WO2022227209A1 - 基于无服务边缘计算的佩戴口罩监测方法、装置及设备 - Google Patents

基于无服务边缘计算的佩戴口罩监测方法、装置及设备 Download PDF

Info

Publication number
WO2022227209A1
WO2022227209A1 PCT/CN2021/097214 CN2021097214W WO2022227209A1 WO 2022227209 A1 WO2022227209 A1 WO 2022227209A1 CN 2021097214 W CN2021097214 W CN 2021097214W WO 2022227209 A1 WO2022227209 A1 WO 2022227209A1
Authority
WO
WIPO (PCT)
Prior art keywords
preset
mask
face image
terminal
edge
Prior art date
Application number
PCT/CN2021/097214
Other languages
English (en)
French (fr)
Inventor
李佳琳
王健宗
Original Assignee
平安科技（深圳）有限公司
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 平安科技（深圳）有限公司 filed Critical 平安科技（深圳）有限公司
Publication of WO2022227209A1 publication Critical patent/WO2022227209A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/16—Human faces, e.g. facial parts, sketches or expressions
    • G06V40/161—Detection; Localisation; Normalisation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks

Definitions

  • the present application relates to the technical field of artificial intelligence, and in particular, to a method, device, computer equipment, and computer-readable storage medium for monitoring wearing a mask based on edge computing without services.
  • cloud computing solutions are limited by the Internet bandwidth of real-time camera video streaming, so they need to be equipped with a relevant database, and the data calculated at the hardware layer and the real-time captured pictures need to be uploaded to the virtual cloud layer. Perform virtualization processing for backup, and then upload it to the database for storage. Especially in places with high traffic flow, very large data is obtained every day.
  • the inventor realized that the mask detection method based on cloud computing increases network pressure, causes system delay, and reduces the efficiency of monitoring wearing masks.
  • the present application provides a monitoring method, device, computer equipment and computer-readable storage medium for wearing a mask based on serviceless edge computing, which can solve the technical problem of low monitoring efficiency of wearing a mask in the traditional technology.
  • the present application provides a method for monitoring wearing a mask based on serviceless edge computing, which is applied to a local preset smart terminal, wherein the method includes: acquiring a face image collected by a local preset collection device; The face image is sent to a local preset edge node for edge computing, and an edge computing result corresponding to the face image is obtained, wherein the local preset edge node is a device that has deployed a preset high-performance neural network inference computing framework NCNN.
  • the local intelligent terminal obtains the edge calculation result, obtains the mask detection result corresponding to the face image according to the edge calculation result, and displays the mask detection result.
  • the present application further provides a monitoring device for wearing a mask based on edge computing without service, which is applied to a local preset smart terminal, wherein the device includes: a first acquisition unit for acquiring a local preset acquisition device a collected face image; a computing unit, configured to send the face image to a local preset edge node for edge calculation, and obtain an edge calculation result corresponding to the face image, wherein the local preset edge node is A local intelligent terminal with a preset high-performance neural network inference calculation framework NCNN is deployed; a display unit is used to obtain the edge calculation result, and according to the edge calculation result, the mask detection result corresponding to the face image is obtained, and the The detection result of the mask is displayed.
  • the present application also provides a computer device, which includes a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program: acquiring a local preset collection device The collected face image; the face image is sent to a local preset edge node for edge calculation, and an edge calculation result corresponding to the face image is obtained, wherein the local preset edge node is configured with a preset high
  • the local intelligent terminal of the performance neural network inference calculation framework NCNN obtains the edge calculation result, obtains the mask detection result corresponding to the face image according to the edge calculation result, and displays the mask detection result.
  • the present application further provides a computer-readable storage medium, where the computer-readable storage medium stores a computer program, and the computer program is executed by the processor as follows: acquiring a face collected by a local preset collection device image; send the face image to a local preset edge node for edge calculation, and obtain the edge calculation result corresponding to the face image, wherein the local preset edge node is a pre-set high-performance neural network reasoning deployed Calculate the local intelligent terminal of the framework NCNN; obtain the edge calculation result, obtain the mask detection result corresponding to the face image according to the edge calculation result, and display the mask detection result.
  • the computer-readable storage medium stores a computer program
  • the computer program is executed by the processor as follows: acquiring a face collected by a local preset collection device image; send the face image to a local preset edge node for edge calculation, and obtain the edge calculation result corresponding to the face image, wherein the local preset edge node is a pre-set high-performance neural network reasoning deployed Calculate the local intelligent terminal of the framework NCNN
  • the present application provides a method, device, computer equipment, and computer-readable storage medium for monitoring wearing a mask based on serviceless edge computing.
  • a local preset smart terminal Through a local preset smart terminal, a face image collected by a local preset collection device is obtained, and the collected image is collected. The face image is sent to the local preset edge node for edge computing, and the edge computing result corresponding to the face image is obtained.
  • Edge computing, using edge computing most traffic loads will be processed at the data source instead of sending all data through the network, network congestion is significantly improved, and then the edge computing results are obtained.
  • the edge computing results the The mask detection result corresponding to the face image, and the mask detection result is displayed, so as to realize the monitoring of the face image wearing a mask, which greatly reduces the mask monitoring performed by the cloud computing-based service architecture in the traditional technology.
  • edge computing based on local preset edge nodes can make full use of preset terminals in the local environment, reduce the demand for local centralized computing resources, reduce the deployment cost of mask monitoring, and improve performance.
  • the convenience of mask monitoring improves the efficiency of mask monitoring on face images.
  • FIG. 1 is a schematic flowchart of a method for monitoring wearing a mask based on edge computing without service provided by an embodiment of the present application
  • FIG. 2 is a schematic diagram of a system architecture in a method for monitoring wearing a mask based on edge computing without service provided by an embodiment of the present application;
  • FIG. 3 is a schematic diagram of a first sub-flow of a method for monitoring wearing a mask based on edge computing without service provided by an embodiment of the present application;
  • FIG. 4 is a schematic diagram of a second sub-flow of a method for monitoring wearing a mask based on edge computing without service provided by an embodiment of the present application;
  • FIG. 5 is a schematic diagram of the third sub-flow of the monitoring method for wearing a mask based on edge computing without service provided by an embodiment of the present application;
  • FIG. 6 is a schematic diagram of the fourth sub-flow of the monitoring method for wearing a mask based on edge computing without service provided by the embodiment of the present application;
  • FIG. 7 is a schematic block diagram of a monitoring device for wearing a mask based on edge computing without service provided by an embodiment of the present application.
  • FIG. 8 is a schematic block diagram of a computer device according to an embodiment of the present application.
  • FIG. 1 is a schematic flowchart of a mask-wearing monitoring method based on no-service edge computing provided by an embodiment of the present application
  • FIG. 2 is a mask-wearing monitoring method based on no-service edge computing provided by an embodiment of the present application Schematic diagram of the system architecture in the method.
  • the monitoring method for wearing a mask based on edge computing without service is applied to a local preset smart terminal, wherein the method includes the following steps S11-S13:
  • face images can be collected in real time through preset collection equipment such as cameras configured on the mask monitoring site, and the face images can be transmitted by wired or wireless transmission (such as WIFI or Bluetooth).
  • device is uploaded to the local preset intelligent terminal set on the mask monitoring site, the local preset intelligent terminal is a device with a chip (such as an ARM chip), including a CPU and a GPU processor, and the local preset intelligent terminal acquires the collection
  • the face image is detected, and based on the face image, it is detected whether a mask is worn on the face image, so as to monitor whether the face wears a mask.
  • the preset high-performance neural network inference computing framework NCNN is an optimized high-performance neural network forward computing framework
  • CNN is the abbreviation of convolutional neural network
  • the N at the beginning can contain multiple layers of meaning, for example, N can be described as New/ Next (new implementation), Naive (ncnn is a naive implementation), Neon (ncnn was originally optimized for mobile phones), etc.
  • the default high-performance neural network inference computing framework NCNN is a cross-platform framework without third-party dependencies. Based on NCNN, it can Transplant deep learning algorithms to smart terminals for efficient execution.
  • An edge node is an intelligent terminal with common capabilities such as edge-side real-time data analysis, local data storage, and real-time network connection. The calculation corresponding to the image processing process on the edge node is called edge computing.
  • each local preset smart terminal has the ability to perform edge computing as an edge node, and the edge computing can be independent of the device (it can be across smart terminals such as computers, laptops, mobile phones or tablets), and is compatible with the main Operating system compatible (eg Windows, MacOS, Linux, Android and iOS).
  • the edge computing can be independent of the device (it can be across smart terminals such as computers, laptops, mobile phones or tablets), and is compatible with the main Operating system compatible (eg Windows, MacOS, Linux, Android and iOS).
  • the face image can be sent to the edge computing node on the local intelligent terminal, that is, the local intelligent terminal is also used as the edge computing node , the face image can also be sent to other local edge nodes based on intelligent terminals, such as edge node 1 and edge node 2 in FIG.
  • the preset intelligent terminal improves the processing ability of face images, improves the processing efficiency of face images, and does not require the use of centralized computing devices with powerful computing power such as servers, which can reduce the demand for local centralized computing resources and reduce the need for masks.
  • the local smart terminal can disconnect from the Internet after the face image data is loaded, reduce data exposure, and provide an infrastructure for the storage and use of key private data, which can ensure the security of user data and satisfy privacy concerns.
  • the need for sexuality also avoids legal and security complications.
  • a preset WASM virtual machine may also be deployed on the local preset edge node, so that edge computing is performed on the face image based on the preset high-performance neural network inference computing framework NCNN and the preset WASM virtual machine.
  • the WASM virtual machine WebAssembly-based virtual machine in English
  • the virtual machine can be deployed on the hardware device through the network card device and the virtual network card of the hardware device.
  • WebAssembly (abbreviation WASM) is based on The binary instruction format of the stack virtual machine, WASM can be used to compile high-level languages such as C/C+/RUST, so that client and server applications can be deployed on the Web.
  • WASM is a low-level language that runs in the browser. It is in the form of machine language, which can be recognized faster by the machine and can be compiled directly in C language. Compared with the commonly used Js language, it is easier for the machine to understand, so that the virtual machine can be deployed on the hardware device, which is convenient for edge processing. calculate.
  • the local preset edge node can perform mask detection on the face image based on the preset YOLO deep learning model by establishing a preset YOLO deep learning model, so as to realize the monitoring of wearing a mask, when pre-training the YOLO neural network , select training sample images that only cover the mouth and nose, non-masks and other objects, no occlusion, mask occlusion, etc., and label the training sample images.
  • the labeled training sample images form a training sample image set, and use the training sample image set to
  • the YOLO deep learning model is trained so that the YOLO deep learning model learns the different situations of wearing a mask in the training sample image set.
