A detection and identification method, edge device and cloud device

Abstract

An embodiment of the present invention provides a detection and identification method, an edge device, and a cloud device, the method comprising: the edge device obtains first video stream data within a preset area; the edge device performs a first feature extraction on the first video stream data to obtain image data of a first object; and the edge device sends the image data of the first object to the cloud device. An edge device is set at a location close to the data source to obtain the first video stream data within the preset area, and performs a first feature extraction on the first video stream data to obtain image data of the first object. The image data required for identifying the first object is transmitted to the cloud device through data screening by the edge device, and the first object is then identified based on the cloud device. In this way, by adopting a distributed computing architecture, when the amount of data accessed is large, the capacity and computing power of the existing cloud server can still meet the needs of detecting and identifying the first object.

Description

Technical Field

[0001] The present invention relates to the field of Internet of Things technology, and in particular to a detection and identification method, an edge device, and a cloud device. Background Art

[0002] As security needs grow, the workload of manual on-site inspections has increased, and traditional manual inspections are no longer sufficient. Currently, multiple cameras are typically installed indoors and outdoors in a park or factory to collect video data within the surveillance area. This data is then transmitted to the cloud for analysis and computation to identify the identities of the objects captured by the cameras.

[0003] However, with the access of tens of millions of IoT data, the capacity and computing power of existing cloud servers are unable to meet the needs of detecting and identifying large amounts of video data. Summary of the Invention

[0004] Embodiments of the present invention provide a detection and identification method, an edge device, and a cloud device to solve the problem in the prior art that the capacity and computing power of the cloud server are difficult to detect and identify large amounts of video data.

[0005] In a first aspect, an embodiment of the present invention provides a detection and identification method, the method comprising:

[0006] The edge device obtains first video stream data in a preset area;

[0007] The edge device performs a first feature extraction on the first video stream data to obtain image data of the first object;

[0008] The edge device sends the image data of the first object to the cloud device.

[0009] Optionally, after obtaining the image data of the first object and before the edge device sends the image data of the first object to the cloud device, the method further includes:

[0010] The edge device performs second feature extraction based on the image data of the first object to obtain facial feature data of the first object;

[0011] The edge device sends the image data of the first object to the cloud device, including:

[0012] The edge device sends the facial feature data of the first object to the cloud device.

[0013] Optionally, the edge device performs second feature extraction based on the image data of the first object to obtain facial feature data of the first object, including:

[0014] performing illumination compensation on the image data of the first object to obtain compensated image data;

[0015] Perform a second feature extraction on the compensated image data to obtain facial feature data of the first object.

[0016] Optionally, performing a second feature extraction on the compensated image data to obtain facial feature data of the first object includes:

[0017] performing a second feature extraction on the compensated image data to obtain data of a plurality of facial feature points of the first object;

[0018] Feature fusion is performed on data of multiple facial feature points of the first object to obtain facial feature data of the first object.

[0019] Optionally, the edge device performs second feature extraction based on the image data of the first object to obtain facial feature data of the first object, including:

[0020] In a case where the edge device needs to perform a second feature extraction on the image data of the first object, the edge device detects whether a parameter for the second feature extraction is updated;

[0021] When the parameters used for the second feature extraction are updated, the second feature extraction is performed on the image data of the first object using the updated parameters.

[0022] In a second aspect, an embodiment of the present invention provides another detection and identification method, the method comprising:

[0023] The cloud device receives image data of the first object from the edge device;

[0024] The cloud device performs identity recognition on the first object based on the image data of the first object.

[0025] Optionally, the image data of the first object is facial feature data of the first object.

[0026] In a third aspect, an embodiment of the present invention provides an edge device, the edge device comprising:

[0027] An acquisition module, configured to acquire first video stream data within a preset area;

[0028] A first feature extraction module, configured to extract a first feature from the first video stream data to obtain image data of a first object;

[0029] A sending module is used to send the image data of the first object to a cloud device.

