Vehicle Recognition with Local-Feature Based Algorithm Using Parallel Vision Board

Abstract

This paper describes a robust method for recognizing vehicles. Our system is based on local-feature configuration, and we have already shown that it works very well in infrared images. The algorithm is based on our previous work, which is a generalization of the eigen-window method. This method has the following three advantages: (1) it can detect even if part of vehicles is occluded. (2) it can detect even if vehicles are translated due to running out of the lanes. (3) it does not require us to segment vehicle areas from input images. It is true that we have first developed our system with infrared images, but it is not essential for our system to employ infrared images. In this paper, applying our system on images of super wide-angle, we have shown that our system is effective to optical images, performing an outdoor experiment. Our system is good at detecting locations of vehicles, hence it will be useful for not only vehicle detection but also such application, ETC, DSRC or so, that system needs to know which vehicle it communicates with.