CN114581959A - Work clothes wearing detection method based on clothes style feature extraction - Google Patents
Work clothes wearing detection method based on clothes style feature extraction Download PDFInfo
- Publication number
- CN114581959A CN114581959A CN202210496533.8A CN202210496533A CN114581959A CN 114581959 A CN114581959 A CN 114581959A CN 202210496533 A CN202210496533 A CN 202210496533A CN 114581959 A CN114581959 A CN 114581959A
- Authority
- CN
- China
- Prior art keywords
- image
- person
- clothes
- wearing
- extracting
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B21/00—Alarms responsive to a single specified undesired or abnormal condition and not otherwise provided for
- G08B21/18—Status alarms
- G08B21/24—Reminder alarms, e.g. anti-loss alarms
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20112—Image segmentation details
- G06T2207/20132—Image cropping
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Abstract
The invention discloses a method for detecting wearing of an industrial garment based on garment style feature extraction, which comprises the following steps: extracting key points of a human body, detecting a person target, recording a work clothes library, extracting features of the work clothes library, extracting features of person clothes in a video stream picture, comparing the features, and triggering an alarm. The method does not need to manually construct the features, effectively reduces the occurrence of false alarm and false alarm situations of a manual feature extraction method, realizes the detection work of wearing of the personnel and the clothing in a video stream picture, greatly improves the online efficiency of a scene model of the clothing, and is suitable for the detection of the clothing of different styles.
Description
Technical Field
The invention belongs to the technical field of image recognition, and particularly relates to a method for detecting wearing of an industrial garment.
Background
In industrial production, the wearing of worker's clothes can bring a great deal of benefit for the enterprise, not only to the culture of enterprise, image and even cohesion, can promote the popularity of enterprise even to a certain extent, improve market competition, and to having special needs in the operating environment, labourers should be according to relevant danger, customize the work clothes that correspond safeguard function, common function has fire prevention, waterproof, prevent wind, dustproof, protect against radiation, prevent static, anti-oxidation, prevent high temperature, prevent low temperature etc. in order to avoid arousing the occurence of failure or injure them. The problem that needs to be solved urgently in industrial production management is to realize the correct wearing of industrial clothes. The wearing of the artificial monitoring work clothes consumes a large amount of manpower, so the labor input is greatly reduced based on the method for monitoring whether the work clothes are worn correctly or not by the intelligent video. The existing model algorithm solution usually needs to collect a large amount of image data in a production environment to carry out customized training of a worker's clothes model, and due to customization and lack of mobility, the scheme cannot realize rapid online of a system model when a new worker's clothes model is met in a new production environment. Therefore, the method for detecting the wearing of the industrial clothes by extracting the features of the clothes style not only meets the requirement of accuracy, but also can adapt to the rapid detection of wearing of different industrial clothes styles.
Disclosure of Invention
The invention aims to provide a method for detecting wearing of an industrial garment based on garment style feature extraction, which aims to solve the problem that the prior art cannot adapt to quick detection of wearing of different types of industrial garments.
In order to achieve the purpose, the invention adopts the following technical scheme:
a method for detecting wearing of an industrial garment based on garment style feature extraction comprises the following steps:
(1) extracting key points of a human body: extracting human key points from all the people in the image by a human key point extraction method, wherein if all the key points of one person are extracted, the target mark of the person is 1, and if not, the target mark is 0;
(2) detecting a person target: identifying a personnel target in the image through a target detection algorithm, and returning the position information of a circumscribed rectangular frame of the personnel target;
(3) recording a work service library: when a bottom library is manufactured for a certain style of work clothes, only one person in a camera collecting picture is ensured, an image of the work clothes correctly worn by the person is collected in the camera picture, whether all key points of the person are in the image is judged by adopting a human body key point extraction method, if the image is stored in the bottom library of the style work clothes, the image of the angle is collected again if the key points are not in the image;
(4) extracting the characteristics of the work service library: cutting out the image of the personnel part in the image by a target detection algorithm, cutting out personnel image pieces, extracting personnel clothing characteristics in the image piecesThe base library of each work clothing style generates three feature vectors:wherein;
(5) Extracting clothing features of people with video streaming pictures: using a target detection algorithm, cutting an image of a person part in the image according to a person detection frame to obtain a person image sheet, extracting person key points in the image sheet by using a human body key point extraction method, and judging whether the whole body part of the person is in the image sheet; let the number of image slices satisfying the condition that the whole body part of the person is within the image slice beNThen extract the firstThe individual image pieces are characterized in thatIn whichAnd is;
(6) And (3) feature comparison: if in step (5)Namely, the condition that the whole body part of a person is in the image slice is not contained in the intercepted video frame picture, and the analysis of the video frame is skipped; if it isFor the second in the video stream picture in step (5)Corresponding feature extracted from personal image sliceRespectively with feature vectorsUsing the Euclidean distance betweenIn whichWhereinExpress the first in the work clothes baseThe feature vectors of the individual person-image slices,representing the first in a video frameFeature vectors of individual person image patches;representing feature vectorsTo (1) aThe number of the elements is one,representing feature vectorsTo (1) aAn element;,and is,Representing feature vectorsAnddimension of (d); mapping the distance between two characteristic vectors, and judging the second time according to a set threshold valueWhether the person in the individual image patch is wearing the frock correctly;
(7) and (3) alarm triggering: and if any characteristic score in the video streaming picture is higher than a set threshold value, triggering a system alarm message to remind a manager to manage and control the personnel who don the work clothes incorrectly in the video picture.
