Multi-mode adaptive fusion three-dimensional target detection method

Abstract

The invention discloses a multi-mode adaptive fusion three-dimensional target detection method, which is used for solving the technical problem of low detection efficiency of an existing three-dimensional target detection method. According to the technical scheme, the method comprise: inputting an RGB image and BEV Map; firstly, using an FPN network structure, comprising an encoder structure and adecoder structure, obtaining and using full-resolution feature maps of the FPN network structure and the encoder structure for being combined with bottom-layer detail information and high-layer semantic information, then extracting features corresponding to the two feature maps through feature clipping to be clipped and fused in a self-adaptive mode, and finally selecting 3D suggestions to achieve 3D object detection. The whole process is two-stage detection. In addition, the RGB image and the point cloud are used as original input, LIDAR FV input is reduced, the calculation amount is reduced, the calculation complexity of the algorithm is reduced, and the efficiency of three-dimensional space vehicle target detection is improved. According to the algorithm, the detection effect on smallobjects and the detection rate of shielded vehicles and intercepted vehicles are effectively improved.

Description

technical field [0001] The invention relates to a three-dimensional target detection method, in particular to a multi-mode self-adaptive fusion three-dimensional target detection method. Background technique [0002] Literature "X.Chen, H.Ma, J.Wan, B.Li, and T.Xia,"Multi-view 3d object detectionnetwork for autonomous driving,"in Proc.IEEE Conf.Conference on ComputerVision and Pattern Recognition,2017, pp.1907-1915." proposed a 3D object detection method based on RGB images and LIDAR point cloud information. The method aims to achieve high-precision 3D object detection in autonomous driving scenarios, and proposes a multi-view 3D network, a sensor fusion framework that takes lidar point clouds and RGB images as input and predicts oriented 3D bounding boxes. The network consists of two sub-networks for 3D object proposal generation and multi-view feature fusion. The 3D candidate boxes generated by the region proposal network can effectively represent the 3D point cloud from ...

AI Technical Summary

Problems solved by technology

[0003] In order to overcome the shortcomings of low detection efficiency of existing 3D target detection methods, the present invention provides a 3D target detection method based on multi-modal adaptive fusion

Images