Research on Clustering Algorithm Based on Aggregate View Object Detection

According to the existing three-dimensional object detection algorithms for fusion of point cloud and image information, after analysing different three-dimensional target detection algorithms, AVOD(Aggregate View Object Detection, AVOD) network is selected as the basic framework for the research and improvement of the fusion algorithm in this paper. The specific research content is as follows: for the problem of inaccurate anchor frame size of AVOD network, use the improved Canopy algorithm as the data preprocessing stage of K-means clustering algorithm, optimise the k-value selection of K-means algorithm, make the anchor frame generated after the final clustering more close to the real size of the object, and carry out the accuracy validation in the KITTI dataset, and finally improve the detection accuracy of the network.