CN114092875A

CN114092875A - Operation site safety supervision method and device based on machine learning

Info

Publication number: CN114092875A
Application number: CN202111283451.7A
Authority: CN
Inventors: 刘建明; 常弘
Original assignee: Shenzhen Digital Power Grid Research Institute of China Southern Power Grid Co Ltd
Current assignee: Shenzhen Digital Power Grid Research Institute of China Southern Power Grid Co Ltd
Priority date: 2021-11-01
Filing date: 2021-11-01
Publication date: 2022-02-25

Abstract

The invention discloses a method and a device for supervising operation site safety based on machine learning, wherein the method comprises the following steps: analyzing a video image of an operation site acquired by acquisition equipment to obtain an image of a target person in the operation site, determining a first area image according to the image of the target person, determining a target area image corresponding to a target article in the first area image according to a predetermined article area determination mode, and analyzing the target area image to obtain an analysis result; and judging whether alarm operation needs to be executed for the target personnel according to the analysis result, and outputting an alarm prompt to the target personnel when the judgment result is yes. Therefore, the method and the device can intelligently analyze the video image of the operation site, and are beneficial to improving the processing efficiency of the video image; and warning operation can be intelligently executed according to the image analysis result, and the safety supervision efficiency of the operation site is favorably improved.

Description

Operation site safety supervision method and device based on machine learning

Technical Field

The invention relates to the technical field of safety supervision, in particular to a method and a device for supervising operation site safety based on machine learning.

Background

In the production process of electric power and various projects, the safety supervision of an operation site has important significance for ensuring the life and property safety of workers. Therefore, how to perform safety supervision on the operation site is very important.

In actual life, the illegal behaviors of workers in a working site are various in types and different in characteristics, and typical illegal behaviors mainly comprise that safety helmets are not correctly worn, masks are not correctly worn, protective clothes are not worn according to requirements, and the like.

The traditional safety supervision method manages an operation field in a mode of manually carrying out field supervision patrol and manually checking videos, but the traditional safety supervision method has low supervision efficiency and poor supervision timeliness, and therefore, how to improve the safety supervision efficiency is very important.

Disclosure of Invention

The technical problem to be solved by the invention is to provide a method and a device for supervising the safety of an operation site based on machine learning, wherein the method and the device are used for processing the video image of the operation site through a preset processing algorithm, so that the safety supervision efficiency of the operation site is improved, and the probability of safety accidents caused by the fact that personnel on the operation site do not wear a mask and/or a safety helmet is reduced.

In order to solve the technical problem, a first aspect of the present invention discloses a method for supervising operation site safety based on machine learning, the method comprising:

analyzing a video image of an operation site acquired by acquisition equipment to obtain an image of a target person in the operation site;

determining a first area image according to the image of the target person, wherein the first area image comprises an area image corresponding to the head contour of the target person;

determining a target area image corresponding to a target object in the first area image according to a predetermined object area determination mode, wherein the object area determination mode comprises a mode determined based on a predetermined head-object position relation and/or a mode determined based on parameters of predetermined different area images, and the different area images comprise a head top area image and/or an oral-nasal area image and a head remaining area image;

analyzing the target area image to obtain an analysis result;

judging whether alarm operation needs to be executed for the target personnel according to the analysis result;

and when the alarm operation needs to be executed for the target personnel, outputting an alarm prompt to the target personnel, wherein the alarm prompt is used for prompting the target personnel not to wear a mask and/or a safety helmet according to the safety regulation requirement of the operation site.

As an optional implementation manner, in the first aspect of the present invention, the determining a first region image according to the target image includes:

determining a distance between the target person and the acquisition device;

judging whether the distance is greater than or equal to a preset distance threshold value;

when the distance is judged to be greater than or equal to the distance threshold, determining an image area, in the target image, of which the gray value change of each adjacent frame of image is within a preset gray value threshold, as a key identification area of the target image;

analyzing the key point identification area based on a preset first image processing algorithm to obtain the outline of the target person and an image area corresponding to the outline of the target person;

and determining a region image corresponding to the head contour of the target person according to the contour of the target person and the image region corresponding to the contour of the target person, and taking the region image as a first region image.

As an alternative implementation, in the first aspect of the present invention, the method further includes:

and when the distance is judged to be smaller than the distance threshold value, analyzing the image of the target person based on a preset face model recognition algorithm and face facial features stored in a database to obtain a region image corresponding to the head contour of the target person, and taking the region image as a first region image.

As an optional implementation manner, in the first aspect of the present invention, the determining, according to a predetermined item region determining manner, a target region image corresponding to a target item in the first region image includes:

after the head outline of the target person is determined, dividing the first area image according to a predetermined head-object position relation and/or a predetermined head-object size proportional relation to obtain a division result;

when the article area determining mode comprises a mode determined based on a predetermined head-article position relation, determining a target area image corresponding to a target article in the first area image according to the dividing result, wherein the target article comprises a face covering the mouth and the nose and/or a head covering; alternatively, the first and second electrodes may be,

and when the article area determining mode comprises a mode determined based on the parameters of the predetermined different area images, determining a target area image corresponding to the target article in the first area image according to the parameters of the predetermined different area images.