  • the subsequent YOLO deep learning model judges that the mask is correctly worn according to the images in the training sample image set with masks covering the mouth and nose. , and other incorrectly wearing masks, such as only covering the mouth or nose, it is displayed as no mask, or when other objects cover it, it is still no mask, and the above situations of correctly wearing a mask and incorrectly wearing a mask are used as the judgment rule for wearing a mask. .
  • Using the YOLO deep learning model as the detection method solves the defect of the rapid recognition of the traditional convolutional neural network.
  • the recognition accuracy is higher, but the response is very slow, because there is no preset candidate area and object recognition, traversing the picture It is extremely slow to search for areas of different sizes in a carpet-like manner, which can improve the efficiency and monitoring effect of wearing masks.
  • the local preset terminal After the local preset terminal acquires the face image, it sends the face image to a local preset edge node for edge computing, and the local preset edge node includes the local preset intelligent terminal, that is, the local preset edge node.
  • the intelligent terminal is used as an edge node to perform edge computing
  • the local preset edge node may also include other local preset intelligent terminals, such as other local preset intelligent terminals connected through the Internet of Things, and each local preset edge node is Edge computing can be performed independently, and the corresponding edge computing results can be obtained.
  • the local preset intelligent terminal After the local preset intelligent terminal acquires the face image, it can perform edge computing on the face image by itself, or perform edge computing on the face image through other local edge nodes, to obtain a corresponding edge computing result, And based on the edge calculation result, identify the mask, and judge whether the mask is detected on the face image and whether the mask is correctly worn.
  • the local preset smart terminal obtains the edge calculation result corresponding to the face object, and according to the edge calculation result, judges whether the face image is wearing a mask correctly, and obtains the mask detection result corresponding to the face image. , and the detection result of the mask is displayed to realize the mask monitoring of the face image.
  • the local preset edge node is a local intelligent terminal deployed with a preset high-performance neural network inference computing framework NCNN, so that the service-free (Serverless in English, Serverless) For edge computing with no server architecture), using edge computing, most of the traffic load will be processed at the data source instead of sending all data through the network, network congestion is significantly improved, and then the edge computing results are obtained, according to the edge
  • the detection result of the mask corresponding to the face image is obtained, and the detection result of the mask is displayed, so as to realize the monitoring of wearing a mask on the face image, compared with the mask performed by the service architecture based on cloud computing in the traditional technology.
  • edge computing based on local preset edge nodes can make full use of preset terminals in the local environment, reducing the demand for local centralized computing resources and reducing the need for mask monitoring.
  • the deployment cost reduces the service cost of mask monitoring, improves the convenience of mask monitoring, and improves the efficiency of mask monitoring on face images.
  • FIG. 3 is a schematic diagram of the first sub-flow of the method for monitoring wearing a mask based on edge computing without service provided by the embodiment of the present application.
  • the step of sending the face image to a local preset edge node for edge calculation, and obtaining the edge calculation result corresponding to the face image includes: S121 , obtaining and
  • the locally preset intelligent terminal is based on the preset terminal identifiers corresponding to other terminals that are locally preset in the connected state of the Internet of Things, and according to the preset terminal identifiers, sends the face image to the corresponding preset terminal identifiers.
  • Preset other terminals locally so that the locally preset other terminals perform edge calculation on the face image, obtain an edge calculation result corresponding to the face image, and return the edge calculation result; S122, receive The edge calculation result returned by the other terminal is preset locally to obtain the edge calculation result corresponding to the face image.
  • the Internet of Things (IoT, Internet of things in English) is the "Internet of all things connected", which is an extension and expansion of the Internet based on the Internet.
  • IoT Internet of things in English
  • IP Internet of all things connected
  • different terminals included in the Internet of Things can send data through the network and protocols (such as HTTP).
  • HTTP HyperText Transfer Protocol
  • the hardware devices based on intelligent terminals in the Internet of Things It is equivalent to an edge node in edge computing.
  • the preset high-performance neural network inference computing framework NCNN is a cross-platform framework without third-party dependencies, based on NCNN, deep learning algorithms can be transplanted to intelligent terminals for efficient execution, and the Internet of Things has realized the integration of various Therefore, based on the Internet of Things, the preset high-performance neural network inference computing framework NCNN can be deployed to various terminals in the Internet of Things, so that various terminals in the Internet of Things can become intelligent terminals, which can be used as edge
  • the nodes perform edge computing, so as to make full use of various terminals in the local environment, including various intelligent terminals and various hardware devices such as desktop computers and notebook computers, and use various terminals in the local environment as edge computing nodes, so that the local Various terminals in the environment can perform edge computing independently, thereby realizing intelligent terminals such as computers, laptops, mobile phones or tablets, and are compatible with major operating systems (such as Windows, MacOS, Linux, Android, and IOS systems).
  • the locally preset smart terminal can obtain other locally preset terminals that are connected to the locally preset smart terminal based on the Internet of Things, and use the
  • the face image is sent to the locally preset other terminal, so that the locally preset other terminal performs edge calculation on the face image, obtains the edge calculation result corresponding to the face image, and performs the edge calculation on the edge calculation result.
  • the locally preset intelligent terminal receives the edge calculation result returned by the locally preset other intelligent terminal, so as to obtain the edge calculation result corresponding to the face image, for example, please continue to refer to FIG.
  • the face image is sent to edge node 1 and edge node 2 to further make full use of various terminals in the local environment. Since the mask monitoring of the face image is only calculated locally, the pressure on network bandwidth is reduced, and the The computing power of face image monitoring for masks further improves the efficiency of mask monitoring on face images.
  • FIG. 4 is a schematic diagram of a second sub-flow of the method for monitoring wearing a mask based on edge computing without service provided by the embodiment of the present application.
  • the step of sending the face image to other locally preset terminals corresponding to the preset terminal identifiers includes: S1211: Acquire a plurality of the locally preset smart terminals that are connected to the locally preset intelligent terminal based on the Internet of Things
  • the preset terminal identifiers corresponding to other terminals S1212, according to the preset time period, determine the preset terminal identifiers that have not sent the face image within the preset time period as the target terminal identifier; S1213, send the face image to the locally preset other terminal corresponding to the target terminal identifier
  • the locally preset smart terminal is connected to a plurality of the locally preset other terminals based on the Internet of Things
  • the face images are concentrated in some of the locally preset other terminals for processing, and different face images can be sent to the device in turn through a preset time period, such as 20 minutes or half an hour.
  • a preset time period such as 20 minutes or half an hour.
  • Different locally preset other terminals perform edge computing, and when the locally preset smart terminal acquires a face image and sends the face image, it can be determined according to a preset time period that there has not been an image within the preset time period.
  • the preset terminal identifier to which the face image is sent is the target terminal identifier, and the target terminal identifier corresponds to other locally preset terminals, and the locally preset other terminals have not received the transmission from the locally preset intelligent terminal within a preset time period. That is, the locally preset other terminal is not sent a face image by the locally preset intelligent terminal within a preset time period, and the face image is sent to the locally preset other terminal corresponding to the target terminal identifier.
  • the preset terminal identifier corresponding to other terminals preset locally may be randomly selected as the target terminal identifier, or the local preset identifier may be determined according to a certain preset order.
  • Set the preset terminal identification corresponding to the other terminal as the target terminal identification and send the face image to the local preset other terminal corresponding to the target terminal identification, so as to balance the processing of face images by different local preset other terminals. It can make full use of the computing power of different local presets and other terminals to improve the efficiency of mask monitoring.
  • FIG. 5 is a schematic diagram of a third sub-flow of the method for monitoring wearing a mask based on edge computing without service provided by the embodiment of the present application.
  • the steps of obtaining the mask detection result corresponding to the face image according to the edge calculation result, and displaying the mask detection result include: S131 , obtaining the mask detection result.
  • the mask recognition accuracy rate included in the edge calculation result, the mask recognition accuracy rate is used to describe the probability that the nose and lips of the face image are both blocked by the mask; S132, determine whether the mask recognition accuracy rate is greater than or equal to a predetermined rate.
  • the deep learning model can be deployed on the terminal corresponding to the edge computing node.
  • the deep learning model can be the Yolo deep learning model, so that the edge computing node can perform edge computing on the face image based on the Yolo deep learning model.
  • the edge computing node performs edge computing on the face image based on the Yolo deep learning model
  • the Yolo deep learning model can detect various objects, for example, it can be set to detect various objects such as masks, paper towels, mobile phones and water cups.
  • the Yolo deep learning model will output the probability of each object contained in the face image.
  • the Yolo deep learning model will detect that the face image contains masks, tissues, mobile phones, water cups, etc.
  • the Yolo deep learning model will output the probability that the object is included in the face image, and the Yolo deep learning model can further determine the probability of each object identified.
  • the face image uses cloth or paper items (such as masks or tissues, etc.) to cover the lips (that is, the mouth) and nose, if the mouth and nose are covered by cloth or paper items, the depth of Yolo
  • the learning model will output the mask-wearing accuracy rate corresponding to the face image (that is, the mask recognition accuracy rate), and if the mouth and nose are covered by cloth or paper items, the output mask-wearing accuracy rate is relatively high.
  • the local preset smart terminal obtains the edge calculation result corresponding to the face object, and according to the edge calculation result, obtains the mask recognition accuracy included in the edge calculation result, and judges the mask Whether the recognition accuracy is greater than or equal to the preset accuracy threshold, if the mask recognition accuracy is greater than or equal to the preset accuracy threshold, it is determined that the face image is correctly wearing a mask, if the mask recognition accuracy is less than the preset accuracy rate threshold, determine that the face image does not wear a mask correctly, so as to determine whether the face image is wearing a mask correctly, and obtain the mask detection result corresponding to the face image, because WASM is based on the binary instruction format of the stack virtual machine.
  • the client can be deployed on the Web, and the mask detection result can be displayed through the preset browser based on the preset WASM virtual machine, so as to realize the mask monitoring of the face image, which is compared with the traditional technology of most devices.
  • the function is mainly used to detect people wearing masks or other obstacles, and it is detected by face recognition, and it is rarely detected whether a mask is worn directly. Improve the accuracy and efficiency of mask monitoring.
  • the edge computing result can be obtained based on the edge computing without service, and then according to the edge computing result, the mask detection result corresponding to the face image is obtained, and the The mask detection results are displayed through a preset browser.
