Obstacle recognition for intelligent vehicle based on radar and vision fusion

Intelligent vehicle has elicited a significant amount of interest from both academe and industry. In order to solve the obstacles detection and recognition problems for intelligent vehicle, this paper puts forward a multi-objective detection method, which is used for the recognition of pedestrians and vehicles based on single-line laser radar and monocular vision fusion. Firstly, the laser radar is used to detect and get the location of the obstacles with the clustering algorithm. Secondly, according to the perspective transformation relationship between radar and vision, mapping the depth of the environment information to the image window, and determining the region of interest (ROI) through obstacle coordinates and coordinate mapping model. Afterwards, the histogram of oriented gradients feature is utilized to extract the feature vectors from ROI, the pre-established support vector machine classifier model is employed to identify and judge the type of obstacles (pedestrian, vehicle). Finally, through a variety of environmental conditions for experiment, the proposed method can detect and identify the obstacles effectively and exactly.