As an optional implementation manner, in the first aspect of the present invention, after determining a target area image corresponding to a target item in the first area image, and before analyzing the target area image to obtain an analysis result, the method further includes:

performing stretching and amplifying processing on the target area image for preset times to obtain a stretched and amplified image;

analyzing image blocks of the stretched and amplified image in the image edge area, of which the gray value changes within a preset gray threshold value, to obtain a plurality of image blocks;

and executing the reduction processing of the preset times on the image blocks to obtain a target image outline with the size consistent with that of the target area image, and executing the operation of analyzing the target area image to obtain an analysis result.

As an alternative implementation, in the first aspect of the present invention, the analyzing the target area image to obtain an analysis result includes:

analyzing respective target parameters of a plurality of images included in the first area image to obtain data corresponding to the respective target parameters of the plurality of images;

and calculating the numerical difference of the data corresponding to the images of two adjacent frames in the plurality of images as an analysis result.

As an optional implementation manner, in the first aspect of the present invention, the determining whether an alarm operation needs to be performed on the target person according to the analysis result includes:

judging whether all the numerical value differences included in the analysis result have numerical values meeting a preset parameter threshold value, and determining that alarm operation needs to be executed for the target personnel when the numerical values meeting the preset parameter threshold value are judged to exist; alternatively, the first and second electrodes may be,

judging whether a numerical value matched with the oral-nasal feature data exists in data corresponding to respective target parameters of all the images or not based on related data stored in a database, wherein the related data comprises face contour data, the oral-nasal feature data, contour data corresponding to the target object and data of parameters of the target object;

and when the data corresponding to the respective target parameters of all the images are judged to have numerical values matched with the oral-nasal characteristic data, determining that alarm operation needs to be executed for the target personnel.

The invention discloses a working site safety supervision device based on machine learning in a second aspect, which comprises:

the analysis module is used for analyzing the video image of the operation site acquired by the acquisition equipment to obtain the image of the target person in the operation site;

the first determining module is used for determining a first area image according to the image of the target person, wherein the first area image comprises an area image corresponding to the head outline of the target person;

a second determining module, configured to determine a target area image corresponding to a target object in the first area image according to a predetermined object area determining manner, where the object area determining manner includes a manner determined based on a predetermined head-object position relationship and/or a manner determined based on parameters of predetermined different area images, and the different area images include a vertex area image and/or a mouth-nose area image, and a head remaining area image;

the analysis module is further used for analyzing the target area image to obtain an analysis result;

the judging module is used for judging whether to execute alarm operation aiming at the target personnel according to the analysis result obtained by the analyzing module;

and the warning module is used for outputting a warning prompt to the target personnel when the judging module judges that the warning operation needs to be executed for the target personnel, and the warning prompt is used for prompting the target personnel not to wear a mask and/or a safety helmet according to the safety standard requirement of the operation site.

As an optional implementation manner, in the second aspect of the present invention, the first determining module includes:

the first determining submodule is used for determining the distance between the target person and the acquisition equipment;

the judgment submodule is used for judging whether the distance determined by the first determination submodule is larger than or equal to a preset distance threshold value or not;

the first determining submodule is further configured to determine, as the key identification area of the target image, an image area in which a change in gray value of each of a plurality of adjacent frames of images in the target image is within a preset gray value threshold when the judging submodule judges that the distance determined by the first determining submodule is greater than or equal to the distance threshold;

the analysis submodule is used for analyzing the key recognition area based on a preset first image processing algorithm to obtain the outline of the target person and an image area corresponding to the outline of the target person;

the first determining submodule is further configured to determine, according to the contour of the target person and the image area corresponding to the contour of the target person, an area image corresponding to the head contour of the target person as a first area image.

As an optional implementation manner, in the second aspect of the present invention, the analysis sub-module is further configured to, when the judgment sub-module judges that the distance determined by the first determination sub-module is smaller than the distance threshold, analyze the image of the target person based on a preset face model recognition algorithm and facial features stored in a database to obtain an area image corresponding to a head contour of the target person, and use the area image as the first area image.

As an optional implementation manner, in the second aspect of the present invention, the second determining module includes:

the dividing submodule is used for dividing the first area image according to a predetermined head-object position relation and/or a predetermined head-object size proportional relation after the head outline of the target person is determined, so that a dividing result is obtained;

a second determining submodule, configured to determine, according to the division result, a target area image corresponding to a target item in the first area image when the item area determination manner includes a manner determined based on a predetermined head-item positional relationship, where the target item includes a face mask and/or a head cover that covers the mouth and nose;

and the third determining submodule is used for determining a target area image corresponding to the target object in the first area image according to the parameters of the predetermined different area images when the object area determining mode comprises a mode determined based on the parameters of the predetermined different area images.

As an alternative embodiment, in the second aspect of the present invention, the apparatus further comprises:

the processing module is used for performing stretching amplification processing on the target area image for a preset number of times to obtain a stretched and amplified image after the second determining module determines the target area image corresponding to the target object in the first area image and before the analyzing module analyzes the target area image to obtain an analysis result;

the analysis module is further configured to analyze image blocks of the stretched and amplified image in the image edge area, where the change in the gray value of the image in the image edge area is within a preset gray threshold value, to obtain a plurality of image blocks;

the processing module is further configured to perform reduction processing on the plurality of image blocks for the preset number of times to obtain a target image contour with a size consistent with that of the target area image, and trigger the analysis module to perform the operation of analyzing the target area image to obtain an analysis result.