  • the mask detection results can be further displayed through a preset browser based on the preset WASM virtual machine, so as to realize the monitoring of face images wearing masks, so as to perform mask monitoring.
  • the step of obtaining the edge calculation result obtaining the mask detection result corresponding to the face image according to the edge calculation result, and displaying the mask detection result, it also includes: Obtain the mask monitoring result and preset target data corresponding to the mask monitoring result, and upload the mask monitoring result and the preset target data to a preset server.
  • the preset server can save the valuable data, such as the mask monitoring time corresponding to the mask monitoring result, the mask monitoring location, the mask monitoring object and whether the mask is properly worn, especially in important cases.
  • the mask monitoring time corresponding to the monitoring result of the user not wearing the mask correctly the mask monitoring locations and mask monitoring objects are uploaded to the preset server for storage to retain evidence.
  • FIG. 6 is a schematic diagram of the fourth sub-flow of the method for monitoring wearing a mask based on edge computing without service provided by the embodiment of the present application.
  • the mask monitoring result and the preset target data corresponding to the mask monitoring result are obtained, and the mask monitoring result and the preset target data are uploaded to a preset
  • the steps of the server include: S61, judging whether the mask monitoring result is that the mask is not properly worn; S62, if the mask monitoring result is that the mask is not properly worn, the mask monitoring result of the incorrectly wearing the mask and the corresponding preset target data are Upload to the preset server; S63, if the mask monitoring result is that the mask is correctly worn, do not upload the mask monitoring result and the corresponding preset target data for incorrectly wearing the mask to the preset server.
  • the face image is sent to a local preset edge device, so that the preset edge device calculates the face image, obtains a corresponding edge calculation result, and returns the edge calculation result.
  • the preset edge device calculates the face image, obtains a corresponding edge calculation result, and returns the edge calculation result.
  • the mask monitoring result is that the mask is not properly worn, and if the mask monitoring result is that the mask is not properly worn, the mask monitoring result of incorrectly wearing the mask and
  • the corresponding preset target data is uploaded to the preset server. If the mask monitoring result is that the mask is not properly worn, the mask monitoring result of the incorrectly wearing the mask and the corresponding preset target data are uploaded to the preset server, so as to further screen out the mask. It is necessary to upload and store the data stored to the server, which can further reduce the data uploaded to the server, further reduce the pressure on network bandwidth, and reduce the demand for storage resources. And these data are only used for one detection, so there is no need to store them.
  • the step of uploading the mask monitoring result of incorrectly wearing the mask and the corresponding preset target data to the preset server if the mask monitoring result is that the mask is not properly worn it also includes: if the mask is not properly worn.
  • the mask monitoring result is that the mask is not properly worn, and an alarm is issued for the face image.
  • the judgment rule corresponding to whether the mask is correctly worn it is judged whether the face corresponding to the face image is wearing a mask correctly.
  • the mask monitoring results corresponding to the face image can be used. Displayed through the browser, the output results of the mask monitoring corresponding to the face image include the accuracy of mask recognition, the Bounding box of the YOLO deep learning model, and the detection results. If the face does not wear a mask correctly, the face The image issues an alarm, which can be displayed as no mask and a warning color such as an orange frame or a red frame, which means that the wearer is not wearing the mask correctly.
  • the mask In the application scenario of the entrance, such as at the entrance of a hospital or an office building, the mask is not properly worn. If the face is wearing a mask correctly, no alarm will be issued for the face image, it will show that there is a mask and a blue box can appear. Further, in the application scenario of the entrance, wear the mask correctly Persons who are allowed to pass can take measures to release and pass, so as to realize automatic control of the passage and improve the efficiency of passage control.
  • FIG. 7 is a schematic block diagram of a monitoring device for wearing a mask based on edge computing without service provided by an embodiment of the present application.
  • the embodiment of the present application further provides a monitoring device for wearing a mask based on no-service edge computing.
  • the monitoring device for wearing a mask based on edge computing without service includes a unit for executing the method for monitoring wearing a mask based on edge computing described above, and the monitoring device for monitoring wearing a mask based on edge computing can be Configured in the local preset smart terminal. Specifically, please refer to FIG.
  • the mask wearing monitoring device 70 based on no-service edge computing includes a first acquisition unit 71 , a calculation unit 72 and a display unit 73 .
  • the first acquisition unit 71 is used to acquire the face image collected by the local preset collection device;
  • the calculation unit 72 is used to send the face image to the local preset edge node for edge calculation to obtain the face image
  • the edge computing result corresponding to the image wherein the local preset edge node is a local intelligent terminal that deploys a preset high-performance neural network inference computing framework NCNN;
  • the display unit 73 is used to obtain the edge computing result, according to the According to the edge calculation result, the mask detection result corresponding to the face image is obtained, and the mask detection result is displayed.
  • the computing unit 72 includes: a first sending sub-unit, configured to obtain a preset terminal identifier corresponding to another locally preset terminal in which the locally preset smart terminal is in a connected state based on the Internet of Things, according to the For the preset terminal identification, the face image is sent to the locally preset other terminal corresponding to the preset terminal identification, so that the locally preset other terminal performs edge calculation on the face image, and obtains the result. the edge calculation result corresponding to the face image, and return the edge calculation result; the receiving subunit is used to receive the edge calculation result returned by the locally preset other terminal, and obtain the corresponding edge calculation result of the face image. Edge computing results.
  • the first sending subunit includes: a first obtaining subunit, configured to obtain a plurality of locally preset other terminals corresponding to the locally preset smart terminal in a connected state based on the Internet of Things. a preset terminal identifier; a determination subunit for determining, according to a preset time period, a preset terminal identifier that has not sent a face image within the preset time period as a target terminal identifier; a second sending subunit for sending The face image is sent to other locally preset terminals corresponding to the target terminal identifier.
  • the display unit 73 includes: a second acquisition subunit for acquiring the mask recognition accuracy included in the edge calculation result; a first judgment subunit for judging whether the mask recognition accuracy is is greater than or equal to the preset accuracy threshold; the first determination subunit is used to determine that the face image is correctly wearing a mask if the mask recognition accuracy is greater than or equal to the preset accuracy threshold; the second determination subunit is used for If the mask recognition accuracy is less than the preset accuracy threshold, it is determined that the face image does not wear a mask correctly; a display subunit is used to display the face image correctly wearing a mask or the face image is not correctly wearing a mask The detection results of masks are displayed through the preset browser.
  • the mask-wearing monitoring device 70 based on no-service edge computing further includes: an uploading unit for acquiring the mask monitoring results and preset target data corresponding to the mask monitoring results, and uploading the mask monitoring results.
  • the mask monitoring results and the preset target data are uploaded to the preset server.
  • the uploading unit includes: a second judging subunit for judging whether the mask monitoring result is that the mask is not properly worn; the uploading subunit is used for if the mask monitoring result is that the mask is not properly worn , and upload the mask monitoring results of incorrectly wearing masks and the corresponding preset target data to the preset server.
  • the uploading unit further includes: an alarm subunit, configured to issue an alarm to the face image if the mask monitoring result is that the mask is not properly worn.
  • the division and connection method of each unit in the above-mentioned no-service edge computing-based mask-wearing monitoring device are only used for illustration.
  • the no-service-edge computing-based mask wearing monitoring device can be divided into different types according to needs. It is also possible to adopt different connection sequences and methods for each unit in the monitoring device for wearing a mask based on edge computing without service, so as to complete all or part of the functions of the monitoring device for wearing a mask based on edge computing without service.
  • the above-mentioned monitoring device for wearing a mask based on serverless edge computing can be implemented in the form of a computer program, and the computer program can run on a computer device as shown in FIG. 8 .
  • the computer device 500 may be a computer device such as a notebook computer, a tablet computer, or other smart terminals, or may be a component or component in other devices.
  • the computer device 500 includes a processor 502, a memory and a network interface 505 connected through a system bus 501, wherein the memory may include a non-volatile storage medium 503 and an internal memory 504, and the memory may also be volatile sexual storage medium.
  • the nonvolatile storage medium 503 can store an operating system 5031 and a computer program 5032 .
  • the computer program 5032 When executed, it can cause the processor 502 to execute the above-mentioned method for monitoring wearing a mask based on edge computing without service.
  • the processor 502 is used to provide computing and control capabilities to support the operation of the entire computer device 500 .
  • the internal memory 504 provides an environment for running the computer program 5032 in the non-volatile storage medium 503.
  • the processor 502 can execute the above-mentioned non-service edge computing-based mask wearing monitoring method.
  • the network interface 505 is used for network communication with other devices.
  • the structure shown in FIG. 8 is only a block diagram of a partial structure related to the solution of the present application, and does not constitute a limitation on the computer device 500 to which the solution of the present application is applied.
  • the specific computer device 500 may include more or fewer components than shown, or combine certain components, or have a different arrangement of components.
  • the computer device may only include a memory and a processor. In such an embodiment, the structures and functions of the memory and the processor are consistent with the embodiment shown in FIG. 8 , which will not be repeated here.
  • the processor 502 is configured to run the computer program 5032 stored in the memory, so as to implement the method for monitoring wearing a mask based on edge computing without service described in the embodiments of the present application.
  • the processor 502 may be a central processing unit (Central Processing Unit, CPU), and the processor 502 may also be other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), Application Specific Integrated Circuit (ASIC), Field-Programmable Gate Array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.
  • the general-purpose processor can be a microprocessor or the processor can also be any conventional processor or the like.
  • the present application also provides a computer-readable storage medium.
  • the computer-readable storage medium may be a non-volatile computer-readable storage medium, or a volatile computer-readable storage medium, and the computer-readable storage medium stores a computer program, and the computer program is executed by the processor At the same time, the processor is caused to execute the steps of the method for monitoring wearing a mask based on edge computing without service described in the above embodiments.
  • the storage medium is a physical, non-transitory storage medium, such as a U disk, a removable hard disk, a read-only memory (Read-Only Memory, ROM), a magnetic disk or an optical disk and other physical storage that can store computer programs. medium.
  • a physical, non-transitory storage medium such as a U disk, a removable hard disk, a read-only memory (Read-Only Memory, ROM), a magnetic disk or an optical disk and other physical storage that can store computer programs. medium.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Molecular Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • Human Computer Interaction (AREA)
  • Multimedia (AREA)
  • Image Analysis (AREA)

