CN111401301B - Personnel dressing monitoring method, device, equipment and storage medium - Google Patents

Personnel dressing monitoring method, device, equipment and storage medium Download PDF

Info

Publication number
CN111401301B
CN111401301B CN202010266704.9A CN202010266704A CN111401301B CN 111401301 B CN111401301 B CN 111401301B CN 202010266704 A CN202010266704 A CN 202010266704A CN 111401301 B CN111401301 B CN 111401301B
Authority
CN
China
Prior art keywords
detected
dressing
target object
frames
image
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.)
Active
Application number
CN202010266704.9A
Other languages
Chinese (zh)
Other versions
CN111401301A (en
Inventor
李斯
赵齐辉
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Dongpu Software Co Ltd
Original Assignee
Dongpu Software Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Dongpu Software Co Ltd filed Critical Dongpu Software Co Ltd
Priority to CN202010266704.9A priority Critical patent/CN111401301B/en
Publication of CN111401301A publication Critical patent/CN111401301A/en
Application granted granted Critical
Publication of CN111401301B publication Critical patent/CN111401301B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/24—Aligning, centring, orientation detection or correction of the image
    • G06V10/245—Aligning, centring, orientation detection or correction of the image by locating a pattern; Special marks for positioning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
    • G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V2201/00—Indexing scheme relating to image or video recognition or understanding
    • G06V2201/07—Target detection
    • 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 relates to the field of image processing, and discloses a method, a device, equipment and a storage medium for monitoring personal dress, which are used for solving the problem of low efficiency of personal dress identification when monitoring personal dress. The personnel dressing monitoring method comprises the following steps: acquiring an image to be detected, wherein the image to be detected is used for displaying the dressing of a target object and comprises a plurality of detection objects; determining a candidate external frame of a target object in an image to be detected by adopting a target detection algorithm; determining a region to be detected in the candidate external frame, and judging whether the region to be detected has a mark to be detected; when the to-be-detected mark exists on the to-be-detected area, comparing pixel point characteristic values of the to-be-detected mark and a preset standard mark based on a Hash algorithm to obtain a dressing recognition result, wherein the dressing recognition result is used for indicating whether a target object wears a specified dressing or not; and if the dressing identification result is that the target object does not wear the appointed dressing, uploading the dressing identification result to an alarm system.