As an optional implementation manner, in the second aspect of the present invention, the analyzing module analyzes the target area image, and the manner of obtaining the analysis result specifically includes:

As an optional implementation manner, in the second aspect of the present invention, the manner in which the determining module determines whether the alarm operation needs to be performed on the target person according to the analysis result obtained by the analyzing module specifically includes:

The invention discloses a third aspect of the device for supervising the safety of the operation site based on machine learning, which comprises:

a memory storing executable program code;

a processor coupled with the memory;

the processor calls the executable program codes stored in the memory to execute the operation field safety supervision method based on machine learning disclosed by the first aspect of the invention.

The fourth aspect of the present invention discloses a computer storage medium, which stores computer instructions, and when the computer instructions are called, the computer instructions are used for executing the method for supervising and managing the operation site safety based on machine learning disclosed in the first aspect of the present invention.

Compared with the prior art, the embodiment of the invention has the following beneficial effects:

in the embodiment of the invention, a video image of an operation site acquired by acquisition equipment is analyzed to obtain an image of a target person in the operation site, a first area image is determined according to the image of the target person, a target area image corresponding to a target article in the first area image is determined according to a predetermined article area determination mode, and the target area image is analyzed to obtain an analysis result; and judging whether alarm operation needs to be executed for the target personnel according to the analysis result, and outputting an alarm prompt to the target personnel when the judgment result is yes. Therefore, the method and the device can intelligently analyze the video image of the operation site, and are beneficial to improving the processing efficiency of the video image; the method can also determine a required target area image according to the image of the target person, intelligently analyze the target area image, and output warning to the target person according to the analysis result of the target area image (for example, the target person who does not wear the mask and/or the safety helmet as required in the operation field exists), thereby reducing the occurrence of safety accidents caused by the target person who does not wear the mask and/or the safety helmet as required in the operation field, and improving the safety supervision efficiency and accuracy of the operation field.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.

FIG. 1 is a flow chart of a method for supervising and managing operation site safety based on machine learning according to an embodiment of the present invention;

FIG. 2 is a flow chart of another work site safety supervision method based on machine learning according to an embodiment of the present invention;

FIG. 3 is a schematic structural diagram of a work site safety supervision device based on machine learning according to an embodiment of the present invention;

FIG. 4 is a schematic structural diagram of another work site safety supervision device based on machine learning according to the embodiment of the present invention;

fig. 5 is a schematic structural diagram of another work site safety supervision device based on machine learning according to an embodiment of the present invention.

Detailed Description

In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

The terms "first," "second," and the like in the description and claims of the present invention and in the above-described drawings are used for distinguishing between different objects and not for describing a particular order. Furthermore, the terms "include" and "have," as well as any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, apparatus, article, or article that comprises a list of steps or elements is not limited to only those steps or elements listed, but may alternatively include other steps or elements not listed, or inherent to such process, method, article, or article.

Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the invention. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. It is explicitly and implicitly understood by one skilled in the art that the embodiments described herein can be combined with other embodiments.

The invention discloses a method and a device for supervising operation site safety based on machine learning, which can intelligently analyze video images of an operation site and are beneficial to improving the processing efficiency of the video images; and intelligently analyzing the target area image, and outputting warning to target personnel according to the analysis result of the target area image (for example, the target personnel who do not wear the mask and/or the safety helmet according to the requirement exist in the operation field), so that the occurrence situation of safety accidents caused by the fact that the target personnel wear the mask and/or the safety helmet according to the requirement can be reduced, the safety supervision efficiency and the safety supervision accuracy of the operation field are improved, and the detailed description is respectively given below.

Example one

Referring to fig. 1, fig. 1 is a schematic flow chart of a method for supervising operation site safety based on machine learning according to an embodiment of the present invention. The operation site safety supervision method based on machine learning described in fig. 1 may be applied to a construction site of electric power, and may also be applied to a construction site of a building industry. As shown in fig. 1, the method for supervising safety of work site based on machine learning may include the following operations:

101. and analyzing the video image of the operation site acquired by the acquisition equipment to obtain the image of the target person in the operation site.

In the embodiment of the present invention, a video capture device of an operation site captures a video image of the operation site in real time, where the capture device may include a camera or a device loaded with a real-time video recording function, and the embodiment of the present invention is not limited. It should be noted that, the camera or the device loaded with the real-time video recording function transmits the acquired video image of the operation site to the background server in real time, so that the background server performs subsequent processing operations on the video image after receiving the acquired video image of the operation site, where the background server may be a local server, a remote server, or a cloud server (also called a cloud server), and when the background server is a non-cloud server, the non-cloud server can be in communication connection with the cloud server, which is not limited in the embodiment of the present invention.

In the embodiment of the invention, the video images of the job site collected by the collecting device comprise frame number images with target personnel and frame number images without target personnel, and the images obtained by analyzing the video images by the background server are the images of the job site comprising the target personnel, so that the image processing operation is performed on the images comprising the target personnel in the following process; if the video image does not include the target person image, the image processing operation is not required to be executed for the image not including the target person by default, so that the image not including the target person is screened out, the number of images needing to be processed is favorably reduced when the processing operation is executed on the video image subsequently, and the interference of the image not including the target person to the image including the target person when the image processing operation is executed is reduced.

In the embodiment of the present invention, it should be noted that the number of the target person may be one or several, and the embodiment of the present invention is not limited.

102. A first region image is determined from the image of the target person.

In the embodiment of the present invention, the first area image includes an area image corresponding to a head contour of the target person, that is, the area image corresponding to the head contour of the target person needs to be determined according to the image including the target person, so that the processing operation is performed on the area image corresponding to the head contour of the target person subsequently.