Abstract

一种基于无服务边缘计算的佩戴口罩监测方法、装置、计算机设备及计算机可读存储介质，通过本地预设智能终端，获取本地预设采集设备采集的人脸图像（S11），将人脸图像发送至本地预设边缘节点进行边缘计算，得到人脸图像对应的边缘计算结果（S12），获取边缘计算结果，根据边缘计算结果，得到人脸图像对应的口罩检测结果，将口罩检测结果进行显示（S13）。

Description

基于无服务边缘计算的佩戴口罩监测方法、装置及设备
本申请要求于2021年04月30日提交中国专利局、申请号为202110481307.8、申请名称为“基于无服务边缘计算的佩戴口罩监测方法、装置及设备”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及人工智能技术领域，尤其涉及一种基于无服务边缘计算的佩戴口罩监测方法、装置、计算机设备及计算机可读存储介质。
背景技术
疫情防控是目前国内一个重大的医疗挑战，针对于目前的疾病(如流感，新冠病毒等)，佩戴口罩是阻断呼吸道感染的最主要方法之一。因此，许多公司研发出口罩监测仪器(平台)，并投放在在医院、机场等公共场所中使用，以便对人员佩戴口罩提供有效的监督措施。
然而，目前的检测系统大多基于云计算方案，然而云计算解决方案受到实时摄像机视频流互联网带宽的限制，因此需要配有相关的数据库，硬件层计算的数据以及实时拍摄的图片需要上传到虚拟云层进行虚拟化处理进行备份，再上传到数据库中进行存储。尤其是在人流量比较大的地方，每天会得到非常大的数据，发明人意识到，基于云计算的口罩检测方式，增加了网络压力，导致系统延迟，降低了对佩戴口罩进行监测的效率。
发明内容
本申请提供了一种基于无服务边缘计算的佩戴口罩监测方法、装置、计算机设备及计算机可读存储介质，能够解决传统技术中佩戴口罩监测效率较低的技术问题。
第一方面，本申请提供了一种基于无服务边缘计算的佩戴口罩监测方法，应用于本地预设智能终端，其中，所述方法包括：获取本地预设采集设备采集的人脸图像；将所述人脸图像发送至本地预设边缘节点进行边缘计算，得到所述人脸图像对应的边缘计算结果，其中，所述本地预设边缘节点为部署了预设高性能神经网络推理计算框架NCNN的本地智能终端；获取所述边缘计算结果，根据所述边缘计算结果，得到所述人脸图像对应的口罩检测结果，并将所述口罩检测结果进行显示。
第二方面，本申请还提供了一种基于无服务边缘计算的佩戴口罩监测装置，应用于本地预设智能终端，其中，所述装置包括：第一获取单元，用于获取本地预设采集设备采集的人脸图像；计算单元，用于将所述人脸图像发送至本地预设边缘节点进行边缘计算，得到所述人脸图像对应的边缘计算结果，其中，所述本地预设边缘节点为部署了预设高性能神经网络推理计算框架NCNN的本地智能终端；显示单元，用于获取所述边缘计算结果，根据所述边缘计算结果，得到所述人脸图像对应的口罩检测结果，并将所述口罩检测结果进行显示。
第三方面，本申请还提供了一种计算机设备，其包括存储器及处理器，所述存储器上存储有计算机程序，所述处理器执行所述计算机程序时实现如下步骤：获取本地预设采集设备采集的人脸图像；将所述人脸图像发送至本地预设边缘节点进行边缘计算，得到所述人脸图像对应的边缘计算结果，其中，所述本地预设边缘节点为部署了预设高性能神经网络推理计算框架NCNN的本地智能终端；获取所述边缘计算结果，根据所述边缘计算结果，得到所述人脸图像对应的口罩检测结果，并将所述口罩检测结果进行显示。
第四方面，本申请还提供了一种计算机可读存储介质，所述计算机可读存储介质存储有计算机程序，所述计算机程序被处理器执行如下步骤：获取本地预设采集设备采集的人脸图像；将所述人脸图像发送至本地预设边缘节点进行边缘计算，得到所述人脸图像对应的边缘计算结果，其中，所述本地预设边缘节点为部署了预设高性能神经网络推理计算框架NCNN的本地智能终端；获取所述边缘计算结果，根据所述边缘计算结果，得到所述人脸图像对应的口罩检测结果，并将所述口罩检测结果进行显示。
本申请提供了一种基于无服务边缘计算的佩戴口罩监测方法、装置、计算机设备及计算 机可读存储介质，通过本地预设智能终端，获取本地预设采集设备采集的人脸图像，并将所述人脸图像发送至本地预设边缘节点进行边缘计算，得到所述人脸图像对应的边缘计算结果，从而不使用服务架构，即可实现基于无服务(英文为Serverless，为无服务架构)的边缘计算，使用边缘计算，大部分流量负载将通过在数据源端处理数据而不是通过网络发送所有数据，网络拥堵明显改善，再获取所述边缘计算结果，根据所述边缘计算结果，得到所述人脸图像对应的口罩检测结果，并将所述口罩检测结果进行显示，以实现对人脸图像佩戴口罩的监测，相比传统技术中基于云计算的服务架构进行的口罩监测，极大减轻了网络带宽的压力，同时，基于本地预设边缘节点进行边缘计算，可以充分利用本地环境中的预设终端，减少了对本地集中计算资源的需求，降低了进行口罩监测的部署成本，提高了进行口罩监测的便利性，提高了对人脸图像进行口罩监测的效率。
附图说明
为了更清楚地说明本申请实施例技术方案，下面将对实施例描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图是本申请的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。
图1为本申请实施例提供的基于无服务边缘计算的佩戴口罩监测方法的一个流程示意图；
图2为本申请实施例提供的基于无服务边缘计算的佩戴口罩监测方法中系统架构示意图；
图3为本申请实施例提供的基于无服务边缘计算的佩戴口罩监测方法的第一个子流程示意图；
图4为本申请实施例提供的基于无服务边缘计算的佩戴口罩监测方法的第二个子流程示意图；
图5为本申请实施例提供的基于无服务边缘计算的佩戴口罩监测方法的第三个子流程示意图；
图6为本申请实施例提供的基于无服务边缘计算的佩戴口罩监测方法的第四个子流程示意图；
图7为本申请实施例提供的基于无服务边缘计算的佩戴口罩监测装置的一个示意性框图；以及
图8为本申请实施例提供的计算机设备的示意性框图。
具体实施方式
下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例是本申请一部分实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本申请保护的范围。
请参阅图1与图2，图1为本申请实施例提供的基于无服务边缘计算的佩戴口罩监测方法的一个流程示意图，图2为本申请实施例提供的基于无服务边缘计算的佩戴口罩监测方法中系统架构示意图。如图1所示，基于无服务边缘计算的佩戴口罩监测方法，应用于本地预设智能终端，其中，所述方法包括以下步骤S11-S13：
S11、获取本地预设采集设备采集的人脸图像。
具体地，请参阅图2，进行口罩监测，可以通过口罩监测现场配置的摄像机等预设采集设备实时的采集人脸图像，并将人脸图像通过有线传输方式或者无线传输方式(例如WIFI或者蓝牙设备)上传至口罩监测现场设置的本地预设智能终端，所述本地预设智能终端为带有芯片(例如ARM芯片)、包括CPU及GPU处理器的设备，所述本地预设智能终端获取采集的人脸图像，并基于所述人脸图像检测人脸图像上是否佩戴有口罩，以实现对人脸是否佩戴口罩进行监测。
S12、将所述人脸图像发送至本地预设边缘节点进行边缘计算，得到所述人脸图像对应的边缘计算结果，其中，所述本地预设边缘节点为部署了预设高性能神经网络推理计算框架NCNN的本地智能终端。
其中，预设高性能神经网络推理计算框架NCNN，为优化的高性能神经网络前向计算框架，CNN为卷积神经网络的缩写，开头的N可以包含多层含义，例如N可以描述为New/Next(全新的实现)，Naive(ncnn是naive实现)，Neon(ncnn最初为手机优化)等，预设高性能神经网络推理计算框架NCNN为无第三方依赖且跨平台的框架，基于NCNN，能够将深度学习算法移植到智能终端高效执行。边缘节点为具备边缘侧实时数据分析、本地数据存储、实时网络联接等共性能力的智能终端，边缘节点上对图像的处理过程对应的计算称为边缘计算。
具体地，由于基于预设高性能神经网络推理计算框架NCNN，能够将深度学习算法移植到智能终端高效执行，将本地的每个预设智能终端上均部署预设高性能神经网络推理计算框架NCNN，本地的每个预设智能终端各自都具备了作为边缘节点进行边缘计算的能力，且进行边缘计算可以与设备无关(可以跨计算机、笔记本电脑、手机或平板电脑等智能终端)，并与主要操作系统兼容(例如Windows、MacOS、Linux、Android和iOS)。将所述人脸图像发送至本地预设边缘节点进行边缘计算，请继续参阅图2，可以将所述人脸图像发送至本地智能终端上的边缘计算节点，即将本地智能终端同时作为边缘计算节点，也可以将所述人脸图像发送至本地其它基于智能终端的边缘节点，例如图2中的边缘节点1及边缘节点2，从而通过各个边缘对所述人脸图像进行计算，得到所述人脸图像对应的边缘计算结果，由于边缘计算靠近数据源端进行数据处理，能够大大地减少系统延迟，由于人脸图像数据收集与处理的时间间隔几乎是实时的，不但可以充分利用本地的多个预设智能终端提高对人脸图像的处理能力，提高对人脸图像的处理效率，而且不需要借助服务器等具有强大计算能力的集中计算设备，能够降低对本地集中计算资源的需求，降低进行口罩监测的部署成本与系统的复杂性，同时，由于本申请实施例的口罩监测是将所有用户数据进行本地处理，无需通过网络将人脸图像等用户数据进行远距离传输至异地端，为了保护个人隐私，本地智能终端可以在人脸图像数据在加载后与Internet断开连接，减少数据暴露，为关键性隐私数据的存储与使用提供了基础设施，能够保障用户数据的安全性，满足对隐私安全性的需求，也避免了法律和安全复杂性的情况。
进一步地，所述本地预设边缘节点上还可以部署预设WASM虚拟机，从而基于预设高性能神经网络推理计算框架NCNN与预设WASM虚拟机将所述人脸图像进行边缘计算。其中，所述WASM虚拟机(英文为基于WebAssembly的虚拟机)为运行在Web平台上的Assembly，虚拟机可以通过网卡设备和硬件设备的虚拟网卡部署在硬件设备上，WebAssembly(缩写WASM)是基于堆栈虚拟机的二进制指令格式，WASM可用于编译C/C+/RUST等高级语言，使客户端和服务器应用程序能够在Web上部署，因此，WASM是一种在浏览器中运行的低级语言，它是机器语言的形式，对于机器来说能够更快的识别，可以直接使用C语言进行编译，相比于常用的Js语言，机器更容易理解，从而将虚拟机部署在硬件设备上，便于进行边缘计算。
其中，所述本地预设边缘节点可以通过建立预设YOLO深度学习模型，基于训练的预设YOLO深度学习模型将人脸图像进行口罩检测，以实现佩戴口罩监测，在预先训练YOLO神经网络的时候，选取只遮挡口鼻、非口罩等物体遮挡、无遮挡、口罩遮挡等训练样本图像，并将训练样本图像进行标注，由标注后的训练样本图像组成训练样本图像集，利用训练样本图像集对YOLO深度学习模型进行训练，以使YOLO深度学习模型学习到训练样本图像集中戴口罩的不同情形，后续YOLO深度学习模型根据训练样本图像集中包含的有口罩遮挡口与鼻的图像判断为正确戴口罩，其它未正确戴口罩，例如只遮挡口或鼻的情形，显示为无口罩，或者其它物体遮挡时依旧为无口罩，将上述正确佩戴口罩与未正确佩戴口罩的情形作为佩戴口罩监测的判断规则。利用YOLO深度学习模型作为检测方法，解决了传统卷积神经网络的快速识别的缺陷，相比传统的CNN识别精准度较高，但是响应很慢，因为没有预先设置候选区和对象识别，遍历图片中所有可能的位置，地毯式搜索不同大小的区域来检测，效率极慢，能够提高佩戴口罩监测的效率与监测效果。