[0030] Optionally, the edge device further includes:

[0031] a second feature extraction module, configured to perform second feature extraction based on the image data of the first object to obtain facial feature data of the first object;

[0032] The sending module includes:

[0033] The sending submodule is used to send the facial feature data of the first object to the cloud device.

[0034] Optionally, the second feature extraction module includes:

[0035] a compensation submodule, configured to perform illumination compensation on the image data of the first object to obtain compensated image data;

[0036] The feature extraction submodule is used to perform a second feature extraction on the compensated image data to obtain facial feature data of the first object.

[0037] Optionally, the feature extraction submodule includes:

[0038] a feature extraction unit, configured to perform a second feature extraction on the compensated image data to obtain data of a plurality of facial feature points of the first object;

[0039] The feature fusion unit is used to perform feature fusion on the data of multiple facial feature points of the first object to obtain facial feature data of the first object.

[0040] Optionally, the second feature extraction module includes:

[0041] a detection submodule, configured to, when the edge device needs to perform second feature extraction on the image data of the first object, detect whether parameters used for the second feature extraction are updated;

[0042] An updating submodule is configured to, when the parameters used for extracting the second feature are updated, perform second feature extraction on the image data of the first object using the updated parameters.

[0043] In a fourth aspect, an embodiment of the present invention provides a cloud device, the cloud device comprising:

[0044] a receiving module, configured to receive image data of a first object from an edge device;

[0045] The recognition module is configured to perform identity recognition on the first object based on the image data of the first object.

[0046] Optionally, the image data of the first object is facial feature data of the first object.

[0047] In an embodiment of the present invention, an edge device is positioned near a data source to obtain first video stream data within a preset area, and first feature extraction is performed on the first video stream data to obtain image data of a first object. The image data required for identifying the first object is then transmitted to a cloud device through data screening by the edge device, and the first object is then identified based on the cloud device. In this way, by adopting a distributed computing architecture, even when a large amount of data is being accessed, the capacity and computing power of the existing cloud server can still meet the needs of detecting and identifying the first object, thereby improving the computing speed of the device and extending its service life. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments or descriptions of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0049] Figure 1 This is a flow chart of a detection and identification method provided by an embodiment of the present invention;

[0050] Figure 2 This is a schematic diagram of the system architecture of a detection and identification method provided by an embodiment of the present invention;

[0051] Figure 3 1 is a flow chart of another detection and identification method provided by an embodiment of the present invention;

[0052] Figure 4 is a schematic structural diagram of an edge device provided by an embodiment of the present invention;

[0053] Figure 5 It is a structural diagram of a cloud device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0054] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0055] The terms "first," "second," and the like in the specification and claims of the present invention are used to distinguish similar objects, and are not used to describe a specific order or precedence. It should be understood that the structures used in this manner are interchangeable where appropriate, so that the embodiments of the present invention can be implemented in an order other than that illustrated or described herein, and that the objects distinguished by "first," "second," and the like are generally of the same type, and do not limit the number of objects. For example, the first object can be one or more. In addition, the term "and / or" in the specification and claims indicates at least one of the connected objects, and the character " / " generally indicates that the objects connected are in an "or" relationship.

[0056] See also Figure 1 , Figure 1 FIG. 1 is a flow chart of a detection and identification method provided by an embodiment of the present invention. Figure 1 As shown, an embodiment of the present invention provides a detection and identification method, comprising the following steps:

[0057] Step 101: The edge device obtains first video stream data in a preset area;

[0058] In one example, multiple cameras can be set up in a preset area to perform 24 / 7 uninterrupted detection of objects within the preset area. The edge device can be an edge gateway device deployed near the camera end. The edge gateway device (with a built-in artificial intelligence (AI) module and network module) can obtain surveillance video data from multiple cameras deployed in the preset area, thereby obtaining the first video stream data within the preset area to detect the first object.

