Night preceding vehicle detection method based on millimeter-wave radar and machine vision

Abstract

The invention relates to a night preceding vehicle detection method based on a millimeter-wave radar and machine vision, which belongs to the field of vehicle control. The night preceding vehicle detection method comprises the following steps: (1) performing camera calibration, so as to obtain a projection matrix of world coordinates to image pixel coordinates, establishing a transition relation between a radar coordinate system and a world coordinate system, and converting radar coordinates into the image pixel coordinates; (2) resolving received millimeter-wave radar data, excluding false targets and determining effective targets through data processing, and synchronously collecting camera images; (3) projecting radar scanning points under the world coordinate system to an image pixel coordinate system, and establishing an region of interest (ROI) on the images according to projective points; (4) detecting whether a vehicle exists in the region of interest based on an image processing method. Compared with the prior art, the night preceding vehicle detection method has the characteristics of fusion of the millimeter-wave radar and the machine vision, high real-time performance and high accuracy.

Description

technical field [0001] The invention belongs to the field of vehicle control, in particular to a detection method for a vehicle ahead at night. Background technique [0002] Vehicle intelligent anti-collision warning has become a hot spot in the current international intelligent transportation system research, and the detection of the vehicle ahead is the most important content. Real-time detection and identification of the vehicle ahead can effectively prevent the occurrence of vicious traffic accidents such as rear-end collisions. [0003] At present, there are a variety of front vehicle detection methods at home and abroad: based on vision (monocular vision, stereo vision), based on multi-sensor fusion (sonar sensor, laser sensor), etc.; Monocular vision-based vehicle detection method" (application number: 201210143389.6); Southeast University patent application "based on binocular vision vehicle distance measurement method" (application number: 200710025166.9); Toyota Mo...

AI Technical Summary

Problems solved by technology

[0004] Among them, the vehicle detection method based on monocular vision is not accurate in measuring the depth information of the target, and it is difficult to determine the position of the target object; the vehicle detection method based on stereo vision is prone to drift in the calibration of the camera, and there are complex algorithms, poor precision, and adaptability. Disadvantages such as small range; vehicle detection methods based on vision and laser sensors are not suitable for high-speed vehicle detection. In addition, laser sensors are greatly affected by obstacles and their surface smoothness, and are very sensitive to interference such as lights, so they are not suitable for complex applications. Road environment; vehicle detection methods based on vision and sonar sensors have a short range, and sonar sensors have poor angular resolution and low measurement accuracy

[0005] Due to the complex lighting conditions at night, the accuracy, real-time performance, and robustness of vehicle detection algorithms are required to be higher. However, the existing detection methods for vehicles ahead at night are mainly based on monocular vision. Therefore, there is an urgent need for a vehicle detection method in this field The new technology combines effective multi-sensor data to realize the detection and accurate positioning of the vehicle in front

Images