Object contact identification method based on depth image information

Abstract

The invention provides an object contact recognition method based on depth image information, and the method comprises the steps: segmenting a foreground moving object and a static object through employing the depth image information provided by a depth camera and combining with a background removal algorithm and a watershed algorithm; and carrying out connected domain and depth information analysis between different moving objects or between the moving object and a static object. Therefore, whether the contact relationship occurs between the objects is clearly determined. According to the method, the defects that common camera shooting information is limited in two-dimensional information and the object contact relation cannot be accurately judged are overcome; secondly, the method provided by the invention is suitable for contact identification between moving and static objects, and particularly provides a rapid and practical discrimination method under partial and complete shieldingconditions. Whether a contact relation exists between a moving object and a static object or not can be judged very accurately under the shielding condition with low operation cost.

Description

technical field [0001] The invention relates to an object contact recognition method in an image, in particular to an object contact recognition method based on depth image information. Background technique [0002] In the field of video surveillance artificial intelligence applications, the core issues that often need to be dealt with include object segmentation, tracking, recognition, and behavior analysis. Among the behavior analysis requirements, the most basic analysis requirement is to judge the direct relative position and contact relationship between objects. The contact relationship is often regarded as the core identification and judgment point. Many common intelligent application requirements, such as body contact discrimination, intrusion into the warning area, and monitoring of objects that are strictly prohibited from touching, etc., all require contact relationship discrimination as the basis. [0003] Ordinary single-camera surveillance video is flat two-di...

AI Technical Summary

Problems solved by technology

However, in practical applications, due to the natural two-dimensional information limitation of a single camera, even if the plane relationship between objects is completely adjacent, it is difficult to determine whether they are at the same depth position, so it cannot accurately distinguish the contact relationship between objects does it really exist

At the same time, when the judged object is completely occluded, it is more difficult to judge whether there is a real contact relationship with the occluder.

Even for a specific scenario, applying a more complex algorithm can partially solve the problem, but it requires a high algorithm design cost

In the final analysis, the problem is that the lack of depth information leads to

Images