Moving target detection system and method based on multi-frame point cloud

Abstract

The invention discloses a moving target detection system and method based on multi-frame point cloud. The system comprises a voxel feature extraction module which carries out the voxelization of a continuous frame point cloud sequence, and extracts a feature tensor sequence, a conversion module which performs matching fusion on the feature tensor sequence through a cross-modal attention module, fuses the first feature tensor and the second feature tensor, fuses the fusion result with the third feature tensor, fuses the fusion result with the fourth feature tensor, and so on to obtain the final fused feature tensor, a cross-modal attention module which is used for fusing the two feature tensors through a convolutional neural network according to an attention mechanism to obtain a fused feature tensor, and an identification module which is used for carrying out feature extraction on the finally fused feature tensor and outputting detection information of the target. The method comprises the following steps: S1 constructing each system module; S2 training the model through training set data; and S3 performing prediction through the trained model.

Description

technical field [0001] The invention relates to the technical field of three-dimensional object detection, in particular to a multi-frame point cloud-based moving object detection system and method. Background technique [0002] At this stage, the application of autonomous driving technology is becoming more and more extensive. Perception technology, especially point cloud-based 3D object detection technology, is one of the most important tasks in autonomous driving technology. At this stage, the 3D object detection technology based on point cloud with good effect includes the papers "Sparsely Embedded Convolutional Detection", "3D Object ProposalGeneration and Detection from Point Cloud" and the patent "A 3D Object Detection System Based on Laser Point Cloud and Its Detection Method", "A 3D Target Detection Method Based on Point Cloud", etc., but there are some problems in the above-mentioned prior art: First, the above-mentioned method does not consider continuous frame po...

AI Technical Summary

Problems solved by technology

At this stage, the 3D object detection technology based on point cloud with good effect includes the papers "Sparsely Embedded Convolutional Detection", "3D Object ProposalGeneration and Detection from Point Cloud" and the patent "A 3D Object Detection System Based on Laser Point Cloud and Its Detection Method", "A 3D Target Detection Method Based on Point Cloud", etc., but there are some problems in the above-mentioned prior art: First, the above-mentioned method does not consider continuous frame point cloud data, not only does not predict the target trajectory, but also affects the detection accuracy of the target ;Secondly, the above method completely depends on the inherent category of the training data set, that is, when there is a category in the actual scene that does not exist in the training set, the target will be missed.

Images