Description

Personnel dressing monitoring method, device, equipment and storage medium
Technical Field
The invention relates to the field of image processing, in particular to a person dressing monitoring method, a person dressing monitoring device, a person dressing monitoring equipment and a storage medium.
Background
With the development of science and technology, more and more technologies are applied to various fields of life, wherein deep learning technologies represented by convolutional neural networks are widely applied to various artificial intelligence tasks, such as object classification, face recognition, identity recognition and the like. In the application scene of inspection personnel dress, whether utilize the dress of deep learning technique differentiation staff to conform to the standard, whether the staff wears the dress of unified institute promptly, in case there is the personnel that dress is irregular to appear in the work environment, then can lead to the work scene to have the potential safety hazard to cause unnecessary economic loss. In the prior art, a method for detecting the dressing of workers generally includes the steps of installing a video camera outside a working scene, shooting personnel entering the working scene through the video camera, and finally judging whether the dressing of the personnel entering the working scene is standard or not by monitoring personnel. For example: in the granary, in order to guarantee the safety of the granary, the occurrence of irregular personnel needs to be timely warned, so that potential safety hazards can be timely eliminated.
When dressing identification is carried out on personnel entering a working scene, a large amount of manual monitoring is needed, a large amount of manpower and material resources are consumed, and dressing identification efficiency is low.
Disclosure of Invention
The invention mainly aims to solve the problem that dressing identification efficiency of monitoring personnel is low when the personnel dress.
The invention provides a person dressing monitoring method in a first aspect, which comprises the following steps: acquiring an image to be detected, wherein the image to be detected is used for displaying the dressing of a target object and comprises a plurality of detection objects; determining a candidate external frame of the target object in the image to be detected by adopting a target detection algorithm, wherein the candidate external frame is an external frame carrying the type of the target object, and the type of the target object is the type information of the target object; determining a to-be-detected area in the candidate external frame, and judging whether to-be-detected identification exists in the to-be-detected area; when the to-be-detected mark exists on the to-be-detected area, comparing pixel point characteristic values of the to-be-detected mark with preset standard marks based on a Hash algorithm to obtain a dressing identification result, wherein the dressing identification result is used for indicating whether the target object wears a designated dressing or not, and the to-be-detected mark is located at a fixed position on the designated dressing; and if the dressing identification result indicates that the target object does not wear the appointed dressing, uploading the dressing identification result to an alarm system.
Optionally, in a first implementation manner of the first aspect of the present invention, the determining, by using a target detection algorithm, a candidate circumscribed frame of the target object in the image to be detected, where the candidate circumscribed frame is a circumscribed frame carrying the category of the target object, and the category information of the target object, which is the category information of the target object, includes: extracting feature points in the image to be detected by adopting a target detection algorithm, and calculating the convolution of all the feature points to obtain a multilayer target feature map; traversing each feature point in the multilayer target feature map by using a sliding window to generate a plurality of prediction external frames, wherein each prediction external frame comprises a detection object in an image to be detected; adjusting the sizes of the plurality of predicted circumscribed frames in the multi-layer target feature map to obtain a plurality of basic circumscribed frames, wherein the plurality of basic circumscribed frames are used for indicating the external contours of different detection objects; classifying the basic external frames by using a classification function to obtain a plurality of classified external frames, and extracting candidate external frames from the classified external frames, wherein the candidate external frames are external frames carrying the target object type, and the target object type is the type information of the target object.
Optionally, in a second implementation manner of the first aspect of the present invention, traversing each feature point in the multi-layer target feature map by using a sliding window to generate a plurality of prediction bounding boxes, where each prediction bounding box includes a detection object in an image to be detected, and the method includes: traversing each feature point in the multilayer target feature map by using a sliding window to generate a plurality of initial external frames; calculating initial confidence degrees of the initial external frames, screening out a plurality of initial confidence degrees of which the numerical values are greater than a standard threshold value, determining the initial confidence degrees as a plurality of basic confidence degrees, taking the initial external frames corresponding to the basic confidence degrees as a plurality of predicted external frames, wherein each predicted external frame comprises a detection object in the image to be detected.
Optionally, in a third implementation manner of the first aspect of the present invention, the adjusting sizes of the multiple predicted bounding boxes in the multi-layer target feature map to obtain multiple basic bounding boxes, where the multiple basic bounding boxes are used to indicate outer contours of different detection objects includes: mapping the plurality of predicted external frames to the same layer of target feature map in the multilayer target feature map to obtain a mapped target feature layer; performing regression processing on the plurality of predicted circumscribed frames in the mapped target feature layer respectively, and adjusting the sizes of the plurality of predicted circumscribed frames to obtain a plurality of regression circumscribed frames; and performing pooling treatment on the plurality of regression outer frames to obtain a plurality of basic outer frames, wherein the plurality of basic outer frames are used for indicating the outer contours of different detection objects.
Optionally, in a fourth implementation manner of the first aspect of the present invention, the utilization classification isClassifying the multiple basic external frames by a function to obtain multiple classified external frames, and extracting candidate external frames from the multiple classified external frames, wherein the candidate external frames are external frames carrying the target object type, and the target object type is the type information of the target object, and the method comprises the following steps: uniformly distributing the plurality of basic external frames by utilizing a classification loss calculation formula to obtain a plurality of uniformly distributed external frames, wherein the classification loss calculation formula is as follows: FL (p) = -alpha (1-p) γ log (p), where FL (p) is the improved cross-entropy loss, α is the weighting factor, γ is the exponential factor, and p is the true distribution of the base bounding box; classifying different detection objects in the plurality of uniformly distributed external frames by using a classification function to obtain a plurality of classified external frames, wherein the plurality of classified external frames carry class information of the corresponding detection objects; and screening out the classification external frames of which the class information corresponding to the detection object is the class of the target object from the plurality of classification external frames, and determining the corresponding classification external frames as candidate external frames, wherein the class of the target object is the class information of the target object.
Optionally, in a fifth implementation manner of the first aspect of the present invention, the determining, in the candidate circumscribed frame, a to-be-detected region, and determining whether the to-be-detected identifier exists in the to-be-detected region includes: screening a classification circumscribed frame of which the category information corresponding to the detection object is a preset standard identification category from the candidate circumscribed frames to obtain a region to be detected; and judging whether the to-be-detected mark exists in the to-be-detected area.
Optionally, in a sixth implementation manner of the first aspect of the present invention, when the to-be-detected identifier exists on the to-be-detected region, comparing pixel point feature values of the to-be-detected identifier with a preset standard identifier based on a hash algorithm to obtain a clothing identification result, where the clothing identification result is used to indicate whether the target object wears a designated clothing, and the location of the to-be-detected identifier in the designated clothing fixed position includes: when the to-be-detected mark exists on the to-be-detected area, calculating a pixel point characteristic value of a target detection image by utilizing a Hash algorithm, wherein the target detection image is used for indicating an image which carries the to-be-detected mark and is framed by a classified circumscribed frame, and the to-be-detected mark is located at a fixed position on the appointed dressing; when the value of the pixel point characteristic value is larger than or equal to a standard characteristic value, determining a first dressing identification result, wherein the first dressing identification result designates dressing for the target object, and the standard characteristic value is the pixel point characteristic value of a preset standard identification image; and when the value of the pixel point characteristic value is smaller than the standard characteristic value, determining a second dressing identification result, wherein the second dressing identification result is that the target object is not dressed with the specified dressing, and the dressing identification result comprises the first dressing identification result and the second dressing identification result.
A second aspect of the present invention provides a personal outfit monitoring apparatus, comprising: the device comprises an acquisition module, a display module and a display module, wherein the acquisition module is used for acquiring an image to be detected, the image to be detected is used for displaying the dressing of a target object, and the image to be detected comprises a plurality of detection objects; a determining module, configured to determine a candidate circumscribed frame of the target object in the image to be detected by using a target detection algorithm, where the candidate circumscribed frame is a circumscribed frame carrying a category of the target object, and the category of the target object is category information of the target object; the judging module is used for determining a to-be-detected area in the candidate external frame and judging whether to-be-detected identification exists in the to-be-detected area; the comparison module is used for comparing pixel point characteristic values of the to-be-detected mark and a preset standard mark based on a Hash algorithm when the to-be-detected mark exists on the to-be-detected area to obtain a dressing identification result, wherein the dressing identification result is used for indicating whether the target object wears a designated dressing or not, and the to-be-detected mark is located at a fixed position on the designated dressing; and the warning module is used for uploading the dressing identification result to a warning system if the dressing identification result indicates that the target object does not wear the specified dressing.
