Browsing by Author "Phalguni"

Filter results by typing the first few letters
Now showing 1 - 2 of 2
  • No Thumbnail Available
    Item
    A step towards development of aid for visually challenged
    (Springer India sanjiv.goswami@springer.co.in, 2017) Phalguni
    The work done towards developing a visual aid to help visually challenged people is described in this paper. An ultrasonic device is used for measuring the distance to a nearby object and SIFT algorithm based approach is used for object recognition from the captured image. Features of the recognized object as well as the distance to the object are converted to voice output which is available to the visually challenged person for navigation. The system developed and results obtained are described and suggestions are given for further work to be done to develop a better and more compact visual aid. © 2017, Indian Academy of Sciences.
  • No Thumbnail Available
    Item
    Design and implementation of an automatic traffic sign recognition system on TI OMAP-L138
    (2013) Phalguni; Ganapathi, K.; Madumbu, V.; Rajendran, R.; Sumam, David S.
    This paper discusses the design and processor implementation of a system that detects and recognizes traffic signs present in an image. Morphological operators, segmentation and contour detection are used for isolating the Regions of Interest (ROIs) from the input image, while five methods - Hu moment matching, histogram based matching, Histogram of Gradients based matching, Euclidean distance based matching and template matching are used for recognizing the traffic sign in the ROI. A classification system based on the shape of the sign is adopted. The performance of the various recognition methods is evaluated by comparing the number of clock cycles used to run the algorithm on the Texas Instruments TMS320C6748 processor. The use of multiple methods for recognizing the traffic signs allows for customization based on the performance of the methods for different datasets. The experiments show that the developed system is robust and well-suited for real-time applications and achieved recognition and classification accuracies of upto 90%. � 2013 IEEE.