Satadal Saha

Localization of License Plates from Surveillance Camera Images: A Color Feature Based ANN Approach

Traffic monitoring system has now become an essential administrative part in most of the developed and developing countries. In general, such systems monitor/identify the vehicles exceeding speed limits, or monitor the vehicles crossing the stop line at red traffic signal. It may also be used for registering the vehicles getting entry in a shopping mall or in a railway station or in an airport. The key modules of these monitoring systems are: (i) localization of license plates within the image and (ii) recognizing the license number using an OCR system. The present work addresses the first module of the system. The color information of the license plate is used as the knowledge base for training an artificial neural network system using back propagation algorithm. The trained network is then used to find the potential license plate region within a new traffic image. The scheme is applied in a real life outdoor environment at some road crossings in an Indian city. The result is found to be quite satisfactory giving an accuracy of around 80%.

Binarization of Document Images Using Hierarchical Histogram Equalization Technique with Linearly Merged Membership Function

Binarization itself is a process of finding a threshold value for converting a grey level image into a binary image. The threshold may vary depending on whether it is found globally or locally. It is found that either of the global and the local threshold itself can not provide a good binarization; rather a combination of the two is a better solution. In the current work, we have applied histogram equalization technique over the complete image and also over all the partitions of the image at different levels of hierarchy. A novel scheme is formulated for giving the membership value to each pixel at each level of hierarchy during histogram equalization. Then the image is binarized depending on the net membership value of each pixel. The technique outperforms when exhaustively tested on document images collected from different sources.

License Plate Localization Using Vertical Edge Map and Hough Transform Based Technique

Automatic License Plate Recognition (ALPR) system is the growing need by government authorities of different developing countries like India for traffic monitoring and control. The purpose of ALPR system is to track the vehicles violating the speed limit or violating the traffic signal at a road crossing. License plate localization is one of the key modules of any ALPR system. The objective of the current work is to localize the license plate of the vehicle from the vertical edge map of the image using statistical distribution of the vertical edges. Hough transform is then used to tune potential area to its actual dimension. In this work, real time vehicle images are captured from a road-side surveillance camera automatically throughout day and night in an unconstrained outdoor environment. The performance of the technique is tested and it provides 91.23% accuracy when compared with the ground truth data.

A Novel Scheme for Searching a Bangla Word within a Bangla Dictionary

Document image has been the area of research for a couple of decades because of its potential application in the area of text recognition, line recognition or any other shape recognition from the image. Text recognition from document image is very much dependent on the language of the text itself. English text recognition algorithms have already been developed and are standardized. Some works on Bangla text recognition has also been published in different literatures. Most of these recognition algorithms use character recognition as a base of text recognition. Recognition of text by recognizing its characters is a costly affair in terms of time and space. Our objective behind this work is to generate a Bangla dictionary and develop a novel technique of matching an input word with the dictionary word. The technique uses the features of the words as a whole rather than the features of each character. This reduces the time and space complexities of the recognition algorithm by many scales. We have tested our algorithm using both dictionary and non- dictionary words as the input words. For the dictionary words, it shows 100% accuracy in matching and for non-dictionary words, it shows 90% accuracy in non-matching. the features of the words themselves to match them as a whole. The whole work is divided in two phases: 1. Formation of Bangla dictionary 2. Automatic recognition of Bangla text A Bangla dictionary is constructed by collecting a huge number of maximum used words from corpuses and Bengali news papers. For the time being, the dictionary contains a collection of maximum used 20,000 Bengali words. The Bangla dictionary acts as a database and the developed technique searches for any dictionary word to find the closest match with the input word. In order to minimize the search complexity, the dictionary words are sorted in terms of some parameters of the words. By comparing different statistical parameters between the input word and the dictionary words the matching word from the dictionary can be found. When finding out different parameters of a word ratios of parameter property is used instead of actual property value. This is done to avoid property value variation with the variation of the physical values (font size, font type thickness etc.) of the words. Lot of works has been done in the field of Bangla character recognition from document images. It is found that matching a word from the dictionary by matching all of its characters is a costly affair. This has generated the idea of using the features of the word itself rather than finding the actual characters therein. In this proposed work, finer statistical parameters of the words are used to recognize it and if it is a dictionary word then a perfect match is obtained with an indexed dictionary word.

Text Localization in Camera Captured Images Using Adaptive Stroke Filter

Most of the text localization techniques are sensitive to text color, size, font and background clutter. They simply exploit the general segmentation rules or the prior knowledge about the text shape/size. As, inherently, a text consists of strokes of different sizes and orientations, so the concept of Stroke Filter is much more effective, particularly where text segmentation is taken into consideration. The problem with traditional stroke filter lies in its fixed width and is capable of segmenting strokes of predefined width. The proposed method uses adaptive stroke filter which can localize text regions, having varying stroke width, within camera captured images. The method is verified by experiment on a database containing 600 images.