Optionally, in a first implementation manner of the second aspect of the present invention, the determining module includes: the extraction unit is used for extracting the feature points in the image to be detected by adopting a target detection algorithm and calculating the convolution of all the feature points to obtain a multilayer target feature map; the generating unit is used for traversing each feature point in the multilayer target feature map by using a sliding window to generate a plurality of prediction external frames, and each prediction external frame comprises a detection object in the image to be detected; an adjusting unit, configured to adjust sizes of the plurality of predicted outliers in the multi-layer target feature map to obtain a plurality of basic outliers, where the plurality of basic outliers are used to indicate outer contours of different detection objects; and the classification unit is used for classifying the basic external frames by using a classification function to obtain a plurality of classified external frames, and extracting candidate external frames from the classified external frames, wherein the candidate external frames are external frames carrying the target object type, and the target object type is the type information of the target object.
Optionally, in a second implementation manner of the second aspect of the present invention, the generating unit is specifically configured to: traversing each feature point in the multilayer target feature map by using a sliding window to generate a plurality of initial external frames; calculating initial confidence degrees of the initial external frames, screening out a plurality of initial confidence degrees of which the numerical values are greater than a standard threshold value, determining the initial confidence degrees as a plurality of basic confidence degrees, taking the initial external frames corresponding to the basic confidence degrees as a plurality of predicted external frames, wherein each predicted external frame comprises a detection object in the image to be detected.
Optionally, in a third implementation manner of the second aspect of the present invention, the adjusting unit is specifically configured to: mapping the plurality of predicted external frames to the same layer of target feature map in the multilayer target feature map to obtain a mapped target feature layer; performing regression processing on the plurality of predicted circumscribed frames in the mapped target feature layer respectively, and adjusting the sizes of the plurality of predicted circumscribed frames to obtain a plurality of regression circumscribed frames; and performing pooling treatment on the plurality of regression external frames to obtain a plurality of basic external frames, wherein the plurality of basic external frames are used for indicating external profiles of different detection objects.
Optionally, in a fourth implementation manner of the second aspect of the present invention, the classifying unit is specifically configured to: benefit toUniformly distributing the plurality of basic external frames by using a classification loss calculation formula to obtain a plurality of uniformly distributed external frames, wherein the classification loss calculation formula is as follows: FL (p) = -alpha (1-p) γ log (p), where FL (p) is the improved cross-entropy loss, α is the weighting factor, γ is the exponential factor, and p is the true distribution of the base bounding box; classifying different detection objects in the plurality of uniformly distributed external frames by using a classification function to obtain a plurality of classified external frames, wherein the plurality of classified external frames carry class information of the corresponding detection objects; and screening the classification external frames of which the class information corresponding to the detection object is the class of the target object from the plurality of classification external frames, and determining the corresponding classification external frames as candidate external frames, wherein the class of the target object is the class information of the target object.
Optionally, in a fifth implementation manner of the second aspect of the present invention, the determining module is specifically configured to: screening a classification circumscribed frame of which the category information corresponding to the detection object is a preset standard identification category from the candidate circumscribed frames to obtain a to-be-detected area; and judging whether the to-be-detected area has the to-be-detected mark.
Optionally, in a sixth implementation manner of the second aspect of the present invention, the comparison module is specifically configured to: when the to-be-detected mark exists on the to-be-detected area, calculating a pixel point characteristic value of a target detection image by utilizing a Hash algorithm, wherein the target detection image is used for indicating an image which carries the to-be-detected mark and is framed by a classified circumscribed frame, and the to-be-detected mark is located at a fixed position on the appointed dressing; when the numerical value of the pixel point characteristic value is greater than or equal to a standard characteristic value, determining a first dressing identification result, wherein the first dressing identification result designates dressing for the target object, and the standard characteristic value is the pixel point characteristic value of a preset standard identification image; and when the numerical value of the pixel point characteristic value is smaller than the standard characteristic value, determining a second dressing identification result, wherein the second dressing identification result indicates that the target object is not dressed with a specified dressing, and the dressing identification result comprises the first dressing identification result and the second dressing identification result.
A third aspect of the present invention provides a person-dressing monitoring apparatus comprising: a memory having instructions stored therein and at least one processor, the memory and the at least one processor interconnected by a line; the at least one processor invokes the instructions in the memory to cause the person-clothing monitoring device to perform the person-clothing monitoring method described above.
A fourth aspect of the present invention provides a computer-readable storage medium having stored therein instructions, which, when run on a computer, cause the computer to perform the above-described person dress monitoring method.
According to the technical scheme, an image to be detected is obtained, the image to be detected is used for displaying the dressing of a target object, and the image to be detected comprises a plurality of detection objects; determining a candidate external frame of the target object in the image to be detected by adopting a target detection algorithm, wherein the candidate external frame is an external frame carrying the type of the target object, and the type of the target object is the type information of the target object; determining a region to be detected in the candidate external frame, and judging whether a mark to be detected exists in the region to be detected; when the to-be-detected mark exists on the to-be-detected area, comparing pixel point characteristic values of the to-be-detected mark with preset standard marks based on a Hash algorithm to obtain a dressing identification result, wherein the dressing identification result is used for indicating whether the target object wears a designated dressing or not, and the to-be-detected mark is located at a fixed position on the designated dressing; and if the dressing identification result indicates that the target object does not wear the specified dressing, uploading the dressing identification result to an alarm system. In the embodiment of the invention, the candidate circumscribed frame of the target object is determined by adopting a target detection algorithm, then whether the mark to be detected is a preset standard mark or not is judged in the candidate circumscribed frame, the dressing recognition result of the personnel is obtained, and when the dressing recognition result indicates that the dressing is not appointed by the personnel wearing, the dressing recognition result is uploaded to an alarm system, so that the dressing recognition efficiency of the personnel wearing is improved by adopting the processing mode.
Drawings
FIG. 1 is a schematic diagram of an embodiment of a method for monitoring personal clothing in accordance with the present invention;
FIG. 2 is a schematic diagram of another embodiment of a person dress monitoring method in an embodiment of the invention;
FIG. 3 is a schematic diagram of one embodiment of a person wear monitoring device in an embodiment of the invention;
FIG. 4 is a schematic view of another embodiment of a personal wear monitoring device in accordance with an embodiment of the present invention;
fig. 5 is a schematic diagram of an embodiment of a person wearing monitoring device in the embodiment of the invention.
Detailed Description
The embodiment of the invention provides a method, a device, equipment and a storage medium for monitoring personal dresses, wherein a target detection algorithm is adopted to determine a candidate external frame of a target object, whether a mark to be detected is a preset standard mark or not is judged in the candidate external frame, a dresses recognition result of a person is obtained, and when the dresses recognition result is that the person does not wear a designated dress, the dresses recognition result is uploaded to an alarm system, so that the dressing recognition efficiency of the person is improved.
The terms "first," "second," "third," "fourth," and the like in the description and in the claims, as well as in the drawings, if any, are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It will be appreciated that the data so used may be interchanged under appropriate circumstances such that the embodiments described herein may be implemented in other sequences than those illustrated or described herein. Moreover, the terms "comprises," "comprising," or "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
For convenience of understanding, a detailed flow of an embodiment of the present invention is described below, and referring to fig. 1, an embodiment of a method for monitoring a personal wear in an embodiment of the present invention includes:
101. acquiring an image to be detected, wherein the image to be detected is used for displaying the dressing of a target object and comprises a plurality of detection objects;
it is to be understood that the executing subject of the present invention may be a person wearing monitoring device, and may also be a terminal or a server, which is not limited herein. The embodiment of the present invention is described by taking a server as an execution subject.
The server acquires an image to be detected for displaying the clothing of the target object, wherein the image to be detected comprises a plurality of detection objects.
When the server detects the clothing of the target object, the server needs to acquire an image to be detected for displaying the clothing of the target object, where the format of the image to be detected may be a BMP format, a JPEG2000 format, a TIFF format, a PSD format, a PNG format, a SWF format, an SVG format, and the like, and the format of the image to be detected is not limited in the application. In addition, the resolution of the image to be detected is not limited herein, and generally the resolution of the image to be detected is 320 × 448PPI, 384 × 512PPI, 416 × 576PPI, 480 × 640PPI, and 544 × 704PPI.
It should be noted that at least one detection object is included in the image to be detected, wherein the image to be detected must include the target object, and in addition, other detection objects may also be included, and when the dressing identification of the target object is performed, only the candidate bounding box with the target object is selected to perform the dressing identification. In addition, the outline of the detection object displayed in the image to be detected must be clear, so that the server can accurately identify the circumscribed frame of the detection object.
102. Determining a candidate external frame of a target object in an image to be detected by adopting a target detection algorithm, wherein the candidate external frame is an external frame carrying a target object type, and the target object type is the type information of the target object;
the server determines a candidate external frame of the target object in the image to be detected by adopting a target detection algorithm, wherein the candidate external frame is an external frame carrying the category of the target object, and the category of the target object is category information of the target object.