本地预设终端获取所述人脸图像后，将所述人脸图像发送至本地预设边缘节点进行边缘计算，所述本地预设边缘节点包括所述本地预设智能终端，即将所述本地预设智能终端作为边缘节点进行边缘计算，所述本地预设边缘节点还可以包括本地的其它预设智能终端，例如通过物联网进行连接的本地其它预设智能终端，每个本地预设边缘节点均可以独立进行边缘计算，得到各自对应的边缘计算结果。所述本地预设智能终端获取人脸图像后，可以通过自身将所述人脸图像进行边缘计算，也可以通过本地其它边缘节点将所述人脸图像进行边缘计算，得到对应的边缘计算结果，并基于边缘计算结果进行识别口罩，判断所述人脸图像上是否检测到口罩及是否正确佩戴口罩。
S13、获取所述边缘计算结果，根据所述边缘计算结果，得到所述人脸图像对应的口罩检测结果，并将所述口罩检测结果进行显示。
具体地，本地预设智能终端，获取所述人脸对象对应的边缘计算结果，根据所述边缘计算结果，判断所述人脸图像是否正确佩戴口罩，得到所述人脸图像对应的口罩检测结果，并将所述口罩检测结果进行显示，实现对人脸图像的口罩监测。
本申请实施例，通过基于本地预设智能终端，获取本地预设采集设备采集的人脸图像，并将所述人脸图像发送至本地预设边缘节点进行边缘计算，得到所述人脸图像对应的边缘计算结果，其中，所述本地预设边缘节点为部署有预设高性能神经网络推理计算框架NCNN的本地的智能终端，从而不使用服务架构，即可实现基于无服务(英文为Serverless，为无服务架构)的边缘计算，使用边缘计算，大部分流量负载将通过在数据源端处理数据而不是通过网络发送所有数据，网络拥堵明显改善，再获取所述边缘计算结果，根据所述边缘计算结果，得到所述人脸图像对应的口罩检测结果，并将所述口罩检测结果进行显示，以实现对人脸图像佩戴口罩的监测，相比传统技术中基于云计算的服务架构进行的口罩监测，极大减轻了网络带宽的压力，同时，基于本地预设边缘节点进行边缘计算，可以充分利用本地环境中的预设终端，减少了对本地集中计算资源的需求，降低了进行口罩监测的部署成本，降低了进行口罩监测的服务成本，提高了进行口罩监测的便利性，提高了对人脸图像进行口罩监测的效率。
在一实施例中，请参阅图3，图3为本申请实施例提供的基于无服务边缘计算的佩戴口罩监测方法的第一个子流程示意图。如图3所示，在该实施例中，所述将所述人脸图像发送至本地预设边缘节点进行边缘计算，得到所述人脸图像对应的边缘计算结果的步骤包括：S121、获取与所述本地预设智能终端基于物联网处于连接状态的本地预设其它终端对应的预设终端标识，根据所述预设终端标识，将所述人脸图像发送至所述预设终端标识对应的本地预设其它终端，以使所述本地预设其它终端对所述人脸图像进行边缘计算，得到所述人脸图像对应的边缘计算结果，并将所述边缘计算结果进行返回；S122、接收所述本地预设其它终端返回的所述边缘计算结果，得到所述人脸图像对应的边缘计算结果。
其中，物联网(IoT，英文为Internet of things)即“万物相连的互联网”，是互联网基础上的延伸和扩展的网络，将各种信息传感设备与网络结合起来而形成的一个巨大网络，实现在任何时间、任何地点，人、机、物的互联互通，物联网包含的不同终端可以通过网络以及协议(例如HTTP)发送数据，本申请实施例，物联网中基于智能终端的硬件设备在边缘计算中相当于边缘节点。
具体地，由于预设高性能神经网络推理计算框架NCNN为无第三方依赖且跨平台的框架，因此，基于NCNN，能够将深度学习算法移植到智能终端高效执行，而物联网又实现了将各种终端进行连接，因此，可以基于物联网，将预设高性能神经网络推理计算框架NCNN部署到物联网中的各种终端上，以使物联网中的各种终端成为智能终端，能够作为边缘节点进行边缘计算，从而充分利用本地环境中的各种终端，包括各种智能终端及台式机电脑、笔记本电脑等各种硬件设备，将本地环境中的各种终端作为边缘计算节点，以使本地环境中的各种终端可以各自独立进行边缘计算，从而实现跨计算机、笔记本电脑、手机或平板电脑等智能终端，并与主要操作系统兼容(例如Windows、MacOS、Linux、Android和IOS系统)。因此， 进行口罩监测，尤其是在口罩监测的数据量大时，所述本地预设智能终端可以获取与所述本地预设智能终端基于物联网处于连接状态的本地预设其它终端，将所述人脸图像发送至本地预设其它终端，以使所述本地预设其它终端将所述人脸图像进行边缘计算，得到所述人脸图像对应的边缘计算结果，并将所述边缘计算结果进行返回，所述本地预设智能终端接收所述本地预设其它智能终端返回的所述边缘计算结果，从而得到所述人脸图像对应的边缘计算结果，例如，请继续参阅图2，可以将所述人脸图像发送至边缘节点1及边缘节点2，以进一步充分利用本地环境中的各种终端，由于对人脸图像进行口罩监测只是在本地进行计算，减轻了网络带宽的压力，提高对人脸图像进行口罩监测的运算能力，进一步提高了对人脸图像进行口罩监测的效率。
在一实施例中，请参阅图4，图4为本申请实施例提供的基于无服务边缘计算的佩戴口罩监测方法的第二个子流程示意图。如图4所示，在该实施例中，所述获取与所述本地预设智能终端基于物联网处于连接状态的本地预设其它终端对应的预设终端标识，根据所述预设终端标识，将所述人脸图像发送至所述预设终端标识对应的本地预设其它终端的步骤包括：S1211、获取与所述本地预设智能终端基于物联网处于连接状态的多个所述本地预设其它终端对应的预设终端标识；S1212、根据预设时间周期，确定所述预设时间周期内未被发送人脸图像的预设终端标识为目标终端标识；S1213、将所述人脸图像发送至所述目标终端标识对应的本地预设其它终端。
具体地，针对所述本地预设智能终端基于物联网连接多个所述本地预设其它终端的情形，为了充分利用多个所述本地预设其它终端的运算能力来提高口罩监测的效率，避免将人脸图像集中于某些所述本地预设其它终端进行处理，可以通过预设时间周期，例如20分钟或者半个小时，在该预设时间周期内，将不同的人脸图像轮流发送至不同的本地预设其它终端进行边缘计算，所述本地预设智能终端获取人脸图像，并将所述人脸图像进行发送时，可以根据预设时间周期，确定该预设时间周期内还未被发送人脸图像的预设终端标识为目标终端标识，所述目标终端标识对应本地预设其它终端，该本地预设其它终端在预设时间周期内未接收到所述本地预设智能终端发送的人脸图像，即本地预设其它终端在预设时间周期内未被所述本地预设智能终端发送人脸图像，将所述人脸图像发送至所述目标终端标识对应的本地预设其它终端，若该预设时间周期内不存在未发送人脸图像的预设终端标识，可以随机选取本地预设其它终端对应的预设终端标识作为目标终端标识，或者按照一定预设顺序确定本地预设其它终端对应的预设终端标识作为目标终端标识，并将所述人脸图像发送至所述目标终端标识对应的本地预设其它终端，从而平衡不同的本地预设其它终端处理人脸图像的数量，可以充分利用不同本地预设其它终端的运算能力，提高口罩监测的效率。
在一实施例中，请参阅图5，图5为本申请实施例提供的基于无服务边缘计算的佩戴口罩监测方法的第三个子流程示意图。如图5所示，在该实施例中，所述根据所述边缘计算结果，得到所述人脸图像对应的口罩检测结果，并将所述口罩检测结果进行显示的步骤包括：S131、获取所述边缘计算结果包含的口罩识别准确率，所述口罩识别准确率用于描述所述人脸图像的鼻子与嘴唇均被口罩遮挡的概率；S132、判断所述口罩识别准确率是否大于或者等于预设准确率阈值；S133、若所述口罩识别准确率大于或者等于预设准确率阈值，判定所述人脸图像正确佩戴口罩；S134、若所述口罩识别准确率小于预设准确率阈值，判定所述人脸图像未正确佩戴口罩；S135、将所述人脸图像正确佩戴口罩或者所述人脸图像未正确佩戴口罩的检测结果通过预设浏览器进行显示。
具体地，可以将深度学习模型部署在边缘计算节点对应的终端，例如，所述深度学习模型可以为Yolo深度学习模型，从而边缘计算节点可以基于Yolo深度学习模型对所述人脸图像进行边缘计算。边缘计算节点若基于Yolo深度学习模型对所述人脸图像进行边缘计算，由于Yolo深度学习模型可以对多种物体进行检测，例如可以设定为检测口罩、纸巾、手机及水杯等多种物体，针对设定的每种物体，Yolo深度学习模型会输出所述人脸图像中包含每种物体的概率，例如，Yolo深度学习模型会检测所述人脸图像中包含口罩、纸巾、手机及水杯等 多种物体各自的概率，从而针对每一种物体，Yolo深度学习模型会输出所述人脸图像中包含该物体的概率，Yolo深度学习模型根据识别出的每种物体的概率，再可以进一步判断所述人脸图像是否利用了布制或者纸质的物品(如口罩或者纸巾等)遮挡住嘴唇(即口)和鼻子，如果口和鼻子均被布制或者纸质的物品遮挡住了，Yolo深度学习模型就会输出所述人脸图像对应的戴口罩的准确率(即口罩识别准确率)，且如果口和鼻子均被布制或者纸质的物品遮挡住了，输出的戴口罩的准确率相对较高，若输出的戴口罩的准确率大于或者等于预设准确率阈值，就认为是正确佩戴了口罩，否则就认为没有正确佩戴口罩。因此，请继续参阅图2，本地预设智能终端通过获取所述人脸对象对应的边缘计算结果，根据所述边缘计算结果，获取所述边缘计算结果包含的口罩识别准确率，判断所述口罩识别准确率是否大于或者等于预设准确率阈值，若所述口罩识别准确率大于或者等于预设准确率阈值，判定所述人脸图像正确佩戴口罩，若所述口罩识别准确率小于预设准确率阈值，判定所述人脸图像未正确佩戴口罩，从而实现判断所述人脸图像是否正确佩戴口罩，得到所述人脸图像对应的口罩检测结果，由于WASM是基于堆栈虚拟机的二进制指令格式，可以使客户端能够在Web上部署，可以基于预设WASM虚拟机将所述口罩检测结果通过预设浏览器进行显示，实现对人脸图像的口罩监测，相比传统技术中大多数设备的功能主要在进行检测人类戴口罩或其他障碍物时，是通过进行人脸识别的方式进行检测，很少有直接检测是否带口罩，本申请实施例通过直接进行检测人脸图像是否正确佩戴口罩，提高了对口罩监测的准确性和效率。
本申请实施例，通过基于本地预设智能终端，获取本地预设采集设备采集的人脸图像，并将所述人脸图像发送至本地预设边缘节点进行边缘计算，得到所述人脸图像对应的边缘计算结果，从而不使用服务架构，即可实现基于无服务边缘计算，获取所述边缘计算结果，再根据所述边缘计算结果，得到所述人脸图像对应的口罩检测结果，并将所述口罩检测结果通过预设浏览器进行显示，例如可以进一步基于预设WASM虚拟机将所述口罩检测结果通过预设浏览器进行显示，以实现对人脸图像佩戴口罩的监测，从而进行口罩监测时，用户只需访问网页，并启用相机权限即可触发软件或者在商场机场使用设备来检测即可，能够实现口罩监测无需服务器、无需安装、方便部署，可以直接在各类智能终端的浏览器中进行口罩监测结果的显示，不需要再安装特定软件，提高了进行口罩监测的便利性，提高了对人脸图像进行口罩监测的效率。目前的商用口罩检测系统通常需要与特定的软件或硬件捆绑在一起才可使用，例如Amazon Recognition需要安装特定软件，同时需要跨平台操作，而Mask detector需要安装特定的硬件，给操作者造成了使用的不便。
在一实施例中，所述获取所述边缘计算结果，根据所述边缘计算结果，得到所述人脸图像对应的口罩检测结果，并将所述口罩检测结果进行显示的步骤之后，还包括：获取所述口罩监测结果及所述口罩监测结果对应的预设目标数据，并将所述口罩监测结果及所述预设目标数据上传至预设服务器。