In the steps (1), (3) and (5), the extraction method of the key points of the human body is an mmpos algorithm.
In the step (1), the key points comprise a head, shoulders, a crotch and four limbs.
In the steps (2), (4) and (5), the target detection algorithm is a Yolov5 target detection algorithm.
And (3) acquiring three images of the person wearing the style frock correctly at different angles in a camera picture, wherein the three images are respectively a front side, a side surface and a back side.
In the step (4), the ResNet network is used for extracting the personnel service characteristics in the image slice.
In the step (5), the ResNet network is used for extracting the secondThe clothing characteristics of the individual image pieces.
In the step (6), the distance between the two feature vectors is mapped to 0% -100% by using the deformation of the sigmoid function.
Has the advantages that: compared with the prior art, the invention has the following advantages:
(1) by applying the technologies of target detection, human body key point detection, clothing feature extraction and the like in the field of computer vision, the method is different from other work clothing wearing detection methods, features do not need to be constructed manually, and the situations of false alarm and false alarm of the manual feature extraction method are effectively reduced;
(2) the method for detecting the wearing of the industrial clothes based on the clothing style feature extraction is different from other methods which need to collect a large number of industrial clothes image data sets and carry out customized training, and the detection work of the wearing of the personnel clothes in a video stream picture can be realized only by providing three pictures shot at different angles of the style industrial clothes as an industrial clothes base, so that the online efficiency of an industrial clothes scene model is greatly improved, and the method is suitable for the detection of the industrial clothes of different styles.
Drawings
FIG. 1 is a flow chart of the method of the present invention.
Detailed Description
The invention is further explained below with reference to the drawings.
As shown in fig. 1, the method for detecting wearing of an industrial garment based on the extraction of the style features of the garment of the present invention includes the following steps:
(1) extracting key points of a human body: extracting human body key points from all the persons contained in the image through an mmpos algorithm, wherein if all the key points of one person are extracted, the person target is marked as 1, and if not, the person target is marked as 0;
(2) detecting a person target: identifying a person target in the image through a Yolov5 target detection algorithm, and returning the position information of a circumscribed rectangular frame of the person target;
(3) recording a work service library: when a bottom library is manufactured for a certain style of industrial garment, only one person in a camera acquisition picture is ensured, and three images of different angles, namely the front side, the side and the back side, of the worker correctly wearing the style of industrial garment are acquired in the camera picture; judging whether all key points of the personnel are in the image by adopting the same method as the step (1), if so, storing the image into a bottom library of the style frock, and if not, re-collecting the image of the angle;
(4) extracting the characteristics of the work service library: image cutting is carried out on the personnel part in the image through a Yolov5 target detection algorithm, personnel image pieces are cut out, and personnel uniform characteristics in the image pieces are extractedThe base library of each work clothes style can generate three feature vectors:in which;
(5) Extracting clothing features of people with video streaming pictures: using a Yolov5 target detection algorithm, cutting images of personnel parts in the images according to the personnel detection frame, cutting personnel image slices, extracting personnel key points in the image slices by using the human body key point extraction method adopted in the step (1), and judging whether the whole body part of the personnel is in the image slices; let the number of image slices satisfying the condition that the whole body part of the person is within the image slice beNThen extract the firstThe individual image pieces are characterized in thatWhereinAnd is;
(6) And (3) feature comparison: if in step (5)Namely, the condition that the whole body part of a person is in the image slice is not contained in the intercepted video frame picture, and the analysis of the video frame is skipped; if it isFor the second in the video stream picture in step (5)Corresponding feature extracted from personal image sliceRespectively with feature vectorsUsing a Euclidean distance betweenWhereinIn whichExpress the first in the work clothes baseThe feature vector of the individual person image slice,representing the first in a video frameFeature vectors of individual person image patches;representing feature vectorsTo (1) aThe number of the elements is one,representing feature vectorsTo (1) aAn element;,and is,Representing feature vectorsAnddimension (d); mapping of the distance between two feature vectors is achieved through deformation of a sigmoid function, the mapping is 0% -100%, and the second judgment is carried out according to a set threshold valueWhether the person in the individual image patch is wearing the frock correctly; the threshold is manually set, generally based on manual experience and specific model detection effect, and is increased if the detection sensitivity is higher, otherwise, the threshold is reduced;
(7) and (3) alarm triggering: and if any characteristic score in the video streaming picture is higher than a set threshold value, triggering a system alarm message to remind a manager to manage and control the personnel who don the work clothes incorrectly in the video picture.