The server determines candidate external frames of target objects in an image to be detected by adopting a target detection algorithm, wherein the target objects refer to workers and other people needing to wear specified dresses, the target detection algorithm is a RetinaNet target detection algorithm, the target detection algorithm aims at the detection of multi-scale target objects, the algorithm adopts a characteristic pyramid structure to predict the detection objects with different scales, the detection speed and the detection precision are both achieved under the condition of detecting a large number of samples, the RetinaNet mainly comprises a main network structure, a characteristic pyramid structure and a prediction structure, the main network structure is mainly used for extracting image characteristics, the characteristic pyramid structure is mainly fused with the characteristics extracted by a main network to improve the multi-scale target detection precision, the prediction structure is mainly used for predicting a boundary frame containing the target objects, namely the candidate external frames carrying the detection object categories are obtained through frame regression, and therefore the candidate external frames with the target object categories can be selected from the plurality of external frames.
103. Determining a region to be detected in the candidate external frame, and judging whether the region to be detected has a mark to be detected;
and the server determines the area to be detected in the candidate external frame and judges whether the identification to be detected exists in the area to be detected.
After determining the candidate circumscribed frame carrying the preset standard identification category, the server needs to determine the area to be detected of the identification to be detected in the candidate circumscribed frame, and by determining whether the identification to be recognized on the fixed position of the clothing worn by the target object is the preset standard identification, whether the target object wears the appointed clothing can be determined. It should be noted that the preset standard mark is fixed at a certain position of the designated clothing, such as: the school uniform of a certain school embroiders the school badge of the school on the left chest of the school uniform, after the server detects the candidate external frame carrying the student information, the server determines the position embroidered with the school badge in the candidate external frame, judges whether the mark to be detected on the position of the school badge worn by the student is the same as the school badge of the school, and further judges whether the student wears the designated school uniform.
104. When the to-be-detected mark exists on the to-be-detected area, comparing pixel point characteristic values of the to-be-detected mark with a preset standard mark based on a Hash algorithm to obtain a dressing identification result, wherein the dressing identification result is used for indicating whether a target object wears a designated dressing or not, and the to-be-detected mark is located at a fixed position on the designated dressing;
when the to-be-detected mark located at the fixed position on the appointed dressing exists in the to-be-detected area, the server compares the pixel point characteristic value of the to-be-detected mark with a preset standard mark based on a Hash algorithm to obtain a dressing identification result for indicating whether the target object wears the appointed dressing.
If the server detects that the to-be-detected mark exists in the to-be-detected area, the to-be-detected mark needs to be compared with a preset standard mark, wherein the preset standard mark is a mark on a target wearing designated dressing, and the steps are as follows: if the target object works in a certain company, the target object needs to wear a designated dress with a certain company logo. And comparing the pixel point characteristic value of the mark to be detected with the pixel point characteristic value of the preset standard mark, and verifying whether the mark to be detected is the same as the preset standard mark, thereby judging whether the target object wears the appointed dressing. When the similarity between the mark to be detected and the preset standard mark is larger than or equal to the standard characteristic value, the dressing identification result of the target object indicates that the target object wears the appointed dressing, and when the similarity between the mark to be detected and the preset standard mark is smaller than the standard characteristic value, the dressing identification result of the target object indicates that the target object does not wear the appointed dressing, and the server performs subsequent processing on the dressing identification result.
105. And if the dressing identification result is that the target object does not wear the appointed dressing, uploading the dressing identification result to an alarm system.
And if the dressing identification result is that the target object does not wear the appointed dressing, the server uploads the dressing identification result to an alarm system.
When the server detects that the dressing identification result of the target object is that the target object does not wear the designated dressing, the target object does not wear the dressing according to the designated requirement, on one hand, the target object has certain potential safety hazard during working, and casualty accidents caused by irregular dressing are easily caused; on the other hand, the target object is not a worker, and similarly, the safety of a working scene cannot be guaranteed, so that certain potential safety hazards exist. Therefore, when the server detects that the dressing identification result of the target object is that the target object does not wear the designated dressing, the dressing identification result of the target object needs to be uploaded to an alarm system, and safety management of a workshop is enhanced.
In the embodiment of the invention, the candidate external frame of the target object is determined by adopting a target detection algorithm, then whether the mark to be detected is a preset standard mark or not is judged in the candidate external frame, the dressing recognition result of a person is obtained, and when the dressing recognition result indicates that the person does not wear a designated dressing, the dressing recognition result is uploaded to an alarm system, so that the dressing recognition efficiency of the person is improved by the processing mode.
Referring to fig. 2, another embodiment of the method for monitoring personal clothing in the embodiment of the present invention includes:
201. acquiring an image to be detected, wherein the image to be detected is used for displaying the dressing of a target object and comprises a plurality of detection objects;
the server acquires an image to be detected for displaying the clothing of the target object, wherein the image to be detected comprises a plurality of detection objects.
When the server detects the target object by means of dressing, an image to be detected for displaying the target object by means of dressing needs to be acquired, the format of the image to be detected can be a BMP format, a JPEG2000 format, a TIFF format, a PSD format, a PNG format, a SWF format, a SVG format and the like, and the format of the image to be detected is not limited in the application. In addition, the resolution of the image to be detected is not limited herein, and is generally 320 × 448PPI, 384 × 512PPI, 416 × 576PPI, 480 × 640PPI, and 544 × 704PPI.
It should be noted that at least one detection object is included in the image to be detected, wherein the image to be detected must include the target object, and in addition, other detection objects may also be included, and when the dressing identification of the target object is performed, only the candidate bounding box with the target object is selected to perform the dressing identification. In addition, the outline of the detection object displayed in the image to be detected must be clear, so that the server can accurately identify the circumscribed frame of the detection object.
In addition, before the image to be detected of the target object is obtained, the position area of the target mark needs to be marked, and an image marking tool is used for marking the target mark.
202. Extracting feature points in the image to be detected by adopting a target detection algorithm, and calculating the convolution of all the feature points to obtain a multilayer target feature map;
and the server adopts a target detection algorithm to extract the characteristic points in the image to be detected and calculates the convolution of all the characteristic points to obtain a multilayer target characteristic diagram.
The principle of the target detection method is that a series of processing and analysis are performed on the feature points in the image to be detected, so that the information of the detected object category in the image to be detected is finally identified. When the server processes the feature points, firstly, convolution calculation is carried out on all the feature points, and therefore a multilayer target feature map is obtained. It should be noted that the multilayer target feature map is arranged in a pyramid structure, and the arrangement of the pyramid structure is suitable for calculating and analyzing the multi-scale image to be detected, so that the features under all scales have rich semantic information.
203. Traversing each feature point in the multilayer target feature map by using a sliding window to generate a plurality of prediction external frames, wherein each prediction external frame comprises a detection object in the image to be detected;
and traversing each feature point by using a sliding window in the multilayer target feature map by the server to generate a plurality of prediction external frames, wherein each prediction external frame comprises a detection object in the image to be detected. Specifically, the method comprises the following steps:
the server firstly traverses each feature point in a multilayer target feature map by using a sliding window to generate a plurality of initial external frames; and then the server calculates the initial confidence degrees of the initial external frames, screens out a plurality of initial confidence degrees of which the initial confidence degree value is greater than a standard threshold value, determines the initial confidence degrees as a plurality of basic confidence degrees, and takes the initial external frames corresponding to the basic confidence degrees as a plurality of predicted external frames, wherein each predicted external frame comprises a detection object in the image to be detected.
The server firstly traverses each feature point by using a sliding window, so that a plurality of initial external frames are generated and predicted, wherein the initial external frames are generated according to the external contour of a detection object in an image to be detected; then the server calculates the basic confidence of each initial circumscribed frame, wherein the basic confidence refers to the probability that the detection object in the initial circumscribed frame is an object of a certain category, so that the number of the initial circumscribed frames is multiple, the server screens the initial circumscribed frames in each layer of feature map by using a standard threshold value, and rejects the circumscribed frames with low basic confidence of the initial circumscribed frames to obtain a plurality of predicted circumscribed frames, such as: the server screens a plurality of initial external frames by using a standard threshold value of 0.05 for each layer of feature graph, eliminates the external frames of which the basic confidence coefficient is less than 0.05, and takes the rest initial external frames as a plurality of predicted external frames; and finally, the server analyzes and processes the plurality of predicted external frames, so that the number of external frames processed by the server is reduced, and the efficiency of the server for processing the external frames is improved.
204. Adjusting the sizes of a plurality of predicted external frames in the multilayer target characteristic diagram to obtain a plurality of basic external frames, wherein the plurality of basic external frames are used for indicating external outlines of different detection objects;
the server adjusts the sizes of the plurality of predicted circumscribed frames in the multi-layer target feature map to obtain a plurality of basic circumscribed frames used for indicating the external outlines of different detection objects. Specifically, the method comprises the following steps:
firstly, a server maps a plurality of predicted external frames to a target feature map of the same layer in a multilayer target feature map to obtain a mapped target feature layer; then the server carries out regression processing on the multiple predicted external frames in the mapped target feature layer respectively, and the sizes of the multiple predicted external frames are adjusted to obtain multiple regression external frames; and finally, the server performs pooling processing on the multiple regression extension frames to obtain the basic extension frames used for indicating the external profiles of different detection objects.
It should be noted that, when the sizes of the multiple basic external frames are adjusted, there are three dimensions and three proportions in the target feature map of each layer, where the three dimensions are three scales
