Target detection method based on vision, laser radar and millimeter wave radar

Abstract

The invention discloses a target detection method based on vision, a laser radar and a millimeter-wave radar. The method comprises the following steps: obtaining original data detected by a camera, the laser radar and the millimeter-wave radar, and carrying out time and space synchronization on the three types of data; resolving millimeter wave radar data according to a radar protocol; generatinga three-dimensional region of interest based on the position, the speed and the radar reflection area of the data calculated by the millimeter wave radar; extracting feature maps of a point cloud aerial view and a camera image; projecting the generated three-dimensional region of interest to a feature map of a point cloud aerial view of the laser radar and a feature map of the camera image; and fusing the feature maps with pre-selected boxes, and performing full connection layer processing on the fused image. Different sensor data are used for data fusion, so the generation number of candidateboxes is greatly reduced, the quality of the candidate boxes is improved, the recognition speed is increased, and the reliability of target detection in different environments is improved through sensor redundancy.

Description

technical field [0001] The invention relates to the technical field of target detection, in particular to a target detection method based on vision, laser radar and millimeter wave radar. Background technique [0002] As the country vigorously promotes the development of artificial intelligence and the rapid formation of the intelligent transportation industry, the fields of intelligent parking, Internet of Vehicles, and automatic driving are developing rapidly. Object detection technology has become an important research direction. Object detection is the most important and fundamental research field in the field of autonomous driving. The automatic driving of vehicles is inseparable from the accurate and fast detection of surrounding obstacles. In target detection, vehicles rely on sensors to obtain information around the vehicle, including lidar, millimeter-wave radar, ultrasonic radar, and visual cameras. Lidar has strong anti-interference ability, high resolution, an...

AI Technical Summary

Problems solved by technology

A large number of anchor points are generated in the visual camera detection stage, which consumes computing resources. The scale and size of the generated anchor points are fixed, and the accuracy of obstacles with different proportions such as large vehicles and pedestrians decreases.

In feature-level fusion, existing patents use millimeter-wave radar to generate candidate areas and process candidate areas. This method cannot accurately detect targets when millimeter-wave radars miss targets.

Images