Front vehicle detection based on multi-sensor fusion for autonomous vehicle

On account of the limitations of single sensor in obstacle detection, the paper investigates an obstacle detection method based on the fusion of 3D LiDAR and monocular visual. The spatial data fusion of the two sensors is realized according to their calibration results, and the time data fusion is realized by using double buffer technology. Considering the aspect ratio of vehicles, the image region of interest is determined based on the obstacle clustering of 3D LiDAR data. By using Haar-like features as effective characteristic of the front vehicle, integral figure is applied to extract Haar-like features of vehicle samples and non-vehicle samples. AdaBoost algorithm is used to choose weak classifiers to constitute strong classifiers, which combine into the cascade classifier. The cascade classifier has been trained to identify the vehicle target in the image region of interest. The relevant experimental results verify the effectiveness and real-time performance of the detection method.