible to several diseases and pests, including bacterial blight (BB). We have used PCR based molecular markers in a backcross-breeding program to introgress three major BB resistance genes (Xa21, xa13 and xa5) into Samba Mahsuri from a donor line (SS1113) in which all the three genes are present in a homozygous condition. At each backcross generation, markers closely linked to the three genes were used to select plants possessing these resistance genes (foreground selection) and microsatellite markers polymorphic between donor and recurrent parent were used to select plants that have maximum contribution from the recurrent parent genome (background selection). A selected BC4F1 plant was selfed to generate homozygous BC4F2 plants with different combinations of BB resistance genes. The three-gene pyramid and two-gene pyramid lines exhibited high levels of resistance against the BB pathogen. Under conditions of BB infection, the three-gene pyramid lines exhibited a significant yield advantage over Samba Mahsuri. Most importantly, these lines retain the excellent grain and cooking qualities of Samba Mahsuri without compromising the yield as determined in multi-location trials. This work demonstrates the successful application of marker-assisted selection for targeted introgression of multiple resistance genes into a premium quality rice variety.

Detection of text orientation in document images is of preliminary concern prior to processing of documents by Optical Character Reader. The text direction in document images should exist generally in a specific orientation, i.e., text direction for any automated document reading system. The flipped text orientation leads to an unambiguous result in such fully automated systems. In this paper, we focus on development of text orientation direction detection module which can be incorporated as the perquisite process in automatic reading system. Orientation direction detection of text is performed through employing directional gradient features of document image and adapts an unsupervised learning approach for detection of flipped text orientation at which the document has been originally fed into scanning device. The unsupervised learning is built on the directional gradient features of text of document based on four possible different orientations. The algorithm is experimented on document samples of printed plain English text as well as filled in pre-printed forms of Telugu script. The outcome attained by algorithm proves to be consistent and adequate with an average accuracy around 94%.

Realization of high accuracies and efficiencies in South Indian character recognition systems is one of the principle goals to be attempted time after time so as to promote the usage of optical character recognition (OCR) for South Indian languages like Telugu. The process of character recognition comprises pre-processing, segmentation, feature extraction, classification and recognition. The feature extraction stage is meant for uniquely recognizing each character image for the purpose of classifying it. The selection of a feature extraction algorithm is very critical and important for any image processing application and mostly of the times it is directly proportional to the type of the image objects that we have to identify. For optical technologies like South Indian OCR, the feature extraction technique plays a very vital role in accuracy of recognition due to the huge character sets. In this work we mainly focus on evaluating the performance of various feature extraction techniques with respect to Telugu character recognition systems and analyze its efficiencies and accuracies in recognition of Telugu character set.

Automatic detection, extraction and recognition of vehicle number plate region in traffic control systems is one of the prominent application in Computer vision. The drastic increase in number of vehicles in the current generation greatly increases the complexity in tracking the vehicles through the human visual system, manual procedure of controlling traffic and enforcement of various laws and rules is not sufficient for smooth control of traffic. This urges the need for development of technology that can automate this process. This paper mainly focuses on the development of an automatic number plate extraction and recognition algorithm by incorporating constructs like edge detection, horizontal and vertical edge processing using fixed threshold technique. The extracted number plate region is again processed using template matching algorithm for the recognition of the characters embossed on the number plate with respect to every individual piece of number plate. The algorithm developed has achieved an accuracy of around 100% and works for both front and rear images of the car.

To assess the efficiency of different molecular markers which are linked to fertility restorer genes Rf3 and Rf4 of WA-CMS system, 103 breeding lines with no information on fertility restoration were screened with the help of molecular markers linked to major fertility genes Rf3 and Rf4. The breeding lines were crossed with APMS6A and the F1s were evaluated for pollen and spikelet fertility to identify restorers and maintainers.The SSR primer RM6100 linked to Rf4 gene of chromosome 10 and RM10313 linked to Rf3 gene of chromosome 1 showed eighty five and eight one percentage efficiency respectively in identifying restorer lines. Therefore, these markers are useful tool for evaluating large number of breeding lines to know about their fertility restoration in a short period of time without generating and evaluating large number of test crosses.The potential restorers may be identified with hundred percentage efficiency based on molecular screening itself, if candidate genes based markers are developed and validated for both Rf4 and Rf3 genes.