VEHICLE RECOGNITION IN INFRA-RED IMAGES USING PARALLEL VISION BOARD
Matasaka Kagesawa, S. Ueno, Katsushi Ikeuchi, H. Kashiwagi · PROCEEDINGS OF 6TH WORLD CONGRESS ON INTELLIGENT TRANSPORT SYSTEMS (ITS), HELD TORONTO, CANADA, NOVEMBER 8-12, 1999 · 1999
This paper describes a method for vehicle recognition, in particular, for recognizing a vehicle's make and model. The system employs infra-red images so that the same algorithm can be used both day and night, The algorithm, based on the vector-quantization, was originally proposed by Krumm, and is implemented on an IMAP parallel image-processing board. The system makes the compressed database of local features, for the algorithm, of a target vehicle from given training images in advance; the system then matches a set of local features in the input image with those in training images for recognition. This method has the following advantages: it can detect even if part of the target vehicle is occluded; it can detect even if the target vehicle is translated due to running out of the lanes; and it does not require segmentation of vehicle part from input images. The above advantages have been confirmed by performing outdoor experiments. (A*) For the covering abstract see ITRD E110327.