Subregional vehicle bottom shadow detection method based on divergent scanning

Abstract

The invention relates to a subregional vehicle bottom shadow detection method based on divergent scanning. The method comprises the steps of first, acquiring regions of interest of a road in real timeaccording to road disappearance points and road binary information; then, dividing subregions of the region of interest, from top to bottom, selecting and processing an image with resolution ratio of720*1280 according to a principle of perspective, and dividing ROI into three regions by horizontal lines with 40 and 90 pixels of longitudinal distance disappearance points; then, with a disappearance point as a source point, traversing, based on a specific size of rectangular window, pixels on three corresponding region scanning lines by emission lines scanning along a reverse direction of a lane, calculating point by point, and determining thresholds of the three regions; and finally, determining a final vehicle bottom shadow region according to vehicle bottom shadow and vehicle body edgecharacteristics. The method can improve robustness, accuracy and real-time performance of vehicle bottom shadow identification under working conditions of complex road surface and weak light.

Description

technical field [0001] The invention relates to a target detection technology, in particular to a vehicle bottom shadow detection method. Background technique [0002] The vehicle detection system can play the role of assisting driving and warning in advance to the driver, so as to better ensure the safety of people and vehicles. Currently commonly used vehicle detection methods are: monocular vision, stereo vision, multi-sensor fusion. Due to the mature monocular vision algorithm, high real-time performance, low integration cost, low power consumption, small footprint, and large monitoring range, the image information collected by the system can simultaneously obtain other useful road condition information such as road signs, lane markings, pedestrians, etc. , so most of the current vehicle detection methods are based on monocular vision. [0003] The core of vehicle detection is to separate vehicles from complex backgrounds. In the real-time vehicle detection system bas...

AI Technical Summary

Problems solved by technology

Target detection based on template matching method. When the pose of the detected target changes, the similarity of template matching will decrease, which will affect the detection results.

Images