Development of a Fusing Module of the Images from Wide View Field and Stereo Cameras for an Autonomous Running Vehicle

Abstract

This study investigates a technique combining middle view field stereo cameras with a rotation mechanism and a wide view field camera to recognize running environment of both front and side of a vehicle. A fusing module of the images from the wide view field and stereo cameras is developed, where complicated data processes such as coordinate transformation, image fusing, etc. are executed on CPU-type processing units and primary processes of image data such as extraction of object regions from color images, labelling of extracted object regions, etc. are executed on CPLDs (Complex Programmable Logic Devices). Algorithms of labelling, compensation of the distortion occurred in wide view field images, and fusing the depth image from the stereo cameras into the image from the wide view field camera are implemented on the fusing module. Several experiments of image fusing for well-ordered indoor scenes show the applicability of the module to indoor situations.