A Front Vehicle Detection Method Applicable to Complex Scenes

Abstract

The invention relates to a front vehicle detection method suitable for complex scenes. First, the road surface and the sky are distinguished by the position of the vanishing point of the image, the road surface area is used as the effective area for detection, and image preprocessing is performed on the effective area to highlight the vehicle. Horizontal line feature; then divide the effective area of ​​the image into multiple blocks according to the width of the vehicle, and calculate the segmentation threshold in each block to achieve multi-threshold local segmentation; third, use the vehicle width feature on the binary map to remove the segmentation result. Noise filtering, using vehicle edge features to construct rectangles to extract vehicle candidate regions; finally, combined with vertical edge features and the principle of mutual matching between left and right edges, the positioning of vehicle position detection is realized. The method of the invention can effectively reduce the influence of uneven illumination, and at the same time enhance the detection effect of vehicles in poor environment; can effectively eliminate the interference of non-vehicle areas; and effectively reduce the time required for vehicle detection.

Description

technical field [0001] The present invention relates to computer vision and image processing technology, in particular to a front vehicle detection method suitable for complex scenes. Background technique [0002] Statistics show that 30% of the national road traffic accidents are caused by human subjective judgment errors or improper operation. If the driver can know the danger 1.0s in advance and take measures, 90% of traffic accidents can be reduced; if the driver knows 0.5s in advance, 50% of rear-end collisions can be avoided. In order to reduce the occurrence of these accidents, forward collision warning has gradually become part of the car driver assistance system. [0003] At present, there are two main types of forward collision warning system technologies: one uses radar as sensor (such as millimeter wave, microwave, infrared, etc.); Due to the characteristics of video analysis technology, which can effectively classify objects, and has low cost and high scalabil...

AI Technical Summary

Problems solved by technology

The optical flow method mainly determines the position of the detected vehicle based on the inconsistency between the optical flow field of the vehicle in front and the background optical flow caused by the camera movement, but in a moving scene, the relative movement between the vehicle in front and the vehicle is small, The generated relative optical flow field is relatively weak, which affects the detection effect

The model-based matching method uses a large number of different road conditions and different types of vehicle images to form a corresponding feature library for matching processing, but the matching results can only roughly obtain the vehicle area, and cannot locate the real boundary of the vehicle.

Based on the feature extraction method, make full use of the vehicle's edge, shape, grayscale and other features to determine the position of the vehicle in the image, but it is often affected by illumination or environmental changes, resulting in insufficient information, thereby reducing the detection effect

Images