具体地，基于无服务边缘计算的口罩监测，不需要将进行口罩监测的全部数据上传至预设服务器，只需要传输口罩监测结果及所述口罩监测结果对应的预设目标数据等有价值的数据至预设服务器，以使预设服务器将该有价值的数据进行保存留底即可，例如口罩监测结果对应的口罩监测时间、口罩监测地点、口罩监测对象及是否正确佩戴口罩，尤其是在重要应用场景中，例如乘坐公共交通工具时，当用户未正确佩戴口罩而对用户产生不利后果，导致容易产生纠纷而需要保底留存证据时，将用户未正确佩戴口罩的监测结果对应的口罩监测时间、口罩监测地点、口罩监测对象上传至预设服务器进行存储，以留存证据，通过将有价值的部分数据上传至服务器，可以极大地减轻网络带宽的压力，且减少了对计算与存储资源的需求。
进一步地，请参阅图6，图6为本申请实施例提供的基于无服务边缘计算的佩戴口罩监测方法的第四个子流程示意图。如图6所示，在该实施例中，所述获取所述口罩监测结果及所述口罩监测结果对应的预设目标数据，将所述口罩监测结果及所述预设目标数据上传至预设服务器的步骤包括：S61、判断所述口罩监测结果是否为未正确佩戴口罩；S62、若所述口 罩监测结果为未正确佩戴口罩，将未正确佩戴口罩的口罩监测结果及对应的预设目标数据上传至预设服务器；S63、若所述口罩监测结果为正确佩戴口罩，不将未正确佩戴口罩的口罩监测结果及对应的预设目标数据上传至预设服务器。
具体地，由于若监测到用户未正确佩戴口罩，更可能对用户产生不利影响而容易产生纠纷，例如，在乘坐公共交通工具时，若用户未正确佩戴口罩，可能会对用户采取拒载的措施，再比如，在医院等入口处，若监测到用户未正确佩戴口罩，可能会对用户采取拒绝通行的措施，因此，若监测到用户对应的人脸图像未正确佩戴口罩，尤其需要将用户未正确佩戴口罩的监测结果进行保存留证，以避免后续产生纠纷时取证。因此，将所述人脸图像发送至本地预设边缘设备，以使所述预设边缘设备将所述人脸图像进行计算，得到对应的边缘计算结果，并将所述边缘计算结果进行返回。得到所述人脸图像对应的口罩检测结果后，可以进一步判断所述口罩监测结果是否为未正确佩戴口罩，若所述口罩监测结果为未正确佩戴口罩，将未正确佩戴口罩的口罩监测结果及对应的预设目标数据上传至预设服务器，若所述口罩监测结果为未正确佩戴口罩，将未正确佩戴口罩的口罩监测结果及对应的预设目标数据上传至预设服务器，从而进一步筛选出有必要进行上传存储至服务器的数据进行上传保存，能够进一步减少上传至服务器的数据，可进一步减轻网络带宽的压力，且减少了对存储资源的需求。并且这些数据仅仅用于一次检测，因此没有存储的必要性。
在一实施例中，所述若所述口罩监测结果为未正确佩戴口罩，将未正确佩戴口罩的口罩监测结果及对应的预设目标数据上传至预设服务器的步骤之后，还包括：若所述口罩监测结果为未正确佩戴口罩，对所述人脸图像发出告警。
具体地，根据是否正确佩戴口罩对应的判断规则，判断所述人脸图像对应的人脸是否正确佩戴口罩，例如，基于YOLO深度学习模型进行佩戴口罩监测，可以将人脸图像对应的口罩监测结果通过浏览器进行显示，所述人脸图像对应的口罩监测的输出结果包括口罩识别的精准度、YOLO深度学习模型的Bounding box以及检测的结果，如果人脸未正确佩戴口罩，对所述人脸图像发出告警，可以显示为无口罩并出现橙色框或者红色框等警示色，则表示佩戴者未正确佩戴口罩，在通行入口应用场景下，例如在医院入口或者办公楼入口处，未正确佩戴口罩的人可以采取不放行通过的措施，如果人脸正确佩戴口罩，不对所述人脸图像发出告警，显示为有口罩并可以出现蓝色框，进一步地，在通行入口应用场景下，正确佩戴口罩的人可以采取放行通过的措施，以实现对通行的自动控制，提高通行控制效率。
需要说明的是，上述各个实施例所述的基于无服务边缘计算的佩戴口罩监测方法，可以根据需要将不同实施例中包含的技术特征重新进行组合，以获取组合后的实施方案，但都在本申请要求的保护范围之内。
请参阅图7，图7为本申请实施例提供的基于无服务边缘计算的佩戴口罩监测装置的一个示意性框图。对应于上述所述基于无服务边缘计算的佩戴口罩监测方法，本申请实施例还提供一种基于无服务边缘计算的佩戴口罩监测装置。如图7所示，该基于无服务边缘计算的佩戴口罩监测装置包括用于执行上述所述基于无服务边缘计算的佩戴口罩监测方法的单元，该基于无服务边缘计算的佩戴口罩监测装置可以被配置于本地预设智能终端中。具体地，请参阅图7，该基于无服务边缘计算的佩戴口罩监测装置70包括第一获取单元71、计算单元72及显示单元73。其中，第一获取单元71，用于获取本地预设采集设备采集的人脸图像；计算单元72，用于将所述人脸图像发送至本地预设边缘节点进行边缘计算，得到所述人脸图像对应的边缘计算结果，其中，所述本地预设边缘节点为部署了预设高性能神经网络推理计算框架NCNN的本地智能终端；显示单元73，用于获取所述边缘计算结果，根据所述边缘计算结果，得到所述人脸图像对应的口罩检测结果，并将所述口罩检测结果进行显示。
在一实施例中，所述计算单元72包括：第一发送子单元，用于获取与所述本地预设智能终端基于物联网处于连接状态的本地预设其它终端对应的预设终端标识，根据所述预设终端标识，将所述人脸图像发送至所述预设终端标识对应的本地预设其它终端，以使所述本地预设其它终端对所述人脸图像进行边缘计算，得到所述人脸图像对应的边缘计算结果，并将所 述边缘计算结果进行返回；接收子单元，用于接收所述本地预设其它终端返回的所述边缘计算结果，得到所述人脸图像对应的边缘计算结果。
在一实施例中，所述第一发送子单元包括：第一获取子单元，用于获取与所述本地预设智能终端基于物联网处于连接状态的多个所述本地预设其它终端对应的预设终端标识；确定子单元，用于根据预设时间周期，确定所述预设时间周期内未被发送人脸图像的预设终端标识为目标终端标识；第二发送子单元，用于将所述人脸图像发送至所述目标终端标识对应的本地预设其它终端。
在一实施例中，所述显示单元73包括：第二获取子单元，用于获取所述边缘计算结果包含的口罩识别准确率；第一判断子单元，用于判断所述口罩识别准确率是否大于或者等于预设准确率阈值；第一判定子单元，用于若所述口罩识别准确率大于或者等于预设准确率阈值，判定所述人脸图像正确佩戴口罩；第二判定子单元，用于若所述口罩识别准确率小于预设准确率阈值，判定所述人脸图像未正确佩戴口罩；显示子单元，用于将所述人脸图像正确佩戴口罩或者所述人脸图像未正确佩戴口罩的检测结果通过预设浏览器进行显示。
在一实施例中，所述基于无服务边缘计算的佩戴口罩监测装置70还包括：上传单元，用于获取所述口罩监测结果及所述口罩监测结果对应的预设目标数据，并将所述口罩监测结果及所述预设目标数据上传至预设服务器。
在一实施例中，所述上传单元包括：第二判断子单元，用于判断所述口罩监测结果是否为未正确佩戴口罩；上传子单元，用于若所述口罩监测结果为未正确佩戴口罩，将未正确佩戴口罩的口罩监测结果及对应的预设目标数据上传至预设服务器。
在一实施例中，所述上传单元还包括：告警子单元，用于若所述口罩监测结果为未正确佩戴口罩，对所述人脸图像发出告警。
需要说明的是，所属领域的技术人员可以清楚地了解到，上述基于无服务边缘计算的佩戴口罩监测装置和各单元的具体实现过程，可以参考前述方法实施例中的相应描述，为了描述的方便和简洁，在此不再赘述。
同时，上述基于无服务边缘计算的佩戴口罩监测装置中各个单元的划分和连接方式仅用于举例说明，在其他实施例中，可将基于无服务边缘计算的佩戴口罩监测装置按照需要划分为不同的单元，也可将基于无服务边缘计算的佩戴口罩监测装置中各单元采取不同的连接顺序和方式，以完成上述基于无服务边缘计算的佩戴口罩监测装置的全部或部分功能。
上述基于无服务边缘计算的佩戴口罩监测装置可以实现为一种计算机程序的形式，该计算机程序可以在如图8所示的计算机设备上运行。
请参阅图8，图8是本申请实施例提供的一种计算机设备的示意性框图。该计算机设备500可以是笔记本电脑、平板电脑或者其它智能终端等计算机设备，也可以是其他设备中的组件或者部件。
参阅图8，该计算机设备500包括通过系统总线501连接的处理器502、存储器和网络接口505，其中，存储器可以包括非易失性存储介质503和内存储器504，所述存储器也可以为易失性存储介质。
该非易失性存储介质503可存储操作系统5031和计算机程序5032。该计算机程序5032被执行时，可使得处理器502执行一种上述基于无服务边缘计算的佩戴口罩监测方法。
该处理器502用于提供计算和控制能力，以支撑整个计算机设备500的运行。
该内存储器504为非易失性存储介质503中的计算机程序5032的运行提供环境，该计算机程序5032被处理器502执行时，可使得处理器502执行一种上述基于无服务边缘计算的佩戴口罩监测方法。
该网络接口505用于与其它设备进行网络通信。本领域技术人员可以理解，图8中示出的结构，仅仅是与本申请方案相关的部分结构的框图，并不构成对本申请方案所应用于其上的计算机设备500的限定，具体的计算机设备500可以包括比图中所示更多或更少的部件，或者组合某些部件，或者具有不同的部件布置。例如，在一些实施例中，计算机设备可以仅 包括存储器及处理器，在这样的实施例中，存储器及处理器的结构及功能与图8所示实施例一致，在此不再赘述。
其中，所述处理器502用于运行存储在存储器中的计算机程序5032，以实现本申请实施例所描述的基于无服务边缘计算的佩戴口罩监测方法。
应当理解，在本申请实施例中，处理器502可以是中央处理单元(Central Processing Unit，CPU)，该处理器502还可以是其他通用处理器、数字信号处理器(Digital Signal Processor，DSP)、专用集成电路(Application Specific Integrated Circuit，ASIC)、现成可编程门阵列(Field-Programmable Gate Array，FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件等。其中，通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。
本领域普通技术人员可以理解的是实现上述实施例的方法中的全部或部分流程，是可以通过计算机程序来完成，该计算机程序可存储于一计算机可读存储介质。该计算机程序被该计算机系统中的至少一个处理器执行，以实现上述方法的实施例的流程步骤。
因此，本申请还提供一种计算机可读存储介质。该计算机可读存储介质可以为非易失性的计算机可读存储介质，也可以为易失性的计算机可读存储介质，该计算机可读存储介质存储有计算机程序，该计算机程序被处理器执行时使处理器执行以上各实施例中所描述的所述基于无服务边缘计算的佩戴口罩监测方法的步骤。
所述存储介质为实体的、非瞬时性的存储介质，例如可以是U盘、移动硬盘、只读存储器(Read-Only Memory，ROM)、磁碟或者光盘等各种可以存储计算机程序的实体存储介质。
本领域普通技术人员可以意识到，结合本文中所公开的实施例描述的各示例的单元及算法步骤，能够以电子硬件、计算机软件或者二者的结合来实现，为了清楚地说明硬件和软件的可互换性，在上述说明中已经按照功能一般性地描述了各示例的组成及步骤。这些功能究竟以硬件还是软件方式来执行，取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能，但是这种实现不应认为超出本申请的范围。
以上所述，仅为本申请的具体实施方式，但本申请明的保护范围并不局限于此，任何熟悉本技术领域的技术人员在本申请揭露的技术范围内，可轻易想到各种等效的修改或替换，这些修改或替换都应涵盖在本申请的保护范围之内。因此，本申请的保护范围应以权利要求的保护范围为准。