In the embodiment of the invention, when the number of the target persons is one, the first area image is determined to be the image of the same target person according to the image of the target person, when the number of the target persons exceeds one (such as 2), the area images where the head outlines of different target persons are located are respectively identified and determined according to a preset image identification algorithm, and the area images belonging to the same target person are classified in the same area image set, so that the processing operation is conveniently executed on the area images of the same target person subsequently, and the interference of the determined area images belonging to the same target person (such as person A) with the area images of other target persons (such as person B) is reduced when the processing operation is executed.

Further examples are as follows: assuming that the number of target persons is 2, the target persons include a person a and a person B, and an image set consisting of a plurality of frame header contour region images belonging to the person a and an image set consisting of a plurality of frame header contour region images belonging to the person B can be respectively determined in the images including the person a and the person B according to a preset image recognition algorithm.

103. And determining a target area image corresponding to the target object in the first area image according to a predetermined object area determination mode.

In the embodiment of the invention, the article area determining mode comprises a mode determined based on a predetermined head-article position relation and/or a mode determined based on parameters of predetermined different area images, wherein the different area images comprise an image of a head top area, an image of a mouth-nose area and an image of a head remaining area.

In the embodiment of the present invention, the target item may include a mask and/or a helmet, that is, a potential area image of the mask and/or the helmet in the first area image is determined based on a predetermined head-mask and/or head-helmet positional relationship stored in the database, and the potential area image is also a target area image; and/or determining a target area image corresponding to the target object in the first area image based on the parameters of the different areas of the head stored in the database, and further, how to determine the target area image corresponding to the target object in the first area image according to the parameters of the different areas of the head is exemplified as follows:

the image of the head area of the target person is divided into a head top area, a covering area of the mask when the mask is normally worn, and an area remaining between the head top area and the covering area of the mask in a height ratio, and in addition, for an example of the division in the height ratio: the proportion of the head part on the whole head is 3-4, the proportion of the rest area is 2-3, the proportion of the mask covering area is 3-4, the proportion is substituted into the area image corresponding to the head of the target person, and therefore the potential area image corresponding to the mask and/or the safety helmet is obtained, and the potential area image is also the target area image.

104. And analyzing the target area image to obtain an analysis result.

In the embodiment of the present invention, after the target area image corresponding to the mask and/or the helmet is determined in step 103, an analysis operation is performed on the target area image, so as to obtain an analysis result of whether the target person wears the mask and/or the helmet.

105. And judging whether the alarm operation needs to be executed for the target person according to the analysis result, triggering to execute the step 106 when the judgment result of the step 105 is yes, and continuing to execute the step 105 when the judgment result of the step 105 is no.

In the embodiment of the invention, when the target person is judged to be wearing the mask but not wearing the safety helmet, wearing the safety helmet but not wearing the mask and not wearing the mask according to requirements, the target person is judged to need to perform alarm operation.

106. And outputting a warning prompt to the target personnel.

In the embodiment of the present invention, the warning prompt is used to prompt the target person not to wear a mask and/or a safety helmet according to the safety standard requirement of the job site, and the warning prompt may be performed by controlling a relevant audible and visual warning device of the job site, such as a broadcasting device loaded with a sound amplifying function or a lighting device loaded with a whistle function and a warning light effect, without limitation in the embodiment of the present invention. In addition, the short message can be sent to the target person through the information of the identified target person in a short message notification mode, or the prompting message is sent to the supervisor at the work site, so that the supervisor can reach the work site to process (warn or expel) the behavior of the target person who does not wear the safety helmet and/or the mask according to the information of the target person who does not wear the safety helmet and/or the mask according to the requirement, wherein the information is included in the prompting message.

Therefore, by implementing the operation site safety supervision method based on machine learning described in fig. 1, the video image of the operation site can be intelligently analyzed, which is beneficial to improving the processing efficiency of the video image; the required target area image can be determined according to the image of the target person, and the target area image which needs to be analyzed can be rapidly determined; in addition, the target area image can be intelligently analyzed to obtain an analysis result, and a warning is output to a target person according to the analysis result of the target area image (for example, the target person who does not wear the mask and/or the safety helmet as required in the operation field exists), so that the occurrence of safety accidents caused by the fact that the target person wears the mask and/or the safety helmet as required in the operation field can be reduced, and the safety supervision efficiency and the safety supervision accuracy of the operation field are improved.

In an optional embodiment, determining the first area image according to the target image specifically includes the following operations:

determining the distance between the target person and the acquisition equipment;

when the distance is judged to be larger than or equal to the distance threshold, determining an image area, in the target image, of which the gray value change of every adjacent frames of images is within a preset gray value threshold, as a key identification area of the target image;

analyzing the key identification area based on a preset first image processing algorithm to obtain the outline of the target person and an image area corresponding to the outline of the target person;

and determining a region image corresponding to the head contour of the target person as a first region image according to the contour of the target person and the image region corresponding to the contour of the target person.

In this alternative embodiment, the preset first image processing algorithm may include a Fisherfaces algorithm or an image processing algorithm obtained by improving the Fisherfaces algorithm, and the embodiment of the present invention is not limited thereto.

