Obstacle Navigation Aid System Based On Machine Vision

Obstacle navigation assisted system plays an important role for the development of intelligent transportation as well as to color blindness and weakness. At the same time, it also one of the difficulties research for the unmanned vehicles. How to correct the assisted navigation is of important significance to the future development of intelligent transportation systems. At present, obstacle navigation aid system is not only has a high rate of failure in detection obstacles and navigation, but also it can’t satisfy the robustness requirements. Therefor this paper propose a simple obstacle navigation aid system based on binocular vision, which includes two parts: traffic location and recognition, the feasible region of road detection.In this system, the images by CCD camera collected must be nonlinear transferred from the color space of RGB into HSV first. For the identification of traffic lights, according to the H and V mutual independence, we can segment and extract the candidate regions from the image of HSV by using the different color values H and V, at the same time, we transfer the RGB image collected into gray-scale image, and then, we use the matter of gray-scale morphological and Hough transform to deal with it that can be located potential traffic lights, we merge two images by HSV extracted and conversion operation. By this way, we can detect and recognize the traffic lights.For road area of the feasible region detection, according to the compared with traditional algorithm and intelligent algorithm in experiments, we find some problems. We analyze this problems carefully and base on the characteristics of the road, we put forward method that combine intensity’s 2d image with Ostu threshold to identify feasible area of road. Experiments show this method can overcome the shadow, drop of water and eliminate virtual obstacles effect.According to simulation of the experiment on multiple images received, it shows that the detection accuracy of traffic lights at more than 95%, and also can correctly divided feasible region of road, this system can satisfy the real-time requirements.