Claims (20)

  1. 一种基于无服务边缘计算的佩戴口罩监测方法，应用于本地预设智能终端，其中，所述方法包括：
    获取本地预设采集设备采集的人脸图像；
    将所述人脸图像发送至本地预设边缘节点进行边缘计算，得到所述人脸图像对应的边缘计算结果，其中，所述本地预设边缘节点为部署了预设高性能神经网络推理计算框架NCNN的本地智能终端；
    获取所述边缘计算结果，根据所述边缘计算结果，得到所述人脸图像对应的口罩检测结果，并将所述口罩检测结果进行显示。
  2. 根据权利要求1所述基于无服务边缘计算的佩戴口罩监测方法，其中，所述将所述人脸图像发送至本地预设边缘节点进行边缘计算，得到所述人脸图像对应的边缘计算结果的步骤包括：
    获取与所述本地预设智能终端基于物联网处于连接状态的本地预设其它终端对应的预设终端标识，根据所述预设终端标识，将所述人脸图像发送至所述预设终端标识对应的本地预设其它终端，以使所述本地预设其它终端对所述人脸图像进行边缘计算，得到所述人脸图像对应的边缘计算结果，并将所述边缘计算结果进行返回；
    接收所述本地预设其它终端返回的所述边缘计算结果，得到所述人脸图像对应的边缘计算结果。
  3. 根据权利要求2所述基于无服务边缘计算的佩戴口罩监测方法，其中，所述获取与所述本地预设智能终端基于物联网处于连接状态的本地预设其它终端对应的预设终端标识，根据所述预设终端标识，将所述人脸图像发送至所述预设终端标识对应的本地预设其它终端的步骤包括：
    获取与所述本地预设智能终端基于物联网处于连接状态的多个所述本地预设其它终端对应的预设终端标识；
    根据预设时间周期，确定所述预设时间周期内未被发送人脸图像的预设终端标识为目标终端标识；
    将所述人脸图像发送至所述目标终端标识对应的本地预设其它终端。
  4. 根据权利要求1所述基于无服务边缘计算的佩戴口罩监测方法，其中，所述根据所述边缘计算结果，得到所述人脸图像对应的口罩检测结果，并将所述口罩检测结果进行显示的步骤包括：
    获取所述边缘计算结果包含的口罩识别准确率；
    判断所述口罩识别准确率是否大于或者等于预设准确率阈值；
    若所述口罩识别准确率大于或者等于预设准确率阈值，判定所述人脸图像正确佩戴口罩；
    若所述口罩识别准确率小于预设准确率阈值，判定所述人脸图像未正确佩戴口罩；
    将所述人脸图像正确佩戴口罩或者所述人脸图像未正确佩戴口罩的检测结果通过预设浏览器进行显示。
  5. 根据权利要求1所述基于无服务边缘计算的佩戴口罩监测方法，其中，所述获取所述边缘计算结果，根据所述边缘计算结果，得到所述人脸图像对应的口罩检测结果，并将所述口罩检测结果进行显示的步骤之后，还包括：
    获取所述口罩监测结果及所述口罩监测结果对应的预设目标数据，并将所述口罩监测结果及所述预设目标数据上传至预设服务器。
  6. 根据权利要求5所述基于无服务边缘计算的佩戴口罩监测方法，其中，所述获取所述口罩监测结果及所述口罩监测结果对应的预设目标数据，将所述口罩监测结果及所述预设目标数据上传至预设服务器的步骤包括：
    判断所述口罩监测结果是否为未正确佩戴口罩；
    若所述口罩监测结果为未正确佩戴口罩，将未正确佩戴口罩的口罩监测结果及对应的预 设目标数据上传至预设服务器。
  7. 根据权利要求6所述基于无服务边缘计算的佩戴口罩监测方法，其中，所述若所述口罩监测结果为未正确佩戴口罩，将未正确佩戴口罩的口罩监测结果及对应的预设目标数据上传至预设服务器的步骤之后，还包括：
    若所述口罩监测结果为未正确佩戴口罩，对所述人脸图像发出告警。
  8. 一种基于无服务边缘计算的佩戴口罩监测装置，应用于本地预设智能终端，其中，所述装置包括：
    第一获取单元，用于获取本地预设采集设备采集的人脸图像；
    计算单元，用于将所述人脸图像发送至本地预设边缘节点进行边缘计算，得到所述人脸图像对应的边缘计算结果，其中，所述本地预设边缘节点为部署了预设高性能神经网络推理计算框架NCNN的本地智能终端；
    显示单元，用于获取所述边缘计算结果，根据所述边缘计算结果，得到所述人脸图像对应的口罩检测结果，并将所述口罩检测结果进行显示。
  9. 一种计算机设备，所述计算机设备包括存储器以及与所述存储器相连的处理器；所述存储器用于存储计算机程序；所述处理器用于运行所述计算机程序，以执行如下步骤：
    获取本地预设采集设备采集的人脸图像；
    将所述人脸图像发送至本地预设边缘节点进行边缘计算，得到所述人脸图像对应的边缘计算结果，其中，所述本地预设边缘节点为部署了预设高性能神经网络推理计算框架NCNN的本地智能终端；
    获取所述边缘计算结果，根据所述边缘计算结果，得到所述人脸图像对应的口罩检测结果，并将所述口罩检测结果进行显示。
  10. 根据权利要求9所述计算机设备，其中，所述将所述人脸图像发送至本地预设边缘节点进行边缘计算，得到所述人脸图像对应的边缘计算结果的步骤包括：
    获取与所述本地预设智能终端基于物联网处于连接状态的本地预设其它终端对应的预设终端标识，根据所述预设终端标识，将所述人脸图像发送至所述预设终端标识对应的本地预设其它终端，以使所述本地预设其它终端对所述人脸图像进行边缘计算，得到所述人脸图像对应的边缘计算结果，并将所述边缘计算结果进行返回；
    接收所述本地预设其它终端返回的所述边缘计算结果，得到所述人脸图像对应的边缘计算结果。
  11. 根据权利要求10所述计算机设备，其中，所述获取与所述本地预设智能终端基于物联网处于连接状态的本地预设其它终端对应的预设终端标识，根据所述预设终端标识，将所述人脸图像发送至所述预设终端标识对应的本地预设其它终端的步骤包括：
    获取与所述本地预设智能终端基于物联网处于连接状态的多个所述本地预设其它终端对应的预设终端标识；
    根据预设时间周期，确定所述预设时间周期内未被发送人脸图像的预设终端标识为目标终端标识；
    将所述人脸图像发送至所述目标终端标识对应的本地预设其它终端。
  12. 根据权利要求9所述计算机设备，其中，所述根据所述边缘计算结果，得到所述人脸图像对应的口罩检测结果，并将所述口罩检测结果进行显示的步骤包括：
    获取所述边缘计算结果包含的口罩识别准确率；
    判断所述口罩识别准确率是否大于或者等于预设准确率阈值；
    若所述口罩识别准确率大于或者等于预设准确率阈值，判定所述人脸图像正确佩戴口罩；
    若所述口罩识别准确率小于预设准确率阈值，判定所述人脸图像未正确佩戴口罩；
    将所述人脸图像正确佩戴口罩或者所述人脸图像未正确佩戴口罩的检测结果通过预设浏览器进行显示。
  13. 根据权利要求9所述计算机设备，其中，所述获取所述边缘计算结果，根据所述边 缘计算结果，得到所述人脸图像对应的口罩检测结果，并将所述口罩检测结果进行显示的步骤之后，还包括：
    获取所述口罩监测结果及所述口罩监测结果对应的预设目标数据，并将所述口罩监测结果及所述预设目标数据上传至预设服务器。
  14. 根据权利要求13所述计算机设备，其中，所述获取所述口罩监测结果及所述口罩监测结果对应的预设目标数据，将所述口罩监测结果及所述预设目标数据上传至预设服务器的步骤包括：
    判断所述口罩监测结果是否为未正确佩戴口罩；
    若所述口罩监测结果为未正确佩戴口罩，将未正确佩戴口罩的口罩监测结果及对应的预设目标数据上传至预设服务器。
  15. 根据权利要求14所述计算机设备，其中，所述若所述口罩监测结果为未正确佩戴口罩，将未正确佩戴口罩的口罩监测结果及对应的预设目标数据上传至预设服务器的步骤之后，还包括：
    若所述口罩监测结果为未正确佩戴口罩，对所述人脸图像发出告警。
  16. 一种计算机可读存储介质，所述存储介质存储有计算机程序，所述计算机程序被处理器执行时可实现如下步骤：
    获取本地预设采集设备采集的人脸图像；
    将所述人脸图像发送至本地预设边缘节点进行边缘计算，得到所述人脸图像对应的边缘计算结果，其中，所述本地预设边缘节点为部署了预设高性能神经网络推理计算框架NCNN的本地智能终端；
    获取所述边缘计算结果，根据所述边缘计算结果，得到所述人脸图像对应的口罩检测结果，并将所述口罩检测结果进行显示。
  17. 根据权利要求16所述计算机可读存储介质，其中，所述将所述人脸图像发送至本地预设边缘节点进行边缘计算，得到所述人脸图像对应的边缘计算结果的步骤包括：
    获取与所述本地预设智能终端基于物联网处于连接状态的本地预设其它终端对应的预设终端标识，根据所述预设终端标识，将所述人脸图像发送至所述预设终端标识对应的本地预设其它终端，以使所述本地预设其它终端对所述人脸图像进行边缘计算，得到所述人脸图像对应的边缘计算结果，并将所述边缘计算结果进行返回；
    接收所述本地预设其它终端返回的所述边缘计算结果，得到所述人脸图像对应的边缘计算结果。
  18. 根据权利要求17所述计算机可读存储介质，其中，所述获取与所述本地预设智能终端基于物联网处于连接状态的本地预设其它终端对应的预设终端标识，根据所述预设终端标识，将所述人脸图像发送至所述预设终端标识对应的本地预设其它终端的步骤包括：
    获取与所述本地预设智能终端基于物联网处于连接状态的多个所述本地预设其它终端对应的预设终端标识；
    根据预设时间周期，确定所述预设时间周期内未被发送人脸图像的预设终端标识为目标终端标识；
    将所述人脸图像发送至所述目标终端标识对应的本地预设其它终端。
  19. 根据权利要求16所述计算机可读存储介质，其中，所述根据所述边缘计算结果，得到所述人脸图像对应的口罩检测结果，并将所述口罩检测结果进行显示的步骤包括：
    获取所述边缘计算结果包含的口罩识别准确率；
    判断所述口罩识别准确率是否大于或者等于预设准确率阈值；
    若所述口罩识别准确率大于或者等于预设准确率阈值，判定所述人脸图像正确佩戴口罩；
    若所述口罩识别准确率小于预设准确率阈值，判定所述人脸图像未正确佩戴口罩；
    将所述人脸图像正确佩戴口罩或者所述人脸图像未正确佩戴口罩的检测结果通过预设浏览器进行显示。
  20. 根据权利要求16所述计算机可读存储介质，其中，所述获取所述边缘计算结果，根据所述边缘计算结果，得到所述人脸图像对应的口罩检测结果，并将所述口罩检测结果进行显示的步骤之后，还包括：
    获取所述口罩监测结果及所述口罩监测结果对应的预设目标数据，并将所述口罩监测结果及所述预设目标数据上传至预设服务器。
PCT/CN2021/097214 2021-04-30 2021-05-31 基于无服务边缘计算的佩戴口罩监测方法、装置及设备 WO2022227209A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202110481307.8A CN113221703A (zh) 2021-04-30 2021-04-30 基于无服务边缘计算的佩戴口罩监测方法、装置及设备
CN202110481307.8 2021-04-30