In this optional embodiment, when the distance between the target person and the acquisition device is greater than or equal to the distance threshold, the acquired image of the target person is an image including the head and the torso of the target person, a preset image processing algorithm is required to be used, a region image corresponding to the head and the torso of the target person is determined through the change of the gray value of the target person in the images of several adjacent frames, and a region image corresponding to the head of the target person is further determined from the region image corresponding to the head and the torso of the target person determined by using the preset algorithm to be used as the first region image.

Therefore, in the optional embodiment, an algorithm for determining the head contour of the target person is provided, and when the distance between the target person and the acquisition device exceeds the distance threshold, the region image corresponding to the head contour of the target person can be intelligently determined according to the acquired image including the target person, so that further processing operation can be favorably performed on the region image corresponding to the head contour of the target person subsequently, and the accuracy rate of identifying whether the target person wears the mask/helmet as required can be favorably improved.

In this optional embodiment, further optionally, when it is determined that the distance is smaller than the distance threshold, analyzing the image of the target person based on a preset face model recognition algorithm and the facial features of the face stored in the database to obtain an area image corresponding to the head contour of the target person, which is used as the first area image.

Therefore, in the optional embodiment, when the distance between the target person and the acquisition equipment is smaller than the distance threshold, the image including the target person can be intelligently analyzed to obtain the region image corresponding to the head contour of the target person, another head contour determination algorithm is provided, so that when the distance between the target person and the acquisition equipment is different, the corresponding head contour determination algorithm can be adaptively applied, and the accuracy of determining the head contour of the target person is improved.

In another optional embodiment, the machine learning-based job site safety supervision method may further include the following operations:

analyzing the video image of the operation site acquired by the acquisition equipment to obtain a related analysis result of the operation site;

judging whether the operation site is an operation site with the risk degree larger than a preset risk degree threshold value or not according to the relevant analysis result;

and when the operation site is judged to be the operation site with the risk degree larger than the preset risk degree threshold value and the target person is identified to enter the operation site, outputting a danger prompt to the target person, wherein the danger prompt is used for informing the target person to leave the operation site.

Therefore, in the optional embodiment, the real-time condition of the operation site can be intelligently monitored, when the danger degree of the monitored operation site is greater than the preset danger degree threshold value and the target person is judged to enter the operation site, a danger prompt is timely output to the target person, and the situation that the target person mistakenly enters the dangerous operation site to cause safety accidents is favorably reduced.

In this optional embodiment, further optionally, the analyzing the video image of the job site acquired by the acquisition device to obtain the related analysis result of the job site specifically includes:

analyzing whether a video image of an operation site acquired by acquisition equipment comprises a danger sign, wherein the danger sign comprises a warning board for prompting that the operation site has danger, and when the video image of the operation site is analyzed to comprise the danger sign, determining that the danger degree of the operation site is greater than a preset danger degree threshold value; and/or the presence of a gas in the gas,

in the video image of the operation site collected by the analysis and collection device, whether dangerous articles exist in the site environment of the operation site or not is analyzed, and the dangerous articles include whether falling objects (such as broken stones and wood chips) exist on the ground of the operation site or not and whether sharp articles (such as nails) exist on the ground and/or the wall of the operation site or not, and when the dangerous articles exist in the site environment of the operation site, the dangerous degree of the operation site is determined to be larger than a preset dangerous degree threshold value.

Therefore, in the optional embodiment, the environment of the operation site can be intelligently analyzed, and when the dangerous marks and/or dangerous goods exist in the operation site, the danger degree of the operation site is determined to be larger than the preset danger degree threshold value, so that the efficiency of finding the dangerous goods existing in the operation site is improved.

Example two

Referring to fig. 2, fig. 2 is a schematic flow chart of another work site safety supervision method based on machine learning according to an embodiment of the present invention. The operation site safety supervision method based on machine learning described in fig. 2 may be applied to a construction site of electric power, and may also be applied to a construction site of a building industry. As shown in fig. 2, the method for supervising safety of work site based on machine learning may include the following operations:

201. and analyzing the video image of the operation site acquired by the acquisition equipment to obtain the image of the target person in the operation site.

202. A first region image is determined from the image of the target person.

203. And determining a target area image corresponding to the target object in the first area image according to a predetermined object area determination mode.

204. And analyzing the target area image to obtain an analysis result.

205. And judging whether the alarm operation needs to be executed for the target person according to the analysis result, triggering to execute the step 206 when the judgment result of the step 205 is yes, and continuing to execute the step 205 when the judgment result of the step 205 is no.

206. And outputting a warning prompt to the target personnel.

In the embodiment of the present invention, please refer to other specific descriptions of steps 101 to 106 in the first embodiment for other descriptions of steps 201 to 206, which is not described again in the embodiment of the present invention.

207. And analyzing respective target parameters of a plurality of images included in the first area image to obtain data corresponding to the respective target parameters of the plurality of images.

In the embodiment of the invention, after the first area image is divided into the plurality of images according to the preset height proportion, the respective target parameters of the plurality of images included in the first area image are analyzed, wherein the target parameters comprise gray values, and data corresponding to the respective gray values of the plurality of images are obtained.

208. And calculating the numerical difference of the data corresponding to two adjacent frames of images in the plurality of images as an analysis result.

In the embodiment of the present invention, after obtaining the data corresponding to the respective gray-level values of the plurality of images in step 207, a numerical difference between the gray-level values of two adjacent frames of images is calculated, which is exemplified as follows: after dividing the area image corresponding to the head into a head top area a, a mask covering area B and a remaining area C, the three areas each have a corresponding gray value, and the gray value difference between a and C, and the gray value difference between B and C are required to be calculated as an analysis result.

