Object recognition method and device, equipment and storage medium

Abstract

The embodiment of the invention discloses an object recognition method and device, equipment and a storage medium, and the method comprises the steps: obtaining object images collected by a plurality of pieces of equipment in a physical space at a current moment through the local connection between all pieces of equipment in the same physical space; recording a mapping relationship between each acquired object image and each device in the plurality of devices; performing synthesis compression processing on each object image to obtain a target image; sending the target image to a server, enabling the server to cut the target image, performing attribute recognition on an object in each cut object image, and returning each recognized attribute recognition result; and after receiving each attribute identification result returned by the server, scheduling each attribute identification result to corresponding equipment according to the recorded mapping relationship. According to the embodiment of the invention, the number of requests for the server can be effectively reduced, so that the network flow is saved, and the operation pressure of the server side is reduced.

Description

Technical Field

[0001] The present application relates to the field of Internet technology, specifically to the field of image processing technology, and in particular to an object recognition method, device, equipment and storage medium. Background Art

[0002] With the popularization of the concept of artificial intelligence (AI), object recognition has become a hot research topic; the so-called object recognition refers to a technology that uses computers to process, analyze and understand object images to identify the attribute information of objects in the object images. As the demand for object recognition increases, multiple devices for image acquisition are usually deployed in the same physical space, so that these devices can capture images of different objects at the same time, and then the server can recognize the attribute information of the corresponding object based on the collected object images.

[0003] At present, when multiple devices are deployed in the same physical space, after each device collects an object image, it will directly send the collected object image to the server through the communication connection between the device and the server, so that the server can identify the attributes of the object in the object image and return the identified attribute information to the corresponding device. This processing method will cause each device in the same physical space to continuously send uplink requests to the server, which will not only cause excessive operating pressure on the server side, but also occupy a lot of network traffic. Summary of the invention

[0004] The embodiments of the present application provide an object identification method, apparatus, device and storage medium, which can effectively reduce the number of requests to the server, thereby saving network traffic and reducing the operating pressure on the server side.

[0005] On the one hand, an embodiment of the present application provides an object recognition method, which is performed by a target device and includes:

[0006] Obtaining, through local connections between devices located in the same physical space, object images respectively captured by multiple devices in the physical space at the current moment; wherein the local connection refers to a communication connection that does not consume network traffic;

[0007] Recording the mapping relationship between each acquired object image and each device in the plurality of devices; and performing synthesis compression processing on each object image to obtain a target image;

[0008] Sending the target image to a server, causing the server to cut the target image, and respectively identify the attributes of the objects in each of the cut object images, and returning each attribute recognition result;

[0009] After receiving each attribute recognition result returned by the server, each attribute recognition result is dispatched to a corresponding device according to the recorded mapping relationship.

[0010] On the other hand, an embodiment of the present application provides an object recognition device, the device running in a target device, the device comprising:

[0011] An acquisition unit, configured to acquire, through a local connection between devices located in the same physical space, the object images respectively acquired by multiple devices in the physical space at the current moment; wherein the local connection refers to a communication connection that does not consume network traffic;

[0012] A processing unit, used to record the mapping relationship between each acquired object image and each device in the plurality of devices; and perform synthesis compression processing on each object image to obtain a target image;

[0013] A communication unit, used to send the target image to a server, so that the server cuts the target image, and respectively identifies the attributes of the objects in each of the cut object images, and returns each attribute identification result;

[0014] The communication unit is further configured to dispatch each attribute recognition result to a corresponding device according to the recorded mapping relationship after receiving each attribute recognition result returned by the server.

[0015] In one embodiment, each acquired object image includes an object image captured by the target device at the current moment;

[0016] Accordingly, the processing unit may be specifically used for:

[0017] At the current moment, an image of an object located within a shooting range is collected to obtain streaming media data of the object; the streaming media data includes a plurality of images, and each image includes the object;

[0018] Performing image quality evaluation on each image in the streaming media data respectively to obtain an image quality score of each image;

[0019] Based on the image quality scores of the respective images, an image is selected from the streaming media data as an object image collected by the target device at the current moment.

[0020] In another implementation manner, when the processing unit is used to respectively evaluate the image quality of each image in the streaming media data to obtain the image quality score of each image, it can be specifically used to:

[0021] Obtain multiple preset evaluation indicators and the evaluation rules corresponding to each evaluation indicator;

[0022] Under each evaluation index, quality evaluation is performed on each image in the streaming media data according to corresponding evaluation rules to obtain a quality score of each image under each evaluation index;

[0023] Dynamically determine the indicator weight value corresponding to each evaluation indicator according to the quality score of each image under each evaluation indicator;

[0024] For any image in the streaming media data, the indicator weight values ​​corresponding to the various evaluation indicators are used to perform weighted summation on the quality score of the any image under each evaluation indicator to obtain the image quality score of the any image.

[0025] In another implementation manner, when the processing unit is used to dynamically determine the indicator weight value corresponding to each evaluation indicator according to the quality score of each image under each evaluation indicator, it can be specifically used to:

[0026] Traversing the multiple evaluation indicators, and taking the currently traversed evaluation indicator as the current evaluation indicator;

[0027] According to the quality scores of the respective images under the current evaluation index, statistically analyzing the score distribution corresponding to the current evaluation index;

[0028] If the score distribution is used to indicate that the corresponding quality scores are concentratedly distributed, the indicator weight value of the current evaluation indicator is determined as a first value;

[0029] If the score distribution is used to indicate that the corresponding quality scores are sparsely distributed, the indicator weight value of the current evaluation indicator is determined as a second value;

[0030] The first value is smaller than the second value.

