Recognizing vehicle in infra-red images using IMAP parallel vision board

Abstract

This paper describes a method to recognize vehicles, in particular to recognize which make it is and which type it is. Our system employs infra-red images so that we can use the same algorithm in day time and at night. The algorithm is based on the vector-quantaization, originally proposed by Krumm, and is implemented on IMA P parallel image-processing board. Our system makes the compressed database of local features, for the algorithm, of a target vehicles from given training images in advance, and then matchs a set of local features in the input image with those in training images for recognition. This method has the following three advantages: (1) it can detect even if part of the target vehicles is occluded.(2) it can detect even if the target vehicles is translated due to running out of the lanes. (3) we do not need to segment a vehicle part from input images.Through outdoor experiments, we have confirmed these advantages.

Local-feature based vehicle class recognition in infra-red images using IMAP parallel vision board

“…We adapt the similarity criterion first and the uniqueness criterion later. We also apply the vector quantization method to implement our algorithm on the IMAP-vision board [5]. Similarity is the criterion that selected windows are common in at least two training images.…”