[0059] In another example, multiple smart cameras with built-in AI and network modules can be deployed in a preset area to monitor objects within the preset area 24 / 7. The edge devices can be the smart cameras themselves. The multiple smart cameras deployed in the preset area acquire first video stream data within the preset area to detect a first object.

[0060] Preset areas can include both indoor and outdoor scenes. Indoor framing requirements include framing directly opposite the machine room door to capture the front view of the subject entering the room, and framing cabinets to capture the cabinet status. Outdoor framing requirements include framing machine room doors, cabinets, ladders, outdoor distribution boxes, and rooftop platforms to capture video stream data from outdoor areas.

[0061] Among them, cameras can include four product types: indoor directional fixed focus, indoor variable-direction fixed focus, indoor variable-direction zoom, and outdoor variable-direction zoom. By formulating technical requirements for cameras, they can be designed, installed and deployed according to on-site conditions in actual applications, and the number of cameras in a station can be refined as much as possible to meet the necessary equipment and facility framing, while taking into account the framing of other equipment and facilities.

[0062] Step 102: The edge device performs a first feature extraction on the first video stream data to obtain image data of a first object.

[0063] The edge device has a built-in AI module that can extract the first feature of the acquired first video stream data to obtain image data of the first object. Compared with video stream data, the image data of the first object has more efficient transmission and reduces interference from irrelevant features such as the environment when detecting and identifying the first object, thereby improving the accuracy of detection and identification of the first object. By filtering the necessary uploaded data from the first video stream data, network transmission traffic is reduced, which can alleviate the pressure on the platform's computing power.

[0064] Step 103: The edge device sends the image data of the first object to the cloud device.

[0065] Data collection is the foundation of edge computing. With the proliferation of cameras, other sensors, and IoT devices, the volume of primary video stream data will also increase. It's unrealistic to burden cloud devices with all the data transmission and computational burdens of front-end devices. Therefore, edge devices (such as edge gateways) are beginning to incorporate computational analysis capabilities. They can process data collected near sensors and other IoT devices, reducing the amount of data before sending it to the cloud. This reduces the burden on cloud devices.

[0066] In this embodiment, an edge device is positioned near a data source to acquire first video stream data within a preset area, perform first feature extraction on the first video stream data, and obtain image data of a first object. The edge device then filters the data and transmits the image data required for identifying the first object to a cloud device, which then identifies the first object. In this way, using a distributed computing architecture, even when accessing large amounts of data, the capacity and computing power of existing cloud servers can still meet the needs of detecting and identifying the first object, thereby increasing the computing speed of the device and extending its service life.

[0067] Optionally, after obtaining the image data of the first object and before the edge device sends the image data of the first object to the cloud device, the method further includes:

[0068] The edge device performs second feature extraction based on the image data of the first object to obtain facial feature data of the first object;

[0069] The edge device sends the image data of the first object to the cloud device, including:

[0070] The edge device sends the facial feature data of the first object to the cloud device.

[0071] In this embodiment, an edge device is set at a position close to the data source to obtain the first video stream data in a preset area. First, the edge device performs a first feature extraction on the first video stream data, and obtains the image data of the first object through screening, thereby reducing network transmission traffic; then, the edge device performs a second feature extraction based on the image data of the first object to obtain the facial feature data of the first object, further reducing the interference of irrelevant features and improving the accuracy of detection and identification of the first object.

[0072] The facial feature data of the first object may be data in a vector format, which has a more efficient transmission and detection and recognition efficiency than the image data of the first object.

[0073] Specifically, in the intelligent operation and maintenance application scenario, the present invention proposes a detection and identification method, which extracts the image data of the first object from the first video stream data through the edge device, and then performs a second feature extraction on the image data of the first object to obtain the facial feature data of the first object. In this way, the edge device obtains pictures according to the video stream, intercepts relevant pictures containing only the facial part of the first object, omits the pictures of the rest of the background part, and improves the accuracy of detection and identification of the first object. On the other hand, the edge device sends the facial feature data of the first object to the cloud device, which reduces the network transmission traffic. The cloud device identifies the first object based on the facial feature data of the first object, which effectively reduces the load pressure on the cloud server when performing identity detection and identification. Among them, the system architecture using the cloud-edge combination includes the edge terminal layer, the transmission layer and the cloud layer, such as Figure 2 shown.

