Work clothes identification method based on registration form and computer readable medium

Abstract

The invention discloses a work clothes identification method based on a registration form and a computer readable medium. The method comprises the following steps: work clothes registration, work clothes detection and work clothes picture enhancement, and extraction of all work clothes features to obtain a work clothes feature set; performing work clothes detection on the to-be-identified frame of picture to obtain a to-be-verified work clothes coordinate set and a to-be-verified work clothes set; extracting features of all work clothes in the to-be-verified work clothes set; tracking the work clothes in the current picture to obtain a work clothes id; filtering the extracted work clothes, and performing feature similarity calculation on the filtered work clothes and a work clothes library feature set to obtain a final recognition result of each work clothes; updating the id information set; inquiring the id information set, and judging according to the multi-frame identification result of the work clothes of the same id to obtain a final identification result of the work clothes; and repeating the steps until the last frame. The method is wide in application range, various clothes specified by a user can be recognized in a multi-scene environment, manual intervention is not needed, and the enterprise efficiency is greatly improved.

Description

technical field [0001] The invention relates to the field of artificial intelligence image recognition, and in particular, to a work clothes recognition method based on a registration form and a computer-readable medium. Background technique [0002] At present, in construction sites or companies that require uniform work clothes, most of them are manually checking whether the personnel are dressed properly, which is inefficient, and a small amount of artificial intelligence is used to extract clothes features or make clothes according to unique features of clothes (such as numbers on clothes). Recognition, but there are general shortcomings: 1. The algorithm is relatively limited, and it can only identify specific styles of clothes in specific scenes, and the scope of application is limited (hereinafter referred to as disadvantage 1) data enhancement; 2. The algorithm only considers the identification of a single image, and there is no Consider the continuous frame picture ...

AI Technical Summary

Problems solved by technology

[0002] At present, most of the scenes that require uniform uniforms on construction sites or companies are manually checked whether the personnel are wearing standard clothes, which is inefficient. A small amount of artificial intelligence is used to extract clothes features or to make clothes according to the unique features of clothes (such as numbers on clothes). Recognition, but there are common disadvantages: 1. The algorithm has relatively high limitations, and can only recognize specific styles of clothes in specific scenes, and the scope of application is limited (hereinafter referred to as disadvantage 1). Data enhancement; 2. The algorithm only considers the recognition of a single image, there is no Consider continuous frame image processing in the case of video (hereinafter referred to as disadvantage 2) (tracking algorithm); 3. The recognition result is generated by relying on a single frame image, and the stability of the recognition result is not strong (hereinafter referred to as disadvantage 3)

[0015] Patent Document 1 has the following disadvantages: the method for selecting the second feature set is unknown, the efficiency of artificial selection is low, and it does not consider the scene conditions when the light changes (that is, there is disadvantage 1), and there is no continuous tracking and continuous extraction of a target in the video The process of features and recognition (that is, there is a disadvantage 2), the single recognition effect is not reliable (that is, there is a disadvantage 3), and the results will be reported frequently

[0017] Patent Document 2 has the following disadvantages: this solution requires a specially designed number suit, and in many cases the work clothes do not have such numbers printed, so this solution can only be applied in specific scenarios (that is, there is disadvantage 1)

At the same time, Patent Document 2 does not consider the processing of continuous frame images in the case of video. It also recognizes a single image and relies on a single frame image to generate a recognition result (that is, there are disadvantages 2 and 3).

[0019] Patent Document 3 has the following disadvantages: the existing technology only satisfies a certain style of clothing, which has high limitations, and when the light changes, the color characteristics will change, and the effect is unreliable (that is, there is disadvantage 1)

At the same time, Patent Document 3 does not consider the processing of continuous frame images in the case of video. It also recognizes a single image and relies on a single frame image to generate a recognition result (that is, there are disadvantages 2 and 3).

[0021] Patent Document 4 has the following disadvantages: it does not describe in detail how to identify security suits, and this method is only for security suits, which has high limitations, and it is not reasonable to frequently calculate a person's security status data in actual situations

Patent Document 4 also has disadvantages 1, 2, and 3

Images