Therefore, by implementing the operation site safety supervision method based on machine learning described in fig. 2, the video image of the operation site can be intelligently analyzed, which is beneficial to improving the processing efficiency of the video image; the required target area image can be determined according to the image of the target person, and the target area image which needs to be analyzed can be rapidly determined; in addition, the target area image can be intelligently analyzed to obtain an analysis result, and a warning is output to a target person according to the analysis result of the target area image (for example, the target person who does not wear the mask and/or the safety helmet as required exists in the operation field), so that the occurrence of safety accidents caused by the fact that the target person wears the mask and/or the safety helmet as required in the operation field can be reduced, and the safety supervision efficiency and the accuracy of the operation field are improved; in addition, the numerical difference of the data corresponding to two adjacent frames of images in the plurality of images included in the first area image can be calculated, and the analysis result of whether the target person wears the mask/safety helmet or not is obtained according to the numerical difference, so that the finding efficiency when the target person does not wear the mask/safety helmet is improved.

In an optional embodiment, determining a target area image corresponding to a target item in the first area image according to a predetermined item area determination manner specifically includes the following operations:

when the object area determining mode comprises a mode determined based on the predetermined head-object position relation, determining a target area image corresponding to a target object in the first area image according to the dividing result, wherein the target object comprises a face covering and/or a head covering for covering the mouth and the nose; alternatively, the first and second electrodes may be,

and when the object area determining mode comprises a mode determined based on the parameters of the predetermined different area images, determining the target area image corresponding to the target object in the first area image according to the parameters of the predetermined different area images.

Therefore, in the optional embodiment, two modes of determining the target area image corresponding to the target object are provided, that is, the mode of determining the area corresponding to the mask and/or the helmet, so that the processing algorithm for determining the area corresponding to the mask and/or the helmet can be adaptively executed for different images, and the accuracy of the result of determining the area corresponding to the mask and/or the helmet is improved.

In this optional embodiment, further, after determining a target area image corresponding to a target item in the first area image, and before analyzing the target area image to obtain an analysis result, the method for supervising work site safety based on machine learning may further include the following steps:

performing stretching and amplifying processing on the target area image for a preset number of times to obtain a stretched and amplified image;

analyzing image blocks of the stretched and amplified image in the image edge area, wherein the gray value change of the image blocks is within a preset gray threshold value, and obtaining a plurality of image blocks;

and performing reduction processing on the plurality of image blocks for preset times to obtain a target image outline with the size consistent with that of the target area image, and performing analysis on the target area image to obtain an analysis result.

Therefore, in the optional embodiment, before analyzing the target area image and obtaining the analysis result, the preset image processing operation can be performed on the target area image to obtain the area image corresponding to the contour of the target image, and the contour of the target image is subsequently performed to perform the analysis operation to obtain the analysis result, so that the accuracy of the result of analyzing whether the target person wears the mask and/or the safety helmet is improved.

In another optional embodiment, judging whether an alarm operation needs to be performed for the target person according to the analysis result specifically includes the following steps:

judging whether all the numerical differences included in the analysis result have numerical values meeting a preset parameter threshold value, and determining that alarm operation needs to be executed for target personnel when the numerical values meeting the preset parameter threshold value are judged to exist; alternatively, the first and second electrodes may be,

judging whether a numerical value matched with the oral-nasal feature data exists in data corresponding to respective target parameters of all the images or not based on related data stored in a database, wherein the related data comprises face contour data, the oral-nasal feature data, contour data corresponding to a target object and data of parameters of the target object;

and when the data corresponding to the respective target parameters of all the images are judged to have numerical values matched with the oral-nasal characteristic data, determining that alarm operation needs to be executed for target personnel.

In this optional embodiment, when it is determined that there is a numerical value that satisfies the preset parameter threshold, that is, when the difference between the gray values in the two adjacent frames of images is smaller than the preset gray value threshold, it is determined that an alarm operation needs to be performed on the target person, for example: after the head area image of the target person is divided, comparing the gray value of the potential area of the head helmet with the gray values of the rest areas except the potential area of the head helmet and the potential area of the mask, and when the gray values of the two areas are not different or the difference value is smaller than a preset gray value threshold value, indicating that the target person does not wear the helmet; alternatively, the first and second electrodes may be,

when the data corresponding to the respective target parameters (such as gray values) of all the images are judged to have the numerical values matched with the oral-nasal characteristic data, namely the oral-nasal contours of the target person are identified, the target person is indicated to not wear the mask and/or wear the mask according to requirements, and therefore the alarm operation needs to be executed for the target person.

Therefore, in the optional embodiment, two judgment modes are provided, the judgment mode matched with the analysis result can be adaptively used according to the processing algorithm of the image of the target person in the front, the result obtained by processing and the analysis result, and the accuracy of the result of analyzing whether the target person wears the mask and/or the safety helmet is improved.

In this optional embodiment, further optionally, judging whether an alarm operation needs to be performed for the target person according to the analysis result, specifically, the method may further include the following steps:

and judging whether a numerical value matched with the mask and/or safety helmet characteristic data exists in the data corresponding to the respective target parameters of all the images or not based on the mask and/or safety helmet characteristic data stored in the database, and if not, determining that the target person does not wear the mask and/or wears the mask as required, thereby determining that the alarm operation needs to be executed for the target person.