Figure BDA0002441548500000121
The three proportions are 1:1, 1:2 and 2:1, the general pyramid structure has five layers, namely, has five layers of target feature maps, and the scale range of the prediction candidate box is 32 × 32, 64 × 64, 128 × 0128, 256 × 256 and 512 × 512. For example: when the input image to be detected is 800 × 800, the sizes of the target feature map in the target detection algorithm are 100 × 100, 50 × 50, 25 × 25, 12 × 12 and 6 × 6, respectively, and then the frame regression processing is performed on the prediction candidate frames in the target feature map, so as to adjust the sizes of the plurality of prediction circumscribed frames.
For multiple predicted bounding boxes, bounding box regression aims to find a relationship that allows multiple input predicted bounding boxes to be mapped to obtain a regression bounding box that is closer to the true bounding box. Here, a regression circumscribed frame that is more in line with the detection target is obtained by using the bounding box regression. And finally, pooling the multiple regression extension frames by the server to obtain a basic extension frame for indicating the external profiles of different detection objects.
205. Classifying the basic external frames by using a classification function to obtain a plurality of classified external frames, and extracting candidate external frames from the classified external frames, wherein the candidate external frames are external frames carrying target object types, and the target object types are the type information of the target object;
the server classifies the basic external frames by using a classification function to obtain a plurality of classified external frames, extracts candidate external frames from the classified external frames to obtain candidate external frames carrying target object classes, wherein the target object classes are class information of the target object. Specifically, the method comprises the following steps:
the server firstly utilizes a classification loss calculation formula to uniformly distribute the plurality of basic external frames to obtain a plurality of uniformly distributed external frames, wherein the classification loss calculation formula is as follows: FL (p) = -alpha (1-p) γ log (p), where FL (p) is the improved cross-entropy loss, α is the weighting factor, γ is the exponential factor, and p is the true distribution of the base bounding box; then the server classifies different detection objects in the plurality of uniformly distributed external frames by using a classification function to obtain a plurality of classified external frames carrying class information of the corresponding detection objects; and finally, the server screens the classified external frames of which the category information corresponding to the detection object is the category of the target object from the plurality of classified external frames, and determines the corresponding classified external frames as candidate external frames, wherein the category of the target object is the category information of the target object.
The server firstly utilizes a classification loss calculation formula to calculate the plurality of basic external frames, and uniformly distributes the plurality of basic external frames, because the content contained in the basic external frames can be a detection object in an image to be detected or a scene in the image to be detected, the distribution of the basic external frames and the scene in the image to be detected needs to be balanced, and the calculation result of the server is more accurate.
It should be noted that, a SoftMax function is used to classify a plurality of basic bounding boxes, and the SoftMax function converts output values of multiple classifications into relative probabilities, that is, objective class confidence degrees, so that data are easier to compare. The formula specifically applied is:
Figure BDA0002441548500000131
in the formula: v i For the output of the preceding output unit of the classifier, i is the class index, C is the total class number, S i Is the current elementThe ratio of the index of the element to the sum of the indices of all elements. The server will preset the category information of a plurality of detection objects to compare with the calculated category information, and obtain the category information of the detection objects in the basic external frame. And finally, the server screens out the classification external frames with the class information of the detection object as the class of the target object from the plurality of classification external frames with the class information of the corresponding detection object, and determines the classification external frames as candidate external frames.
206. Determining a region to be detected in the candidate external frame, and judging whether the region to be detected has a mark to be detected;
and the server determines the area to be detected in the candidate external frame and judges whether the identification to be detected exists in the area to be detected. Specifically, the method comprises the following steps:
the server firstly screens a classification external frame of which the category information corresponding to the detection object is a preset standard identification category in the candidate external frame to obtain a to-be-detected area; and then the server judges whether the to-be-detected mark exists in the to-be-detected area.
After determining the candidate external frame carrying the preset standard mark, the server needs to determine the area to be detected of the mark to be detected in the candidate external frame, and can determine whether the target object wears the appointed dress by determining whether the mark to be identified on the fixed position of the clothes worn by the target object is the preset standard mark. It should be noted that the preset standard mark is fixed at a certain position of the designated clothing, such as: the school uniform of a certain school embroiders the school badge of the school on the left chest of the school uniform, after the server detects the candidate external frame carrying the student information, the server determines the position embroidered with the school badge in the candidate external frame, judges whether the mark to be detected on the position of the school badge worn by the student is the same as the school badge of the school, and further judges whether the student wears the designated school uniform.
207. When the to-be-detected mark exists on the to-be-detected area, comparing pixel point characteristic values of the to-be-detected mark with a preset standard mark based on a Hash algorithm to obtain a dressing identification result, wherein the dressing identification result is used for indicating whether a target object wears a designated dressing or not, and the to-be-detected mark is located at a fixed position on the designated dressing;
when the to-be-detected mark exists in the to-be-detected area, the server compares the to-be-detected mark located at the fixed position on the appointed dressing with a preset standard mark on the basis of a Hash algorithm to obtain a dressing identification result for indicating whether the target object wears the appointed dressing. Specifically, the method comprises the following steps:
when the to-be-detected mark exists in the to-be-detected area, the server firstly utilizes a Hash algorithm to calculate a pixel point characteristic value of a target detection image, the target detection image is used for indicating an image which carries the to-be-detected mark and is framed by a classification extension frame, and the to-be-detected mark is located at a fixed position on a designated dressing; when the numerical value of the pixel point characteristic value is greater than or equal to the standard characteristic value, the server determines a first dressing identification result, the first dressing identification result designates dressing for the target object, and the standard characteristic value is the pixel point characteristic value of the preset standard identification image; and when the numerical value of the pixel point characteristic value is smaller than the standard characteristic value, the server determines a second dressing identification result, the second dressing identification result is that the target object is not dressed with the appointed dressing, and the dressing identification result comprises a first dressing identification result and a second dressing identification result.
If the server detects that the to-be-detected mark exists in the to-be-detected area, the to-be-detected mark needs to be compared with a preset standard mark, wherein the preset standard mark is a mark on a target wearing designated dressing, and the steps are as follows: if the target object works in a certain company, the target object needs to wear a designated dress with a certain company logo. The characteristic value of the pixel point of the mark to be detected is compared with the characteristic value of the pixel point of the preset standard mark, whether the mark to be detected is the same as the preset standard mark is verified, and therefore whether the target object wears the appointed dressing or not is judged. When the similarity between the mark to be detected and the preset standard mark is larger than or equal to the standard characteristic value, the dressing identification result of the target object indicates that the target object wears the appointed dressing, and when the similarity between the mark to be detected and the preset standard mark is smaller than the standard characteristic value, the dressing identification result of the target object indicates that the target object does not wear the appointed dressing, and the server performs subsequent processing on the dressing identification result.
Whether the mark to be detected is the same as a preset standard mark or not is judged by calculating the characteristic value of a pixel point of the mark to be detected, a Hash algorithm is used, firstly, a server zooms a target detection image carrying the mark to be detected to an appointed size, and the zoomed appointed size is determined by the information content and the complexity of the target detection image, such as: when the mark to be detected in the target detection image is a simple icon, the information content of the mark to be detected is small, the complexity is low, the target detection image can be zoomed to a small size, and when complex scenes such as scenery exist in the target detection image, the information content is large, and the complexity is high, the target detection image cannot be zoomed to a small size, so that important information is easily lost; secondly, the server performs gray processing on the target detection image to reduce the complexity of later-stage calculation; then the server respectively and sequentially calculates and records the average value of each line of pixel points in the target detection image, wherein the average value of each pixel point corresponds to one line of characteristics; then the server calculates the obtained pixel point characteristic value and the variance of the pixel point, and the obtained variance is the characteristic value of the target detection image; and finally, comparing the obtained characteristic value of the target detection image with the characteristic value of the pixel point of the preset standard identification image, thereby obtaining a dressing identification result.
208. And if the dressing identification result is that the target object does not wear the appointed dressing, uploading the dressing identification result to an alarm system.
And if the dressing identification result is that the target object does not wear the appointed dressing, the server uploads the dressing identification result to an alarm system.
When the server detects that the dressing identification result of the target object is that the target object does not wear the designated dressing, the target object does not wear the dressing according to the designated requirement, on one hand, the target object has certain potential safety hazard during working, and casualty accidents caused by irregular dressing are easily caused; on the other hand, the target object is not a worker, and similarly, the safety of a working scene cannot be guaranteed, so that certain potential safety hazards exist. Therefore, when the server detects that the dressing identification result of the target object is that the target object does not wear the designated dressing, the dressing identification result of the target object needs to be uploaded to an alarm system, and safety management of a workshop is enhanced.
