Vehicle target detection method and system based on Leiyu semantic segmentation adaptive fusion

Abstract

The invention provides a vehicle target detection method and system based on Rayview semantic segmentation adaptive fusion, a semantic segmentation method is adopted to segment a camera depth map and radar point cloud data, and an adaptive Rayview information fusion method PC-ARVF based on reflection points and confidence is provided. And fusing the depth segmentation image and a point cloud segmentation result, reconstructing a fused point cloud, and ensuring supplement and fusion of multi-source data. A single-stage target detection model CDA-SSD based on a central point, a distance and an angle is provided, a vehicle surrounding frame is drawn by means of a cylindrical region, a loss function of target position regression is designed, and the complexity of the vehicle detection model is reduced. Compared with a previous vehicle target detection method, the method is higher in accuracy and lower in complexity, and is of great significance to fusion of radar and video images and vehicle target detection.

Description

technical field [0001] The invention relates to the general field of image data processing or generation; in particular, to the technical field of traffic and computer vision, and in particular to a vehicle target detection method and system based on Leivision semantic segmentation adaptive fusion. Background technique [0002] High-resolution video images provide vital data information for current vehicle object detection. Widely used in intelligent detection, automatic driving, driving safety and other fields. [0003] However, the information contained in video images is limited, and with the in-depth research of machine learning and hardware devices, many image-based object detection cannot meet people's requirements for the robustness of object detection algorithms. Among them, video image information is easily affected by environmental factors such as light and weather, which affects the accuracy of its detection. Millimeter-wave radars have the advantages of short r...

AI Technical Summary

Problems solved by technology

However, due to the characteristics of radar point cloud data, although the detection method of converting 2D video images into pseudo point cloud and merging the original point cloud data saves a large amount of data, both the processing of point cloud data and the detection of 3D objects will produce Huge amount of calculation, high complexity

Some researchers consider extracting the bird's-eye view, front view and other views of the radar point cloud, and realize target detection on the basis of quantizing the radar point cloud into a two-dimensional image. However, due to the sparse nature of point cloud data and the quantization The process will cause the loss of point cloud information, which will affect the accuracy of detection and lose the meaning of multi-source data fusion

Images