[0074] At the edge layer, cameras are installed to detect people and faces inside the computer room and around the base station, and facial image data is collected in real time to achieve 24 / 7 uninterrupted monitoring. Edge devices (such as edge gateways) are also set up at the data source to obtain the first video stream data collected by cameras inside the computer room and around the base station.

[0075] At the transport layer, the fourth generation mobile communication technology / fifth generation mobile communication technology (the 4 thGeneration Mobile Communication Technology / 5 th The 4G / 5G (Generation Mobile Communication Technology) network is used as a data transmission link to realize the connection between the devices in the edge layer and the cloud layer, and the transmitted data includes the facial feature data of the first object collected by the edge.

[0076] At the cloud layer, remote centralized management of terminal devices such as cameras can be achieved, and centralized face recognition can be performed on the facial feature data of the first object to determine whether the first object is a staff member.

[0077] By filtering data at the edge, the required recognition data is transmitted to the cloud, and then the cloud is used to perform facial recognition, which reduces the computing pressure on the cloud, improves the system speed and extends the service life of the machine.

[0078] Optionally, the edge device performs second feature extraction based on the image data of the first object to obtain facial feature data of the first object, including:

[0079] performing illumination compensation on the image data of the first object to obtain compensated image data;

[0080] Perform a second feature extraction on the compensated image data to obtain facial feature data of the first object.

[0081] In this embodiment, an edge device is positioned near a data source to obtain first video stream data within a preset area. First, the edge device performs a first feature extraction on the first video stream data, filtering and obtaining image data of a first object, thereby reducing network transmission traffic. The edge device then performs illumination compensation based on the image data of the first object to obtain compensated image data, thereby preventing poor illumination conditions that could reduce recognition accuracy of the first object. The key algorithm can be a modified Resnet18 network, using a histogram equalization algorithm to perform illumination compensation on the image data of the first object.

[0082] Then, a second feature extraction is performed on the compensated image data to obtain facial feature data of the first subject. The facial feature data of the first subject can be in vector form, which is more efficient in transmission and detection and recognition than the image data of the first subject. This further reduces interference from irrelevant features and improves the accuracy of detection and recognition of the first subject.

[0083] Optionally, performing a second feature extraction on the compensated image data to obtain facial feature data of the first object includes:

[0084] performing a second feature extraction on the compensated image data to obtain data of a plurality of facial feature points of the first object;

[0085] Feature fusion is performed on data of multiple facial feature points of the first object to obtain facial feature data of the first object.

[0086] In this embodiment, an edge device is positioned near a data source to obtain first video stream data within a preset area. First, the edge device performs a first feature extraction on the first video stream data, filtering and obtaining image data of a first object, thereby reducing network transmission traffic. The edge device then performs illumination compensation based on the image data of the first object to obtain compensated image data, thereby preventing poor illumination conditions that could reduce recognition accuracy of the first object. The key algorithm can be a modified Resnet18 network, using a histogram equalization algorithm to perform illumination compensation on the image data of the first object.

[0087] Then, a second feature extraction is performed on the compensated image data to obtain data on multiple facial feature points of the first subject. Feature fusion is then performed on the data on multiple facial feature points of the first subject to obtain facial feature data of the first subject. The facial feature data of the first subject can be in vector form, which allows for more efficient transmission and detection and recognition than image data of the first subject. This further reduces interference from irrelevant features and improves the accuracy of detection and recognition of the first subject.

