Storefront violation identification method and device based on semantic segmentation, and storage medium

Abstract

The invention discloses a storefront violation identification method and device based on semantic segmentation and a storage medium. The storefront violation identification method comprises the steps:acquiring frame images of one or more storefronts; identifying the frame image according to a preset semantic segmentation violation identification model to obtain an identification area belonging toviolation in the frame image; filtering each identification area in the frame image according to a preset mask, and reserving the identification area of the interested storefront; and outputting an identification result of the identification area of the interested storefront. According to the storefront violation identification method and device, each identification area in the frame image is filtered according to the preset mask; therefore, the interference of the non-storefront area in the frame image can be eliminated to the greatest extent, and only the identification area of the interested storefront is concerned with the violation situation, so that the influence of external factors is effectively avoided, and the situations of false detection and missing detection are greatly reduced, and the storefront violation identification accuracy is improved.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a semantic segmentation-based storefront violation identification method, device, and storage medium. Background technique [0002] With the proposal and continuous implementation of smart cities and smart communities, the intelligent management of street shops is imperative. There are two main methods of monitoring store facade violations: the first is the monitoring method of fixed cameras, such as the surveillance camera or ball camera installed on the street wall. Since the area captured by the fixed camera is relatively fixed, the scene analysis is relatively simple. , the area of ​​concern can be segmented through some auxiliary means, and then the violations caused by the random placement of items outside the store in the area of ​​interest can be detected; the second is the monitoring method of mobile cameras, such as vehicle body enforcement devices, etc. In th...

AI Technical Summary

Problems solved by technology

There are two main methods of monitoring store facade violations: the first is the monitoring method of fixed cameras, such as the surveillance camera or ball camera installed on the street wall. Since the area captured by the fixed camera is relatively fixed, the scene analysis is relatively simple. , the area of ​​concern can be segmented through some auxiliary means, and then the violations caused by the random placement of items outside the store in the area of ​​interest can be detected; the second is the monitoring method of mobile cameras, such as vehicle body enforcement devices, etc. Movement status, there are many uncertain factors in terms of movement speed, distance from the store, environmental occlusion, etc. At this time, it will become very difficult to obtain the area of ​​interest, and the same method is difficult to apply

[0003] When processing the shooting content of the camera, the target in the image or video is often detected with the help of deep learning, and the accuracy and speed of target detection are improved through various detection methods. However, no matter what type of target detection method is currently used, there are A problem: simply use a rectangular box to give the position of the target

In fact, when this detection method encounters more complex problems, such as irregular edges of items placed outside the store, overlapping between items, and close proximity of different categories of items, it will lead to illegal operation of the store. The accuracy of the detection is reduced, the actual application effect becomes worse, and it is easy to cause false detection or missed detection of the target.

Images