Adaptive Obstacle Image Detection Method

Abstract

The present invention provides an adaptive obstacle image detection method, which first captures an original image and converts it into a color attribute mode, then captures at least one hue and color saturation component, and divides it into multiple image blocks After that, select one of the blocks as the background block. Then judge whether the rest of the image block is similar to the background block according to an obstacle judgment equation, if similar, keep the image block, and delete the image block if not similar, so as to obtain the initial image range of the obstacle. Finally, three orderly motion flow equations are used to judge which is an obstacle. The invention can detect self-adaptive obstacles, and can issue warnings when the obstacles are approaching.

Description

technical field [0001] The present invention relates to a detection method for obstacles and moving obstacles, in particular to an adaptive obstacle image detection method for obtaining the position of the obstacle by filtering out the background. Background technique [0002] In the past, in the dynamic obstacle detection methods of various image formats, most of them used edge detection or image stream to find the position of the moving object, and then used mathematical morphology to find the possible position of the object, and used the graphic comparison method to compare different images. Area search to determine which is an obstacle or a moving obstacle, and the comparison and judgment ability depends on the applicability of the template, that is, when there are too few types of samples, the misjudgment situation will increase greatly, and Search time is long. Or calculate the image feature values ​​of various possible moving objects in the image, and establish a set...

AI Technical Summary

Problems solved by technology

Or calculate the image feature values ​​of various possible moving objects in the image, and establish a logic mechanism to judge whether it is a moving object. Its technology is not easy to transplant into embedded systems, and it is also time-consuming and lengthy

[0003] Among them, there are previous cases of edge detection, such as Taiwan Patent No.: I327536, which is a method for detecting obstacles. It first takes multiple original images and marginalizes the original images to generate edge objects and corresponding to each edge object. The object information, and then adjust the focal length and horizontal distance of the camera according to the object information to generate the relative distance of the object corresponding to each edge object, compare the relative distance of the object with the threshold distance, and the edge object whose distance is smaller than the threshold distance is an obstacle, but compared The ability to judge depends on the applicability of the template, that is, when there are too few types of samples, the misjudgment situation will increase greatly, and the search time will be longer

[0004] There are previous cases using image flow, such as U.S. Bulletin No. 7,346,191, which discloses the conversion of the optical flow of obstacles captured by using them into a straight line graph to compare whether they are obstacles. Although they are all measurement obstacles, It is different from the algorithm in this case, which uses the comparison between the background and the screen to find the moving object, and the calculation is complicated, which makes the calculation efficiency not good

Images