[0088] In the first convolutional layer of the Resnet18 network, two BasicBlocks can be improved. The input feature map is first subjected to BN regularization before entering the activation function. The activation function ReLU function is improved to LeakyReLU function, and the output data is subjected to 3*3 convolution. This operation is repeated once to fuse the data of multiple facial feature points of the first object. The second and fourth convolutional layers remain unchanged. The two 3*3 convolutions in the two BasicBlocks of the third convolutional layer are replaced with a single 5*5 convolution. The improved Resnet18 network is better at extracting deep features of the input image, and the detection accuracy is significantly improved.

[0089] Optionally, the edge device performs second feature extraction based on the image data of the first object to obtain facial feature data of the first object, including:

[0090] In a case where the edge device needs to perform a second feature extraction on the image data of the first object, the edge device detects whether a parameter for the second feature extraction is updated;

[0091] When the parameters used for the second feature extraction are updated, the second feature extraction is performed on the image data of the first object using the updated parameters.

[0092] In this embodiment, the edge device sends image data of a first subject to the cloud device, which then compares facial feature information in a database to perform identity recognition. If the distance between the first subject's image data and a reference feature is less than or equal to a threshold, the first subject is determined to be the user identity corresponding to the reference feature; if the distance between the first subject's image data and the reference feature is greater than the threshold, the first subject is determined to be an unknown identity.

[0093] Among them, the threshold is predetermined manually, and the level of the threshold will have a great impact on the reliability and convenience of the identity recognition system. Specifically, if the threshold is defined to be lower, the probability of identity misjudgment during identity recognition can be reduced, but the probability of identity misjudgment will increase, wherein identity misjudgment refers to the identity recognition system misjudging person B as identity a, and identity omission refers to the identity recognition system being unable to determine person A as identity a (because the current image of person A is different from the image at the time of face login). On the other hand, if the threshold is defined to be higher, the probability of identity misjudgment during identity recognition can be reduced, but the probability of identity omission will increase. In the detection and recognition method of the present invention, when the edge device needs to perform a second feature extraction on the image data of the first object, the edge device detects whether the parameters used for the second feature extraction are updated; when the parameters used for the second feature extraction are updated, the updated parameters are used to perform a second feature extraction on the image data of the first object. In this way, the cloud device performs identity recognition based on the feature data of the first object after performing second feature extraction based on the image data of the first object. Since the algorithm has undergone multiple iterative optimizations, a threshold that generally meets the project requirements can be selected as the final threshold for setting. That is, during the verification process, the threshold achieves good results in both the missed detection rate and the false detection rate of the verification set.

[0094] See also Figure 3 , Figure 3 FIG. 1 is a flow chart of another detection and identification method provided by an embodiment of the present invention. Figure 3 As shown, another detection and identification method provided by an embodiment of the present invention includes the following steps:

[0095] Step 301: The cloud device receives image data of a first object from the edge device;

[0096] Step 302: The cloud device performs identity recognition on the first object based on the image data of the first object.

[0097] Optionally, the image data of the first object is facial feature data of the first object.

[0098] In this embodiment, an edge device is positioned near a data source to obtain first video stream data within a preset area. First, the edge device performs a first feature extraction on the first video stream data, filtering and obtaining image data of a first object, thereby reducing network transmission traffic. The edge device then performs illumination compensation based on the image data of the first object to obtain compensated image data, thereby preventing poor illumination conditions that could reduce recognition accuracy of the first object. The key algorithm can be a modified Resnet18 network, using a histogram equalization algorithm to perform illumination compensation on the image data of the first object.

[0099] Then, a second feature extraction is performed on the compensated image data to obtain data on multiple facial feature points of the first subject. This data is then feature-fused to obtain facial feature data of the first subject. The facial feature data of the first subject can be in vector form, which allows for more efficient transmission and detection and recognition compared to the image data of the first subject. This further reduces interference from irrelevant features, improving the accuracy of detection and recognition of the first subject. Specifically, in the first convolutional layer of the Resnet18 network, two BasicBlocks can be improved: the input feature map is first subjected to BN regularization before entering the activation function. The activation function is then improved to a LeakyReLU function, and the output data is subjected to a 3x3 convolution. This process is repeated once to fuse the data on multiple facial feature points of the first subject. The second and fourth convolutional layers remain unchanged, while the two 3x3 convolutions in the two BasicBlocks of the third convolutional layer are replaced with a single 5x5 convolution. This improved Resnet18 network is better able to extract deep features from the input image, significantly improving detection accuracy.

