Abstract

A document image is composed of a variety of physical entities or regions such as text blocks, lines, words, figures, tables, and background. We could also assign functional or logical labels such as sentences, titles, captions, author names, and addresses to some of these regions. The process of document structure and layout analysis tries to decompose a given document image into its component regions and understand their functional roles and relationships. The processing is carried out in multiple steps, such as preprocessing, page decomposition, structure understanding, etc. We will look into each of these steps in detail in the following sections. Document images are often generated from ph

Abstract

One of the prime challenges in designing a classifier for large-class problems such as Indian language OCRs is the presence of a large similar looking character set. The nature of the character set introduces problems with accuracy and efficiency of the classifier. Hierarchical classifiers such as Binary Hierarchical Decision Trees (BHDTs) using SVMs as component classifiers have been effectively used to tackle such large-class classification problems. The accuracy and efficiency of a BHDT classifier will depend on: i) the accuracy of the component classifiers, ii) the separability of the clusters at each node in a hierarchical classifier, and iii) the balance of the BHDT. We propose methods to tackle each of the above problems in the case of binary character images. We present a new distance measure, which is intuitively suitable when Support Vector Machines are used as component classifiers. We also propose a novel method for balancing the BHDT to improve its efficiency, while maintaining the accuracy. Finally we propose a method to generate overlapping partitions to improve the accuracy of BHDTs. Comparison of the method with other forms of classifier combination techniques such as 1vs1, 1vsRest and Decision Directed Acyclic Graphs shows that the proposed approach is highly efficient, while being comparable with the more expensive techniques in terms of accuracy. The experiments are focused on the problem of Indian language OCR, while the framework is usable for other problems as well

Abstract

Accurate registration of images is essential for many computer vision algorithms for medical image analysis, super-resolution, and image mosaicing. Performance of traditional correspondence-based approaches is restricted by the reliability of the feature detector. Popular frequency domain approaches use the magnitude of global frequencies for registration, and are limited in the class of transformations that can be estimated. We propose the use of local phase information for accurate image registration as it is robust to noise and illumination conditions and the estimates are obtained at sub-pixel accuracy without any correspondence computation. We form an overdetermined system of equations from the phase differences to estimate the parameters of image registration. We demonstrate the effectiveness of the approach for affine transformation under Gaussian white noise and varying illumination conditions.

Abstract

Forensic document verification presents a different and interesting set of challenges as opposed to traditional writer identification and verification tasks using natural handwriting. The handwritten data presented to a forensic examiner is often deliberately altered, in addition to being limited in quantity. Specifically, the alterations can be either forged, where one imitates another person’s handwriting; or repudiated, where one deliberately distorts his handwriting in order to avoid identification. In this paper, we present a framework to detect repudiation in forensic documents, where we only have one pair of documents to arrive at a decision. The approach generates a statistically significant confidence score from matching two documents, which can be used to screen the documents that are passed on to an expert examiner. The approach can be extended for detection of forgeries as well.

Abstract

Post-processors are critical to the performance of language recognizers like OCRs, speech recognizers, etc. Dictionary-based post-processing commonly employ either an algorithmic approach or a statistical approach. Other linguistic features are not exploited for this purpose. The language analysis is also largely limited to the prose form.This paper proposes a framework to use the rich metric and formal structure of classical poetic forms in Indian languages for post-processing a recognizer like an OCR engine. We show that the structure present in the form of the vrtta and prasa ¯ can be efficiently used to disambiguate some cases that may be difficult for an OCR. The approach is efficient, and complementary to other post-processing approaches and can be used in conjunction with them.