The foregoing is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, various modifications and decorations can be made without departing from the principle of the present invention, and these modifications and decorations should also be regarded as the protection scope of the present invention.
Claims (8)
1. A method for detecting wearing of industrial clothes based on clothes style feature extraction is characterized in that: the method comprises the following steps:
(1) extracting key points of a human body: extracting human key points from all the people in the image by a human key point extraction method, wherein if all the key points of one person are extracted, the target mark of the person is 1, and if not, the target mark is 0;
(2) detecting a person target: identifying a personnel target in the image through a target detection algorithm, and returning the position information of a circumscribed rectangular frame of the personnel target;
(3) recording a work service library: when a bottom library is manufactured for a certain style of work clothes, only one person in a camera collecting picture is ensured, an image of the work clothes correctly worn by the person is collected in the camera picture, whether all key points of the person are in the image is judged by adopting a human body key point extraction method, if the image is stored in the bottom library of the style work clothes, the image of the angle is collected again if the key points are not in the image;
(4) extracting the characteristics of the work service library: cutting out the image of the personnel part in the image by a target detection algorithm, cutting out personnel image pieces, extracting personnel clothing characteristics in the image piecesThe base library of each work clothing style generates three feature vectors:wherein;
(5) Extracting clothing features of people with video streaming pictures: using a target detection algorithm, cutting an image of a person part in the image according to a person detection frame to obtain a person image sheet, extracting person key points in the image sheet by using a human body key point extraction method, and judging whether the whole body part of the person is in the image sheet; let the number of image slices satisfying the condition that the whole body part of the person is within the image slice beNThen extract the firstThe individual image pieces are characterized in thatWhereinAnd is;
(6) And (3) feature comparison: if in step (5)Namely, the condition that the whole body part of a person is in the image slice is not contained in the intercepted video frame picture, and the analysis of the video frame is skipped; if it isFor the second in the video stream picture in step (5)Corresponding feature extracted from personal image sliceRespectively with feature vectorsUsing the Euclidean distance betweenWhereinWhereinExpress the first in the work clothes baseThe feature vector of the individual person image slice,representing the first in a video frameFeature vectors of individual person image patches;representing feature vectorsTo (1)The number of the elements is one,representing feature vectorsTo (1) aAn element;,and is,Representing feature vectorsAnddimension (d); mapping the distance between two characteristic vectors, and judging the second time according to a set threshold valueWhether the person in the individual image patch is wearing the frock correctly;
(7) and (3) alarm triggering: and if any characteristic score in the video streaming picture is higher than a set threshold value, triggering a system alarm message to remind a manager to manage and control the personnel who don the work clothes incorrectly in the video picture.
2. The method for detecting the wearing of the industrial clothes based on the extraction of the features of the clothes style according to claim 1, wherein: in the steps (1), (3) and (5), the extraction method of the key points of the human body is an mmpos algorithm.
3. The method for detecting the wearing of the industrial clothes based on the extraction of the features of the clothes style according to claim 1, wherein: in the step (1), the key points comprise a head, shoulders, a crotch and four limbs.
4. The method for detecting the wearing of the industrial clothes based on the extraction of the features of the clothes style according to claim 1, wherein: in the steps (2), (4) and (5), the target detection algorithm is a Yolov5 target detection algorithm.
5. The method for detecting the wearing of the industrial clothes based on the extraction of the features of the clothes style according to claim 1, wherein: and (3) acquiring three images of the person wearing the style frock correctly at different angles in a camera picture, wherein the three images are respectively a front side, a side surface and a back side.
6. The method for detecting the wearing of the industrial clothes based on the extraction of the features of the clothes style according to claim 1, wherein: in the step (4), the ResNet network is used for extracting the personnel service characteristics in the image slice.