In the embodiment of the invention, the candidate circumscribed frame of the target object is determined by adopting a target detection algorithm, then whether the mark to be detected is a preset standard mark or not is judged in the candidate circumscribed frame, the dressing recognition result of the personnel is obtained, and when the dressing recognition result indicates that the dressing is not appointed by the personnel wearing, the dressing recognition result is uploaded to an alarm system, so that the dressing recognition efficiency of the personnel wearing is improved by adopting the processing mode.
In the above, the method for monitoring personal clothing in the embodiment of the present invention is described, and in the following, referring to fig. 3, the device for monitoring personal clothing in the embodiment of the present invention is described, and one embodiment of the device for monitoring personal clothing in the embodiment of the present invention includes:
an obtaining module 301, configured to obtain an image to be detected, where the image to be detected is used to display the dressing of a target object, and the image to be detected includes multiple detection objects;
a determining module 302, configured to determine a candidate circumscribed frame of the target object in the image to be detected by using a target detection algorithm, where the candidate circumscribed frame is a circumscribed frame carrying a category of the target object, and the category of the target object is category information of the target object;
a determining module 303, configured to determine a to-be-detected region in the candidate circumscribed frame, and determine whether the to-be-detected region has a to-be-detected identifier;
a comparison module 304, configured to, when the to-be-detected identifier exists on the to-be-detected region, compare pixel point feature values of the to-be-detected identifier with a preset standard identifier based on a hash algorithm to obtain a dressing identification result, where the dressing identification result is used to indicate whether the target object wears a designated dressing, and the to-be-detected identifier is located at a fixed position on the designated dressing;
and an alarm module 305, configured to upload the dressing identification result to an alarm system if the dressing identification result indicates that the target object does not wear a specific dressing.
In the embodiment of the invention, the candidate external frame of the target object is determined by adopting a target detection algorithm, then whether the mark to be detected is a preset standard mark or not is judged in the candidate external frame, the dressing recognition result of a person is obtained, and when the dressing recognition result indicates that the person does not wear a designated dressing, the dressing recognition result is uploaded to an alarm system, so that the dressing recognition efficiency of the person is improved by the processing mode.
Referring to fig. 4, another embodiment of the personal wear monitoring device in the embodiment of the present invention includes:
the acquiring module 301 is configured to acquire an image to be detected, where the image to be detected is used for displaying the dressing of a target object, and the image to be detected includes multiple detection objects;
a determining module 302, configured to determine a candidate circumscribed frame of the target object in the image to be detected by using a target detection algorithm, where the candidate circumscribed frame is a circumscribed frame carrying a category of the target object, and the category of the target object is category information of the target object;
a determining module 303, configured to determine a to-be-detected region in the candidate circumscribed frame, and determine whether the to-be-detected region has a to-be-detected identifier;
a comparison module 304, configured to, when the to-be-detected identifier exists on the to-be-detected region, compare pixel point feature values of the to-be-detected identifier with a preset standard identifier based on a hash algorithm to obtain a dressing identification result, where the dressing identification result is used to indicate whether the target object wears a designated dressing, and the to-be-detected identifier is located at a fixed position on the designated dressing;
and an alarm module 305, configured to upload the dressing identification result to an alarm system if the dressing identification result indicates that the target object does not wear a specified dressing.
Optionally, the determining module 302 includes:
an extracting unit 3021, configured to extract feature points in the image to be detected by using a target detection algorithm, and calculate convolutions of all the feature points to obtain a multi-layer target feature map;
a generating unit 3022, configured to traverse each feature point in the multilayer target feature map by using a sliding window, and generate a plurality of prediction external frames, where each prediction external frame includes a detection object in the image to be detected;
an adjusting unit 3023, configured to adjust sizes of the plurality of predicted outliers in the multi-layer target feature map to obtain a plurality of basic outliers, where the plurality of basic outliers are used to indicate outer contours of different detection objects;
a classifying unit 3024, configured to classify the multiple basic frames by using a classification function to obtain multiple classification frames, and extract a candidate frame from the multiple classification frames, where the candidate frame is a frame that carries the target object class, and the target object class is the class information of the target object.
Optionally, the generating unit 3022 may be further specifically configured to:
traversing each feature point in the multilayer target feature map by using a sliding window to generate a plurality of initial external frames;
calculating initial confidence degrees of the initial external frames, screening out a plurality of initial confidence degrees of which the numerical values are greater than a standard threshold value, determining the initial confidence degrees as a plurality of basic confidence degrees, taking the initial external frames corresponding to the basic confidence degrees as a plurality of predicted external frames, wherein each predicted external frame comprises a detection object in the image to be detected.
Optionally, the adjusting unit 3023 may be further specifically configured to:
mapping the plurality of predicted external frames to the same layer of target feature map in the multilayer target feature map to obtain a mapped target feature layer;
performing regression processing on the plurality of predicted circumscribed frames in the mapped target feature layer respectively, and adjusting the sizes of the plurality of predicted circumscribed frames to obtain a plurality of regression circumscribed frames;
and performing pooling treatment on the plurality of regression external frames to obtain a plurality of basic external frames, wherein the plurality of basic external frames are used for indicating external profiles of different detection objects.
Optionally, the classification unit 3024 may be further specifically configured to:
uniformly distributing the plurality of basic external frames by utilizing a classification loss calculation formula to obtain a plurality of uniformly distributed external frames, wherein the classification loss calculation formula is as follows: FL (p) = -alpha (1-p) γ log (p), where FL (p) is the improved cross-entropy loss, α is the weighting factor, γ is the exponential factor, and p is the true distribution of the base bounding box;
classifying different detection objects in the plurality of uniformly distributed external frames by using a classification function to obtain a plurality of classified external frames, wherein the plurality of classified external frames carry class information of the corresponding detection objects;
and screening out the classification external frames of which the class information corresponding to the detection object is the class of the target object from the plurality of classification external frames, and determining the corresponding classification external frames as candidate external frames, wherein the class of the target object is the class information of the target object.
Optionally, the determining module 303 may be further specifically configured to:
screening a classification circumscribed frame of which the category information corresponding to the detection object is a preset standard identification category from the candidate circumscribed frames to obtain a to-be-detected area;
and judging whether the to-be-detected area has the to-be-detected mark.
Optionally, the comparison module 304 may be further specifically configured to:
when the to-be-detected mark exists on the to-be-detected area, calculating a pixel point characteristic value of a target detection image by utilizing a Hash algorithm, wherein the target detection image is used for indicating an image which carries the to-be-detected mark and is framed by a classified circumscribed frame, and the to-be-detected mark is located at a fixed position on the appointed dressing;
when the numerical value of the pixel point characteristic value is greater than or equal to a standard characteristic value, determining a first dressing identification result, wherein the first dressing identification result designates dressing for the target object, and the standard characteristic value is the pixel point characteristic value of a preset standard identification image;
and when the numerical value of the pixel point characteristic value is smaller than the standard characteristic value, determining a second dressing identification result, wherein the second dressing identification result indicates that the target object is not dressed with a specified dressing, and the dressing identification result comprises the first dressing identification result and the second dressing identification result.
In the embodiment of the invention, the candidate external frame of the target object is determined by adopting a target detection algorithm, then whether the mark to be detected is a preset standard mark or not is judged in the candidate external frame, the dressing recognition result of a person is obtained, and when the dressing recognition result indicates that the person does not wear a designated dressing, the dressing recognition result is uploaded to an alarm system, so that the dressing recognition efficiency of the person is improved by the processing mode.
Fig. 3 and 4 describe the personnel dressing monitoring device in the embodiment of the present invention in detail from the perspective of the modular functional entity, and the personnel dressing monitoring device in the embodiment of the present invention is described in detail from the perspective of hardware processing.
Fig. 5 is a schematic diagram of a personal clothing monitoring device according to an embodiment of the present invention, where the personal clothing monitoring device 500 may have a relatively large difference due to different configurations or performances, and may include one or more processors (CPUs) 510 (e.g., one or more processors) and a memory 520, and one or more storage media 530 (e.g., one or more mass storage devices) storing applications 533 or data 532. Memory 520 and storage media 530 may be, among other things, transient or persistent storage. The program stored on storage medium 530 may include one or more modules (not shown), each of which may include a series of instructions operating on the person wearing monitoring device 500. Further, processor 510 may be configured to communicate with storage medium 530 to execute a series of instruction operations in storage medium 530 on personal apparel monitoring device 500.
The personnel apparel monitoring device 500 may also include one or more power supplies 540, one or more wired or wireless network interfaces 550, one or more input-output interfaces 560, and/or one or more operating systems 531, such as Windows Server, mac OS X, unix, linux, freeBSD, and the like. Those skilled in the art will appreciate that the configuration of the personal-worn monitoring device illustrated in FIG. 5 does not constitute a limitation of the personal-worn monitoring device, and may include more or fewer components than illustrated, or some components in combination, or a different arrangement of components.
The present invention also provides a computer readable storage medium, which may be a non-volatile computer readable storage medium, which may also be a volatile computer readable storage medium, having stored therein instructions, which, when run on a computer, cause the computer to perform the steps of the person clothing monitoring method.
It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a read-only memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.
The above-mentioned embodiments are only used for illustrating the technical solutions of the present invention, and not for limiting the same; although the present invention has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.