Therefore, in the optional embodiment, a judgment means for judging whether the target person wears the mask and/or the safety helmet as required is added, so that the judgment means can be selected in a self-adaptive manner according to the analysis result of the head contour of the target person, and the accuracy of the result of judging whether the target person wears the mask and/or the safety helmet as required is improved.

EXAMPLE III

Referring to fig. 3, fig. 3 is a schematic structural diagram of a work site safety supervision apparatus based on machine learning according to an embodiment of the present invention. The device described in fig. 3 may be applied to a construction site of electric power, or a construction site of a building industry, and the present invention is not limited thereto. As shown in fig. 3, the apparatus may include:

the analysis module 301 is configured to analyze a video image of an operation site acquired by the acquisition device to obtain an image of a target person in the operation site;

a first determining module 302, configured to determine a first area image according to the image of the target person obtained by the analyzing module 301, where the first area image includes an area image corresponding to a head contour of the target person;

a second determining module 303, configured to determine, according to a predetermined article area determining manner, an object area image corresponding to the object article in the first area image determined by the first determining module 302, where the article area determining manner includes a manner determined based on a predetermined head-article position relationship and/or a manner determined based on parameters of predetermined different area images, and the different area images include a head top area image and/or a mouth-nose area image, and a head remaining area image;

the analysis module 301 is further configured to analyze the target area image determined by the second determination module 303 to obtain an analysis result;

a judging module 304, configured to judge whether an alarm operation needs to be performed for a target person according to an analysis result obtained by the analyzing module 301;

and the warning module 305 is used for outputting a warning prompt to the target personnel when the judging module 304 judges that the warning operation needs to be executed for the target personnel, wherein the warning prompt is used for prompting the target personnel not to wear the mask and/or the safety helmet according to the safety standard requirement of the operation site.

Therefore, the operation site safety supervision device based on machine learning, which is described by implementing fig. 3, can intelligently analyze the video images of the operation site, which is beneficial to improving the processing efficiency of the video images; the required target area image can be determined according to the image of the target person, and the target area image which needs to be analyzed can be rapidly determined; in addition, the target area image can be intelligently analyzed to obtain an analysis result, and a warning is output to a target person according to the analysis result of the target area image (for example, the target person who does not wear the mask and/or the safety helmet as required exists in the operation field), so that the occurrence of safety accidents caused by the fact that the target person wears the mask and/or the safety helmet as required in the operation field can be reduced, and the safety supervision efficiency and the safety supervision accuracy of the operation field are improved.

In an alternative embodiment, as shown in fig. 4, the first determination module 302 includes a first determination submodule 3021, a determination submodule 3022, and an analysis submodule 3023, wherein:

a first determining submodule 3021 configured to determine a distance between the target person and the acquisition apparatus;

a judgment submodule 3022, configured to judge whether the distance determined by the first determination submodule 3021 is greater than or equal to a preset distance threshold;

the first determining submodule 3021 is further configured to, when the judging submodule 3022 judges that the distance determined by the first determining submodule 3021 is greater than or equal to the distance threshold, determine an image area in the target image, where a change in a gray value of each of a plurality of adjacent frames of images is within a preset gray value threshold, as a key identification area of the target image;

the analysis submodule 3023 is configured to analyze the key recognition area determined by the first determining submodule 3021 based on a preset first image processing algorithm to obtain the contour of the target person and an image area corresponding to the contour of the target person;

the first determining submodule 3021 is further configured to determine, as a first region image, a region image corresponding to the head contour of the target person according to the contour of the target person analyzed by the analyzing submodule 3023 and the image region corresponding to the contour of the target person.

Therefore, in the optional embodiment, an algorithm for determining the head contour of the target person is provided, and when the distance between the target person and the acquisition device exceeds the distance threshold, the region image corresponding to the head contour of the target person can be intelligently determined according to the acquired image including the target person, so that further processing operation can be favorably performed on the region image corresponding to the head contour of the target person subsequently, and the identification accuracy of whether the target person wears the mask/helmet as required is improved.

In an alternative embodiment, as shown in fig. 4, the analysis sub-module 3023 is further configured to, when the judgment sub-module 3022 judges that the distance determined by the first determination sub-module 3021 is smaller than the distance threshold, analyze the image of the target person based on a preset face model recognition algorithm and facial features stored in the database to obtain an area image corresponding to the head contour of the target person, which is used as the first area image.

In an alternative embodiment, as shown in fig. 4, the second determining module 303 includes a dividing sub-module 3031, a second determining sub-module 3032, and a third determining sub-module 3033, wherein:

the dividing submodule 3031 is used for dividing the first region image according to the predetermined head-object position relation and/or the predetermined head-object size proportional relation after the head contour of the target person is determined, so as to obtain a dividing result;

a second determining submodule 3032, configured to determine, according to the division result obtained by the dividing submodule 3031, a target area image corresponding to the target item in the first area image when the item area determination manner includes a manner determined based on the predetermined head-item positional relationship, where the target item includes a face mask and/or a head cover that covers the mouth and nose;

a third determining submodule 3033, configured to determine, when the article region determining manner includes a manner determined based on the predetermined parameter of the different region image, a target region image corresponding to the target article in the first region image according to the predetermined parameter of the different region image after the dividing submodule 3031 obtains the dividing result.

