3D Object Detection Based on Graph Network Fusion Sampling Strategy

In the 3D target detection technology based on point cloud, there are problems like high cost of point cloud calculation and large gap between target scales, which lead to low target detection efficiency. In response, this paper proposes a 3D object detection algorithm based on fusion sampling and graph networks. Firstly, the point cloud fusion sampling technology is introduced to sample the original point cloud to reduce the computational complexity. Secondly, the K-NN algorithm is used to construct the graph of the sampled point cloud, and sub-image sampling is introduced to solve the problem of over-smooth graph convolution. Finally, the features of graph nodes are updated through feature interaction to improve the feature extraction ability of the network, thereby improving the target detection effect. This study conducted experiments on the KITTI3D dataset. Compared with the benchmark model Point-GNN, the detection accuracy of the car target was improved by 3. 89%. In the simple scene and the medium difficulty scene, the detection accuracy of the cyclist target is increased by 6. 60% and 4. 36% , respectively.