Claims (10)

1. A person dress monitoring method, characterized in that the person dress monitoring method comprises:
acquiring an image to be detected, wherein the image to be detected is used for displaying the dressing of a target object and comprises a plurality of detection objects;
determining a candidate external frame of the target object in the image to be detected by adopting a target detection algorithm, wherein the candidate external frame is an external frame carrying the type of the target object, and the type of the target object is the type information of the target object;
determining a to-be-detected area in the candidate external frame, and judging whether to-be-detected identification exists in the to-be-detected area;
when the to-be-detected mark exists on the to-be-detected area, comparing pixel point characteristic values of the to-be-detected mark with preset standard marks based on a Hash algorithm to obtain a dressing identification result, wherein the dressing identification result is used for indicating whether the target object wears a designated dressing or not, and the to-be-detected mark is located at a fixed position on the designated dressing;
and if the dressing identification result indicates that the target object does not wear the specified dressing, uploading the dressing identification result to an alarm system.
2. The method for monitoring personal clothing according to claim 1, wherein the determining a candidate outline of the target object in the image to be detected by using a target detection algorithm, the candidate outline being an outline carrying the category of the target object, the category information of the target object including:
extracting feature points in the image to be detected by adopting a target detection algorithm, and calculating the convolution of all the feature points to obtain a multilayer target feature map;
traversing each feature point in the multilayer target feature map by using a sliding window to generate a plurality of prediction external frames, wherein each prediction external frame comprises a detection object in an image to be detected;
adjusting the sizes of the plurality of predicted circumscribing frames in the multi-layer target feature map to obtain a plurality of basic circumscribing frames, wherein the plurality of basic circumscribing frames are used for indicating the external contours of different detection objects;
classifying the basic external frames by using a classification function to obtain a plurality of classified external frames, and extracting candidate external frames from the classified external frames, wherein the candidate external frames are external frames carrying the target object type, and the target object type is the type information of the target object.
3. The method for monitoring personal clothing as claimed in claim 2, wherein traversing each feature point in the multi-layer target feature map by using a sliding window to generate a plurality of predicted bounding boxes, each predicted bounding box including a detected object in the image to be detected comprises:
traversing each feature point in the multilayer target feature map by using a sliding window to generate a plurality of initial external frames;
and calculating the initial confidence coefficients of the initial external frames, screening out a plurality of initial confidence coefficients of which the initial confidence values are greater than a standard threshold value, determining the initial confidence coefficients as a plurality of basic confidence coefficients, taking the initial external frames corresponding to the basic confidence coefficients as a plurality of predicted external frames, and enabling each predicted external frame to comprise a detection object in the image to be detected.
4. The method of claim 2, wherein the resizing the plurality of predicted outlier boxes in the multi-level target feature map to obtain a plurality of base outlier boxes, the plurality of base outlier boxes indicating outer contours of different inspection objects comprises:
mapping the plurality of predicted external frames to the same layer of target feature map in the multilayer target feature map to obtain a mapped target feature layer;
performing regression processing on the plurality of predicted circumscribed frames in the mapped target feature layer respectively, and adjusting the sizes of the plurality of predicted circumscribed frames to obtain a plurality of regression circumscribed frames;
and performing pooling treatment on the plurality of regression outer frames to obtain a plurality of basic outer frames, wherein the plurality of basic outer frames are used for indicating the outer contours of different detection objects.
5. The method according to claim 2, wherein the classifying the plurality of basic frames by using a classification function to obtain a plurality of classified frames, and extracting candidate frames from the plurality of classified frames, wherein the candidate frames are frames carrying the target object category, and the target object category is category information of the target object, and the method comprises:
uniformly distributing the plurality of basic external frames by utilizing a classification loss calculation formula to obtain a plurality of uniformly distributed external frames, wherein the classification loss calculation formula is as follows: FL (p) = -alpha (1-p) γ log (p), where FL (p) is the improved cross-entropy loss, α is the weighting factor, γ is the exponential factor, and p is the true distribution of the base bounding box;
classifying different detection objects in the plurality of uniformly distributed external frames by using a classification function to obtain a plurality of classified external frames, wherein the plurality of classified external frames carry class information of corresponding detection objects;
and screening the classification external frames of which the class information corresponding to the detection object is the class of the target object from the plurality of classification external frames, and determining the corresponding classification external frames as candidate external frames, wherein the class of the target object is the class information of the target object.
6. The method for monitoring personal clothing according to claim 4, wherein the determining the area to be detected in the candidate circumscribing frame and the determining whether the identifier to be detected exists in the area to be detected comprises:
screening a classification circumscribed frame of which the category information corresponding to the detection object is a preset standard identification category from the candidate circumscribed frames to obtain a region to be detected;
and judging whether the to-be-detected area has the to-be-detected mark.
7. The personnel dressing monitoring method according to any one of claims 1 to 6, wherein when the to-be-detected mark exists on the to-be-detected area, comparing the to-be-detected mark with a preset standard mark based on a hash algorithm to obtain a dressing recognition result, wherein the dressing recognition result is used for indicating whether the target object wears a designated dressing, and the position of the to-be-detected mark at the fixed position on the designated dressing comprises:
when the to-be-detected mark exists on the to-be-detected area, calculating a pixel point characteristic value of a target detection image by utilizing a Hash algorithm, wherein the target detection image is used for indicating an image which carries the to-be-detected mark and is framed by a classified circumscribed frame, and the to-be-detected mark is located at a fixed position on the appointed dressing;
when the value of the pixel point characteristic value is larger than or equal to a standard characteristic value, determining a first dressing identification result, wherein the first dressing identification result designates dressing for the target object, and the standard characteristic value is the pixel point characteristic value of a preset standard identification image;
and when the value of the pixel point characteristic value is smaller than the standard characteristic value, determining a second dressing identification result, wherein the second dressing identification result is that the target object is not dressed with the specified dressing, and the dressing identification result comprises the first dressing identification result and the second dressing identification result.
8. A personal wear monitoring device, comprising:
the device comprises an acquisition module, a display module and a display module, wherein the acquisition module is used for acquiring an image to be detected, the image to be detected is used for displaying the dressing of a target object, and the image to be detected comprises a plurality of detection objects;
a determining module, configured to determine a candidate circumscribed frame of the target object in the image to be detected by using a target detection algorithm, where the candidate circumscribed frame is a circumscribed frame carrying a category of the target object, and the category of the target object is category information of the target object;
the judging module is used for determining a to-be-detected region in the candidate external frame and judging whether the to-be-detected mark exists in the to-be-detected region;
the comparison module is used for comparing pixel point characteristic values of the to-be-detected mark with preset standard marks based on a Hash algorithm when the to-be-detected mark exists on the to-be-detected area to obtain a dressing identification result, wherein the dressing identification result is used for indicating whether the target object wears a designated dressing or not, and the to-be-detected mark is located at a fixed position on the designated dressing;
and the warning module is used for uploading the dressing identification result to a warning system if the dressing identification result indicates that the target object does not wear the specified dressing.
9. A personal wear monitoring device, comprising: a memory having instructions stored therein and at least one processor, the memory and the at least one processor interconnected by a line;
the at least one processor invokes the instructions in the memory to cause the person-dress monitoring device to perform the person-dress monitoring method of any one of claims 1-7.
10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out a method for person clothing monitoring according to any one of claims 1 to 7.
CN202010266704.9A 2020-04-07 2020-04-07 Personnel dressing monitoring method, device, equipment and storage medium Active CN111401301B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010266704.9A CN111401301B (en) 2020-04-07 2020-04-07 Personnel dressing monitoring method, device, equipment and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010266704.9A CN111401301B (en) 2020-04-07 2020-04-07 Personnel dressing monitoring method, device, equipment and storage medium