Abstract

Local localization of a fully autonomous mobile robot in a partial map is an important aspect from the view point of accurate map building and safe path planning. The problem of correcting the location of a robot in a partial map worsens when sonar sensors are used. When a mobile robot is exploring the environment autonomously, it is rare to get the consistent pair of features or readings from two different positions using sonar sensors. So the approaches, which rely on readings or features matching, are prone to fail without exhaustive mathematical calculations of sonar modeling and environment modeling. This paper introduces link graph based robust two step feature chain based localization for achieving online SLAM (Simultaneous Localization And Mapping) using sonar data only. Instead of relying completely on matching of feature to feature or point to point, our approach finds possible associations between features to localize. The link graph based approach removes many false associations enhancing the SLAM process. We also map features onto Occupancy Grid (OG) framework taking advantage of its dense representation of the world. Combining features onto OG overcomes many of its limitations such as the independence assumption between cells and provides for better modeling of the sonar providing more accurate maps.

Abstract

This paper studies the problem of traversing a rough terrain by wheeled vehicles. The criterion for mobility of a wheeled vehicle in any terrain is formally developed, providing insights into the mechanical structure requirements of the vehicle. A vehicle structure with an actively articulated suspension is found as a solution to improved rough terrain mobility. The contact forces of the vehicle with the surface being traversed are identified as the critical factor in determining the traversability of the surface. Hence a control strategy involving the control of the contact forces (normal and traction) is proposed. The key feature of the locomotion strategy, thus, developed is that it provides a solution for mobility in terrain which cannot be traversed using a solution involving assumptions that ignore the dynamics of the main body of the vehicle.

Abstract

This paper presents a methodology for integrating features within the occupancy grid (OG) framework. The OG maps provide a dense representation of the environment. In particular they give information for every range measurement projected onto a grid. However independence assumptions between cells during updates as well as not considering sonar models lead to inconsistent maps, which may also lead the robot to take some decisions which may be unsafe or which may introduce an unnecessary overhead of run-time collision avoidance behaviors. Feature based maps provide more consistent representation by implicitly considering correlation between cells. But they are sparse due to sparseness of features in a typical environment. This paper provides a method for integrating feature based representations within the standard Bayesian framework of OG and provides a dense, more accurate and safe representation than standard OG methods.

Abstract

Different methodologies have been employed to solve the multi-sensor multi-target detection problem in a variety of scenarios. In this paper, we devise a time-step optimal algorithm for this problem when all but a few parameters of the sensor/target system are unknown. Using the concept of covering graph, we find an optimum solution for a single sensor, which is extended to multiple sensors by a tagging operation. Both covering graph and tagging are novel concepts, developed in the context of the detection problem for the first time, and bring a mathematical elegance to its solution. Furthermore, an implementation of the resulting algorithm is found to perform better than other notable approaches. The strong theoretical foundation, combined with the practical efficacy of the algorithm, makes it a very attractive solution to the problem

Abstract

Collision avoidance is one of the essential pillars of a wheeled robotic system. A wheeled mobile robot (called mobile robot for conciseness henceforth) requires for effective functioning an integrated system of modules for (i) map building,(ii) localization,(iii) exploration,(iv) planning and (v) collision avoidance. Often (i) and (ii) are entailed to be done simultaneously by robots resulting in a vast array of literature under the category SLAM, simultaneous localization and mapping. In this chapter we focus on the aspect of collision avoidance specifically between multiple robots, the remaining themes being too vast to even get a brief mention here. We present a cooperative conflict resolution strategy between multiple robots through a purely velocity control mechanism (where robots do not change their directions) or by a direction control method. The conflict here is in the sense of multiple robots competing for the same space over an overlapping time window. Conflicts occur as robots navigate from one location to another while performing a certain task. Both the control mechanisms attack the conflict resolution problem at three levels, namely (i) individual,(ii) mutual and (iii) tertiary levels. At the individual level a single robot strives to avoid its current conflict without anticipating the conflicting robot to cooperate. At the mutual level a pair of robots experiencing a conflict mutually cooperates to resolve it. We also denote this as mutually cooperative phase or simply cooperative phase succinctly. At tertiary level a set of robots cooperate to avoid one or more conflicts amidst them