[0100] Edge recognition then sends the facial feature data of the first subject to the cloud device, which then identifies the first subject. This cloud-edge distributed computing architecture allows the existing cloud server capacity and computing power to meet the needs of detecting and identifying the first subject, even when the amount of data being received is large. This improves the computing speed of the device and extends its service life.

[0101] In one example, industrial gateways can be used within a factory to attach sensor devices to various industrial products and equipment, enabling them to collect data, coordinate communication between different data sources, and analyze and transmit data. This data is then transmitted to the platform, where software processing and optimization are used to create corresponding services. Edge computing utilizes a distributed computing architecture, distributing computations to devices close to the data source, offloading the workload from the cloud. Data collection is the foundation of edge computing. As the number of devices increases, it becomes unrealistic to burden the cloud with data transmission and computation for all devices. Therefore, devices at the edge of the network (such as edge computing gateways) are beginning to incorporate computational analysis capabilities. They can preprocess data collected near sensors and other IoT devices, reducing the amount of data before transmitting it back to the cloud. This reduces the pressure on the cloud.

[0102] See also Figure 4 , Figure 4 Schematic diagram of the structure of the edge device provided by the embodiment of the present invention. Figure 4 As shown, the edge device 400 includes:

[0103] An acquisition module 401 is configured to acquire first video stream data within a preset area;

[0104] A first feature extraction module 402 is configured to perform first feature extraction on the first video stream data to obtain image data of a first object;

[0105] The sending module 403 is configured to send the image data of the first object to a cloud device.

[0106] Optionally, the edge device 400 further includes:

[0107] a second feature extraction module, configured to perform second feature extraction based on the image data of the first object to obtain facial feature data of the first object;

[0108] The sending module 403 includes:

[0109] The sending submodule is used to send the facial feature data of the first object to the cloud device.

[0110] Optionally, the second feature extraction module includes:

[0111] a compensation submodule, configured to perform illumination compensation on the image data of the first object to obtain compensated image data;

[0112] The feature extraction submodule is used to perform a second feature extraction on the compensated image data to obtain facial feature data of the first object.

[0113] Optionally, the feature extraction submodule includes:

[0114] a feature extraction unit, configured to perform a second feature extraction on the compensated image data to obtain data of a plurality of facial feature points of the first object;

[0115] The feature fusion unit is used to perform feature fusion on the data of multiple facial feature points of the first object to obtain facial feature data of the first object.

[0116] Optionally, the second feature extraction module includes:

[0117] a detection submodule, configured to, when the edge device needs to perform second feature extraction on the image data of the first object, detect whether parameters used for the second feature extraction are updated;

[0118] An updating submodule is configured to, when the parameters used for extracting the second feature are updated, perform second feature extraction on the image data of the first object using the updated parameters.

[0119] The edge device provided by the embodiment of the present invention can achieve Figure 1 The various processes implemented in the method embodiment shown can achieve the same beneficial effects, and will not be described again here to avoid repetition.

[0120] See also Figure 5 , Figure 5 This is a schematic diagram of the structure of the cloud device provided by the embodiment of the present invention. Figure 5 As shown, the cloud device 500 includes:

[0121] A receiving module 501 is configured to receive image data of a first object from an edge device;

[0122] The identification module 502 is configured to perform identity recognition on the first object based on the image data of the first object.

[0123] Optionally, the image data of the first object is facial feature data of the first object.

