A remaining object detection method and device

Abstract

The invention provides a remaining object detection method and device, and the method comprises the steps: firstly, judging the motion characteristics of pixels through employing a frame difference method according to the real-time rapidity requirement of remaining object detection, and finding a target object which is changed from dynamic to static, i.e., a suspected remaining object; Then, usingHOG features and SVM to performing classification, distinguishing people and articles, and filtering out the people; Judging whether the staying duration of the target object exceeds a preset duration or not by adopting an evidence accumulation mode; And finally, performing feature matching on the person near the target object and the owner of the target object, and if the person is not matched with the owner of the target object, determining the target object as a left object. On one hand, the calculation complexity is low, real-time performance and high efficiency are achieved; the method solves the problems of false alarm and shielding in the detection process of the left object, also solves the problem of distinguishing whether the left object is a person or an object, and also solvesthe problem of distinguishing whether the left object really has no owner, so that the detection accuracy is very high, and the detection problem of the left object is effectively solved.

Description

technical field [0001] Embodiments of the present invention relate to video image processing technologies, and in particular, to a method and device for detecting leftover objects. Background technique [0002] In places with dense crowds, such as subways and other transportation hubs, objects left behind, or abnormally stuck items, such as packages without owners, sometimes appear. Such leftover items may be ordinary items left by the owner, or dangerous items such as bombs. Therefore, in order to detect these leftover items in time, real-time monitoring of transportation hubs and other occasions becomes very important. In addition, there are other extensive applications for the detection of leftover objects, such as dangerous obstacles on railways, falling rocks, littering, debris on traffic arteries / highways / airstrips, accumulations on the ground Slippery liquids (may cause accidents such as pedestrians slipping and falling, car accidents), etc. [0003] In the exis...

AI Technical Summary

Problems solved by technology

[0004] However, the detection of relic objects is a particularly complex problem, because the environment that needs to be detected is usually very complex, such as public places such as transportation hubs, there are many occlusions

At present, there are some solutions for legacy object detection. For example, the fusion of multi-camera information can be used to solve the occlusion problem that is easy to exist in intelligent video surveillance scenes, but accurate camera models are often difficult to obtain, which leads to the ineffectiveness of the algorithm in practical applications. Not ideal; mixed Gaussian background modeling can be used, and on this basis, the state transition of each Gaussian distribution can be observed to detect objects that become stationary after moving into the scene. However, mixed Gaussian background modeling requires a large amount of training and the amount of calculation, it is difficult to meet the real-time needs of the monitoring system

[0005] In the process of implementing the present invention, the inventor found that these solutions in the prior art are either not accurate enough, or have a large amount of calculation, and there are many problems. Some cannot solve the problems of occlusion and false alarms, and some cannot judge Whether it is a person or an object that stays abnormally, and some cannot judge whether the person near the leftover object is its owner or not is not perfect. It can be seen that the detection problem of leftover objects has not been effectively solved in the prior art

Images