A Novel On-Road Object Detection Approach Based on Vision

This paper proposed a simple and novel approach for on-road object detection based on vision. Firstly, a simple method is applied to detect the interest pixels of object in images by the defined interest pixels function and a single strategy is applied to reduce the redundant computation in the process of computation gray mean of pixels in squared window; Secondly, all the detected interest pixels neighbored each other are grouped to object interest region; after that, the non-interest pixels in are set to interest pixels and the images are binarized; The binary images could be easily segmented applying simple contour tracking algorithms. Lastly, several images captured from ITS or ALV system are used to test the proposed algorithm, the result indicate that the approach is valid and feasible.