8. The method for detecting the wearing of the industrial clothes based on the extraction of the features of the clothes style according to claim 1, wherein: in the step (6), the distance between the two feature vectors is mapped to 0% -100% by using the deformation of the sigmoid function.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210496533.8A CN114581959A (en) | 2022-05-09 | 2022-05-09 | Work clothes wearing detection method based on clothes style feature extraction |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210496533.8A CN114581959A (en) | 2022-05-09 | 2022-05-09 | Work clothes wearing detection method based on clothes style feature extraction |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114581959A true CN114581959A (en) | 2022-06-03 |
Family
ID=81767550
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210496533.8A Pending CN114581959A (en) | 2022-05-09 | 2022-05-09 | Work clothes wearing detection method based on clothes style feature extraction |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114581959A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114743154A (en) * | 2022-06-14 | 2022-07-12 | 广州英码信息科技有限公司 | Work clothes identification method based on registration form and computer readable medium |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2012221437A (en) * | 2011-04-14 | 2012-11-12 | Hitachi Ltd | Object recognition device and object recognition method |
CN111860471A (en) * | 2020-09-21 | 2020-10-30 | 之江实验室 | Work clothes wearing identification method and system based on feature retrieval |
CN113392713A (en) * | 2021-05-20 | 2021-09-14 | 上海可深信息科技有限公司 | Work clothes wearing identification method |
CN113705294A (en) * | 2021-03-04 | 2021-11-26 | 腾讯科技(深圳)有限公司 | Image identification method and device based on artificial intelligence |
CN113792607A (en) * | 2021-08-19 | 2021-12-14 | 辽宁科技大学 | Neural network sign language classification and identification method based on Transformer |
-
2022
- 2022-05-09 CN CN202210496533.8A patent/CN114581959A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2012221437A (en) * | 2011-04-14 | 2012-11-12 | Hitachi Ltd | Object recognition device and object recognition method |
CN111860471A (en) * | 2020-09-21 | 2020-10-30 | 之江实验室 | Work clothes wearing identification method and system based on feature retrieval |
CN113705294A (en) * | 2021-03-04 | 2021-11-26 | 腾讯科技(深圳)有限公司 | Image identification method and device based on artificial intelligence |
CN113392713A (en) * | 2021-05-20 | 2021-09-14 | 上海可深信息科技有限公司 | Work clothes wearing identification method |
CN113792607A (en) * | 2021-08-19 | 2021-12-14 | 辽宁科技大学 | Neural network sign language classification and identification method based on Transformer |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114743154A (en) * | 2022-06-14 | 2022-07-12 | 广州英码信息科技有限公司 | Work clothes identification method based on registration form and computer readable medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106203458B (en) | Crowd video analysis method and system | |
CN111814638B (en) | Security scene flame detection method based on deep learning | |
CN109299683B (en) | Security protection evaluation system based on face recognition and behavior big data | |
CN110648352B (en) | Abnormal event detection method and device and electronic equipment | |
CN109635697A (en) | Electric operating personnel safety dressing detection method based on YOLOv3 target detection | |
CN108053427A (en) | A kind of modified multi-object tracking method, system and device based on KCF and Kalman | |
CN109309808A (en) | A kind of monitoring system and method based on recognition of face | |
CN108846365A (en) | It fights in video detection method, device, storage medium and the processor of behavior | |
CN109558810A (en) | Divided based on position and merges target person recognition methods | |
CN106355154B (en) | Method for detecting frequent passing of people in surveillance video | |
CN114937232B (en) | Wearing detection method, system and equipment for medical waste treatment personnel protective appliance | |
WO2009048513A2 (en) | Apparel as event marker | |
CN109190475A (en) | A kind of recognition of face network and pedestrian identify network cooperating training method again | |
CN112200841B (en) | Cross-domain multi-camera tracking method and device based on pedestrian posture | |
CN110442742A (en) | Retrieve method and device, processor, electronic equipment and the storage medium of image | |
CN108010242A (en) | A kind of security alarm method, system and storage medium based on video identification | |
CN112488034A (en) | Video processing method based on lightweight face mask detection model | |
CN114581959A (en) | Work clothes wearing detection method based on clothes style feature extraction | |
US20220130148A1 (en) | System and Method for Identifying Outfit on a Person | |
CN114885119A (en) | Intelligent monitoring alarm system and method based on computer vision | |
CN109146913B (en) | Face tracking method and device | |
CN114241542A (en) | Face recognition method based on image stitching | |
CN108615020A (en) | A kind of floating population number statistical method in video monitoring regional | |
CN102360450A (en) | Method for counting number of people based on masses | |
JP5752976B2 (en) | Image monitoring device |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20220603 |