Publications (1)

Publication Number Publication Date
WO2022227209A1 true WO2022227209A1 (zh) 2022-11-03

Family

ID=77090493

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/097214 WO2022227209A1 (zh) 2021-04-30 2021-05-31 基于无服务边缘计算的佩戴口罩监测方法、装置及设备

Country Status (2)

Country Link
CN (1) CN113221703A (zh)
WO (1) WO2022227209A1 (zh)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114821388A (zh) * 2022-03-10 2022-07-29 慧之安信息技术股份有限公司 基于边缘计算的室内口罩检测方法和装置
CN115884006B (zh) * 2023-02-23 2023-06-09 启实(烟台)数据技术有限公司 基于AIoT的校园安全防控系统和方法

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109460729A (zh) * 2018-11-01 2019-03-12 浩云科技股份有限公司 一种嵌入式多人脸检测方法及系统
US20200026963A1 (en) * 2018-07-19 2020-01-23 International Business Machines Corporation Reducing computational costs of deep reinforcement learning by gated convolutional neural network
CN111241975A (zh) * 2020-01-07 2020-06-05 华南理工大学 一种基于移动端边缘计算的人脸识别检测方法及系统
CN111428559A (zh) * 2020-02-19 2020-07-17 北京三快在线科技有限公司 一种口罩佩戴情况检测方法、装置、电子设备及存储介质
CN111985621A (zh) * 2020-08-24 2020-11-24 西安建筑科技大学 口罩佩戴实时检测的神经网络模型的搭建方法及实施系统

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111539338A (zh) * 2020-04-26 2020-08-14 深圳前海微众银行股份有限公司 行人口罩佩戴管控方法、装置、设备及计算机存储介质
CN112085010B (zh) * 2020-10-28 2022-07-12 成都信息工程大学 一种基于图像识别的口罩检测和部署系统及方法
CN112434578B (zh) * 2020-11-13 2023-07-25 浙江大华技术股份有限公司 口罩佩戴规范性检测方法、装置、计算机设备和存储介质

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20200026963A1 (en) * 2018-07-19 2020-01-23 International Business Machines Corporation Reducing computational costs of deep reinforcement learning by gated convolutional neural network
CN109460729A (zh) * 2018-11-01 2019-03-12 浩云科技股份有限公司 一种嵌入式多人脸检测方法及系统
CN111241975A (zh) * 2020-01-07 2020-06-05 华南理工大学 一种基于移动端边缘计算的人脸识别检测方法及系统
CN111428559A (zh) * 2020-02-19 2020-07-17 北京三快在线科技有限公司 一种口罩佩戴情况检测方法、装置、电子设备及存储介质
CN111985621A (zh) * 2020-08-24 2020-11-24 西安建筑科技大学 口罩佩戴实时检测的神经网络模型的搭建方法及实施系统

Also Published As

Publication number Publication date
CN113221703A (zh) 2021-08-06

Similar Documents

Publication Publication Date Title
US11483370B2 (en) Preprocessing sensor data for machine learning
WO2020185233A1 (en) Responding to machine learning requests from multiple clients
CN111556294B (zh) 安全监测方法、装置、服务器、终端及可读存储介质
WO2022227209A1 (zh) 基于无服务边缘计算的佩戴口罩监测方法、装置及设备
CN109583278B (zh) 人脸识别报警的方法、装置、系统及计算机设备
WO2021180004A1 (zh) 视频分析方法、视频分析的管理方法及相关设备
JP6764214B1 (ja) 混雑情報報知システム
US11240136B2 (en) Determining attributes using captured network probe data in a wireless communications system
US10999345B2 (en) Real-time video delivery for connected home applications
EP3844667A1 (en) Methods and apparatus for reducing false positives in facial recognition
JP2022515908A (ja) 歩行者行為の検出方法及び装置、電子機器並びに記憶媒体
US11914774B1 (en) Dynamic interface flow based on device location
US20180165500A1 (en) Detection of Traffic Light Signal Changes
US9576467B2 (en) Emergency detection device, emergency detection system, recording medium, and method therefor
JP6875326B2 (ja) 画像解析に基づくワークフローを実施するサービス
US20230205817A1 (en) Systems and methods for identifying events within video content using intelligent search query
CN108664847B (zh) 一种对象识别方法、设备和系统
JP2020154620A (ja) 地域監視システム、分析装置、地域監視方法、及びプログラム
CN110401603B (zh) 用于处理信息的方法和装置
US20210377580A1 (en) Live or local environmental awareness
JP7198052B2 (ja) 画像監視システム
CN112866332A (zh) 基于云边融合实现突发事件识别预警的系统、方法、装置、处理器及其存储介质
US11983909B2 (en) Responding to machine learning requests from multiple clients
CN111813983A (zh) 一种异常体温预警方法、装置及系统
CN110796826A (zh) 一种用于识别烟雾火焰的报警方法及系统

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21938679

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21938679

Country of ref document: EP

Kind code of ref document: A1