[0124] The cloud device provided by the embodiment of the present invention can achieve Figure 3 The various processes implemented in the method embodiment shown can achieve the same beneficial effects, and will not be described again here to avoid repetition.

[0125] It should be noted that, in this article, the terms "comprise", "include" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a..." does not exclude the presence of other identical elements in the process, method, article or device comprising the element. In addition, it should be pointed out that the scope of the methods and devices in the embodiments of the present invention is not limited to performing functions in the order discussed, but may also include performing functions in a substantially simultaneous manner or in the opposite order according to the functions involved. For example, the described method may be performed in an order different from that described, and various steps may also be added, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.

[0126] Through the description of the above embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus the necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better embodiment. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present invention.

[0127] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the present invention and the claims, all of which are protected by the present invention.

Claims

1. A detection and identification method, characterized in that: The method comprises: The edge device obtains first video stream data in a preset area; The edge device performs a first feature extraction on the first video stream data to obtain image data of the first object; The edge device sends the image data of the first object to the cloud device; After obtaining the image data of the first object and before the edge device sends the image data of the first object to the cloud device, the method further includes: The edge device performs a second feature extraction based on the image data of the first object to obtain facial feature data of the first object, including: performing illumination compensation on the image data of the first object to obtain compensated image data; performing a second feature extraction on the compensated image data to obtain facial feature data of the first object, wherein an improved Resnet18 network is adopted and a histogram equalization algorithm is used to perform illumination compensation on the image data of the first object, wherein the improvement of the Resnet18 network includes: in the first convolution layer of the Resnet18 network, two Basic Blocks are improved, the input feature map is first BN regularized, and then enters the activation function, the activation function ReLU function is improved to the LeakyReLU function, the output data is 3*3 convolved, the above operation is repeated once, and the data of multiple facial feature points of the first object are fused; the second and fourth convolution layers remain unchanged, and the two 3*3 convolutions in the two BasicBlocks of the third convolution layer are replaced with one 5*5 convolution.

2. The method according to claim 1, characterized in that The edge device sends the image data of the first object to the cloud device, including: The edge device sends the facial feature data of the first object to the cloud device.

3. The method according to claim 1, characterized in that The performing a second feature extraction on the compensated image data to obtain facial feature data of the first object includes: performing a second feature extraction on the compensated image data to obtain data of a plurality of facial feature points of the first object; Feature fusion is performed on data of multiple facial feature points of the first object to obtain facial feature data of the first object.

4. The method according to claim 2, characterized in that The edge device performs second feature extraction based on the image data of the first object to obtain facial feature data of the first object, including: In a case where the edge device needs to perform a second feature extraction on the image data of the first object, the edge device detects whether a parameter for the second feature extraction is updated; When the parameters used for the second feature extraction are updated, the second feature extraction is performed on the image data of the first object using the updated parameters.

5. A detection and identification method, characterized in that: The method comprises: The cloud device receives image data of the first object from the edge device in the detection and recognition method as claimed in claim 1; The cloud device performs identity recognition on the first object based on the image data of the first object.

6. The method according to claim 5, characterized in that The image data of the first object is facial feature data of the first object.

7. An edge device, configured to execute the detection and identification method according to any one of claims 1 to 4, characterized in that: The edge device includes: An acquisition module, configured to acquire first video stream data within a preset area; A first feature extraction module, configured to extract a first feature from the first video stream data to obtain image data of a first object; A sending module is used to send the image data of the first object to a cloud device.

8. The edge device according to claim 7, wherein: The edge device further includes: a second feature extraction module, configured to perform second feature extraction based on the image data of the first object to obtain facial feature data of the first object; The sending module includes: The sending submodule is used to send the facial feature data of the first object to the cloud device.

9. A cloud device for executing the detection and identification method according to claim 5 or 6, characterized in that: The cloud device includes: a receiving module, configured to receive image data of a first object from an edge device; The recognition module is configured to perform identity recognition on the first object based on the image data of the first object.

Images