Publications (2)

Publication Number Publication Date
CN111401301A CN111401301A (en) 2020-07-10
CN111401301B true CN111401301B (en) 2023-04-18

Family

ID=71433074

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010266704.9A Active CN111401301B (en) 2020-04-07 2020-04-07 Personnel dressing monitoring method, device, equipment and storage medium

Country Status (1)

Country Link
CN (1) CN111401301B (en)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112036251A (en) * 2020-08-04 2020-12-04 汇纳科技股份有限公司 Method, system, medium and terminal for detecting user's dressing
CN112395999A (en) * 2020-11-19 2021-02-23 平安普惠企业管理有限公司 Wearing standard judging method based on image recognition and related equipment
CN113537878A (en) * 2021-06-28 2021-10-22 上海欢猫科技有限责任公司 Package delivery method, device, equipment and storage medium
CN113887310A (en) * 2021-09-03 2022-01-04 珠高电气检测有限公司 Worker standard dressing identification method, system and medium

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109409354A (en) * 2017-08-18 2019-03-01 深圳市道通智能航空技术有限公司 UAV Intelligent follows target to determine method, unmanned plane and remote controler
CN109815868A (en) * 2019-01-15 2019-05-28 腾讯科技（深圳）有限公司 A kind of image object detection method, device and storage medium

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109409354A (en) * 2017-08-18 2019-03-01 深圳市道通智能航空技术有限公司 UAV Intelligent follows target to determine method, unmanned plane and remote controler
CN109815868A (en) * 2019-01-15 2019-05-28 腾讯科技（深圳）有限公司 A kind of image object detection method, device and storage medium

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
王晓宁 ; 宫法明 ; 时念云 ; 吕轩轩 ; .基于卷积神经网络的复杂场景目标检测算法.计算机系统应用.2019,(06),全文. *

Also Published As

Publication number Publication date
CN111401301A (en) 2020-07-10

Similar Documents

Publication Publication Date Title
CN111401301B (en) Personnel dressing monitoring method, device, equipment and storage medium
Racki et al. A compact convolutional neural network for textured surface anomaly detection
JP6398979B2 (en) Video processing apparatus, video processing method, and video processing program
US20140169639A1 (en) Image Detection Method and Device
CN108038424B (en) Visual automatic detection method suitable for high-altitude operation
CN110781836A (en) Human body recognition method and device, computer equipment and storage medium
CN110728252B (en) Face detection method applied to regional personnel motion trail monitoring
CN111860471B (en) Work clothes wearing identification method and system based on feature retrieval
CN112991269A (en) Identification and classification method for lung CT image
Seong et al. Vision-based safety vest detection in a construction scene
CN112613454A (en) Electric power infrastructure construction site violation identification method and system
CN108664875A (en) Underground belt-conveying monitoring method based on image recognition
CN108898623A (en) Method for tracking target and equipment
Rentao et al. Indoor smoking behavior detection based on yolov3-tiny
CN108460320A (en) Based on the monitor video accident detection method for improving unit analysis
CN115512134A (en) Express item stacking abnormity early warning method, device, equipment and storage medium
CN116386120A (en) Noninductive monitoring management system
Patel et al. Top-down and bottom-up cues based moving object detection for varied background video sequences
CN117115147B (en) Textile detection method and system based on machine vision
CN108416304B (en) Three-classification face detection method using context information
CN106548195A (en) A kind of object detection method based on modified model HOG ULBP feature operators
CN111461260A (en) Target detection method, device and equipment based on feature fusion and storage medium
CN116311347A (en) Person on Shift detection method, electronic device, and computer-readable storage medium
CN115995097A (en) Deep learning-based safety helmet wearing standard judging method
KR20160073490A (en) System for assessment of safety level at construction site based on computer vision

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
GR01 Patent grant
GR01 Patent grant