Obstacle detection method and system based on vehicle-mounted fisheye lens under motion background

Abstract

The invention provides an obstacle detection method and system based on a vehicle-mounted fisheye lens under a motion background. Moving objects around a vehicle can be accurately and efficiently detected when a vehicle runs, the reliability of a detection result is good, and the method comprises the following steps of step 1, collecting an image , extracting feature points, conducting optical flow calculation and optical flow clustering on the feature points to acquire candidate targets in the image; step 2, collecting a vehicle signal to obtain predicted position coordinates of the vehicle at each moment so as to obtain driving displacement of the vehicle; step 3, comparing the displacement of the coordinate corresponding to the feature point of each candidate target with the driving displacement of the vehicle, and detecting whether the candidate target is an obstacle target or not; step 4, tracking an obstacle target, and obtaining a movement track of the obstacle target accordingto the position change of the obstacle target in each frame of input image; and step 5, outputting the position of the obstacle target on the image.

Description

technical field [0001] The invention relates to the fields of assisted driving and image processing, in particular to an obstacle detection method and system based on a vehicle-mounted fisheye lens under a moving background. Background technique [0002] With the development of the automobile industry, while the number of automobiles is increasing rapidly, the safety problems caused by traffic are becoming more and more prominent, and traffic accidents, especially vicious traffic accidents, are on the rise. Therefore, it is particularly urgent to research and develop vehicle auxiliary driving. [0003] So far, although products using ultrasonic and radar sensors for reversing assistance have been widely sold in the market, products using vision sensors for vehicle assistance are still in the research stage. Currently, vision-based target detection algorithms mainly include feature-based, stereo-based Vision and detection algorithms based on optical flow. Feature-based targe...

AI Technical Summary

Problems solved by technology

Feature-based target detection is only suitable for detecting specific types of targets, and is not suitable for detecting unknown types of targets; target detection based on stereo vision has a large amount of computation, usually requires special hardware, and the cost is high; optical flow-based The target detection algorithm is difficult to distinguish between the point of the obstacle and the point on the road, the detection result is inaccurate, and it is difficult to meet the real-time requirements

Images