Therefore, two modes of determining the target area image corresponding to the target object are provided, namely the mode of determining the area corresponding to the mask and/or the safety helmet, so that the processing algorithm for determining the area corresponding to the mask and/or the safety helmet can be performed in a self-adaptive manner aiming at different images, and the accuracy of the result of determining the area corresponding to the mask and/or the safety helmet is improved.

In an alternative embodiment, as shown in fig. 4, the machine learning based job site safety supervision apparatus further includes:

the processing module 306 is configured to, after the second determining module 303 determines the target area image corresponding to the target item in the first area image, and before the analyzing module 301 analyzes the target area image to obtain an analysis result, perform stretching and amplifying processing on the target area image for a preset number of times to obtain an image after stretching and amplifying;

the analysis module 301 is further configured to analyze the image blocks of the stretched and amplified image in the image edge area, where the change of the gray value in the image edge area is within the preset gray threshold value, obtained by the processing module 306, to obtain a plurality of image blocks;

the processing module 306 is further configured to perform reduction processing on the plurality of image blocks for a preset number of times to obtain a target image contour with a size consistent with that of the target area image, and trigger the analysis module 301 to perform an operation of analyzing the target area image to obtain an analysis result.

In an alternative embodiment, the analyzing module 301 analyzes the target area image, and the manner of obtaining the analysis result specifically includes:

and calculating the numerical difference of the data corresponding to two adjacent frames of images in the plurality of images as an analysis result.

Therefore, in the optional embodiment, the numerical difference of the data corresponding to two adjacent frames of images in the plurality of images included in the first area image can be calculated, and the analysis result of whether the target person wears the mask/safety helmet is obtained according to the numerical difference, so that the finding efficiency of whether the target person does not wear the mask/safety helmet is improved.

In an optional embodiment, the determining module 304 determines, according to the analysis result obtained by the analyzing module 301, whether to execute the alarm operation for the target person, specifically includes:

Example four

Referring to fig. 5, fig. 5 is a schematic structural diagram of another work site safety supervision device based on machine learning according to an embodiment of the present invention. As shown in fig. 5, the machine learning-based job site safety supervision apparatus may include:

a memory 401 storing executable program code;

a processor 402 coupled with the memory 401;

the processor 402 calls the executable program code stored in the memory 401 to execute the steps in the method for supervising the safety of the job site based on machine learning according to the first embodiment or the second embodiment of the present invention.

EXAMPLE five

The embodiment of the invention discloses a computer storage medium, which stores computer instructions, and when the computer instructions are called, the computer instructions are used for executing the steps of the operation site safety supervision method based on machine learning described in the first embodiment or the second embodiment of the invention.

EXAMPLE six

The embodiment of the invention discloses a computer program product, which comprises a non-transitory computer storage medium storing a computer program, wherein the computer program is operable to make a computer execute the steps of the method for supervising and managing the safety of the operation field based on machine learning described in the first embodiment or the second embodiment.

The above-described embodiments of the apparatus are merely illustrative, and the modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, may be located in one place, or may be distributed on a plurality of network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.

Through the above detailed description of the embodiments, those skilled in the art will clearly understand that the embodiments may be implemented by software plus a necessary general hardware platform, and may also be implemented by hardware. Based on such understanding, the above technical solutions may be embodied in the form of a software product, which may be stored in a computer storage medium, wherein the storage medium includes a Read-Only Memory (ROM), a Random Access Memory (RAM), a Programmable Read-Only Memory (PROM), an Erasable Programmable Read-Only Memory (EPROM), a One-time Programmable Read-Only Memory (OTPROM), an Electrically Erasable Programmable Read-Only Memory (EEPROM), a Compact Disc-Read-Only Memory (CD-ROM) or other magnetic disk, a magnetic tape Memory, a magnetic tape, a magnetic disk, a, Or any other medium which can be used to carry or store data and which can be read by a computer.

Finally, it should be noted that: the method and apparatus for supervising and managing safety of operation site based on machine learning disclosed in the embodiments of the present invention are only the preferred embodiments of the present invention, and are only used for illustrating the technical solution of the present invention, not for limiting the same; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those skilled in the art; the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for supervising safety of a work site based on machine learning, which is characterized by comprising the following steps:

analyzing the target area image to obtain an analysis result;

2. The machine-learning based job site safety supervision method according to claim 1, wherein the determining a first region image from the target image comprises:

determining a distance between the target person and the acquisition device;

3. The machine-learning based job site safety-oversight method of claim 2, further comprising:

4. The machine learning-based work site safety supervision method according to claim 2 or 3, wherein determining a target area image corresponding to a target item in the first area image according to a predetermined item area determination manner comprises:

5. The machine-learning-based job site safety supervision method according to claim 4, wherein after determining a target area image corresponding to a target object in the first area image, and before analyzing the target area image to obtain an analysis result, the method further comprises:

6. The machine-learning-based job site safety supervision method according to claim 5, wherein the analyzing the target area image to obtain an analysis result comprises:

7. The machine learning-based job site safety supervision method according to claim 6, wherein the determining whether an alarm operation needs to be performed for the target person according to the analysis result comprises:

8. A job site safety supervision apparatus based on machine learning, the apparatus comprising:

9. A job site safety supervision apparatus based on machine learning, the apparatus comprising:

a memory storing executable program code;

a processor coupled with the memory;

the processor calls the executable program code stored in the memory to execute the machine learning-based job site safety supervision method according to any one of claims 1-7.

10. A computer storage medium storing computer instructions for performing the machine-learning based job site safety supervision method according to any one of claims 1-7 when invoked.