[0031] In another implementation manner, when the processing unit is used to perform synthesis compression processing on the respective object images to obtain the target image, it can be specifically used to:

[0032] Acquire a blank image, where the blank image refers to an image without storing image content;

[0033] The image contents in the respective object images are extracted respectively, and the extracted image contents are sequentially filled into the blank images to obtain the target image.

[0034] In another embodiment, the processing unit may also be used to: determine the indication information of each image content according to the arrangement of each image content in the target image; the indication information of any image content is used to indicate: the position of the corresponding image content in the target image, and the size of the object image corresponding to the corresponding image content;

[0035] The communication unit may also be used to: send the indication information of the contents of each image to the server, so that the server performs segmentation processing on the target image according to the indication information of the contents of each image;

[0036] The cutting process includes: extracting corresponding image content from the target image according to the indication information of any image content, and generating an object image by using the extracted image content.

[0037] In another implementation manner, the acquisition unit acquires the object images respectively acquired by the multiple devices at the current moment after the target device is elected as the main device; accordingly, the processing unit may also be used to:

[0038] Acquire network performance information of the target device, the network performance information including at least one of the following: network bandwidth, network delay, network jitter, and network packet loss rate;

[0039] Acquire network performance information of each device other than the target device in the physical space through local connections between the devices in the physical space;

[0040] According to the network quality determination rule that has been agreed upon by each device in the physical space, the network quality of the corresponding device is determined according to each acquired network performance information;

[0041] According to the network quality of the target device and the network quality of each other device, a device with the highest network quality is selected from each device in the physical space as the main device.

[0042] In another implementation manner, the network performance information of any device is generated and signed by the server; accordingly, the processing unit may also be used to:

[0043] Whenever network performance information of a device is obtained, the signature information of the server is searched in the currently obtained network performance information;

[0044] If the signature information of the server is found, it is determined that the currently acquired network performance information is credible information; if the signature information of the server is not found, it is determined that the currently acquired network performance information is suspicious information;

[0045] After determining that each acquired network performance information is credible information, the step of triggering the execution of the network quality determination rule that has been agreed upon by each device in the physical space, and determining the network quality of the corresponding device according to each acquired network performance information.

[0046] In another implementation manner, the acquisition unit acquires the object images respectively acquired by the multiple devices at the current moment after the target device is elected as the main device; accordingly, the processing unit may also be used to:

[0047] Acquire network performance information of the target device, the network performance information including at least one of the following: network bandwidth, network delay, network jitter, and network packet loss rate;

[0048] Acquiring network performance information of each device other than the target device in the physical space through local connections between the devices in the physical space, and sending the network performance information of the target device to each of the other devices;

[0049] According to the obtained network performance information, voting is performed among the devices in the physical space to elect a master device to obtain an election result; and through the local connection between the devices in the physical space, the election results obtained by other devices based on the network performance information are obtained;

[0050] According to each election result, a master device is determined from each device in the physical space.

[0051] In another implementation manner, the acquisition unit acquires the object images respectively acquired by the multiple devices at the current moment after the target device is elected as the main device; accordingly, the processing unit may also be used to:

[0052] Receiving a heartbeat instruction sent by the server;

[0053] Returning specified data to the server in response to the heartbeat instruction, so that the server determines the current network communication level of the target device according to the specified data returned by the target device;

[0054] The server selects a master device from the devices in the physical space according to the current network communication level of the target device and the current network communication level of other devices in the physical space except the target device.

[0055] In another aspect, an embodiment of the present application provides a computer device, the computer device comprising an input interface and an output interface, the computer device further comprising:

[0056] processor and computer storage media;

[0057] The processor is suitable for implementing one or more instructions, the computer storage medium stores one or more instructions, and the one or more instructions are suitable for being loaded by the processor and executing the above-mentioned object recognition method.

[0058] On the other hand, an embodiment of the present application provides a computer storage medium, which stores one or more instructions, and the one or more instructions are suitable for being loaded by a processor and executing the above-mentioned object recognition method.

[0059] On the other hand, an embodiment of the present application provides a computer program product, which includes one or more instructions; when one or more instructions in the computer program are executed by a processor, the above-mentioned object recognition method is implemented.

[0060] The embodiment of the present application can establish a local connection between various devices in the same physical space, so that the object images collected by multiple devices in the same physical space at the same time can be aggregated through the established local connection, so that when the object images are aggregated, the network traffic of each device does not need to be consumed, and the network traffic occupied by the object images is saved. Secondly, the mapping relationship between the aggregated object images and each device can be recorded, and the aggregated object images can be synthesized and compressed, so that the target image obtained by the synthesis compression is sent to the server, so as to realize the cutting of the target image on the back-end server side, and the attribute recognition of the objects in the cut object images respectively; this can effectively reduce the uplink request, thereby effectively reducing the number of requests to the server, reducing the occupation of the network traffic by the object image, and further saving the network traffic and reducing the operating pressure of the server. Then, the server can return the identified attribute recognition results to the communicating device, so that the device can perform unified result distribution according to the recorded mapping relationship, so as to realize the dispatching of each attribute recognition result back to the corresponding device; this can make the server not need to communicate with each device, effectively reduce the downlink request of the server, thereby further saving the network traffic and reducing the operating pressure of the server. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0062] Figure 1 is a schematic diagram of the architecture of an object recognition system provided in an embodiment of the present application;

[0063] Figure 2 It is a flowchart of an object recognition method provided in an embodiment of the present application;

[0064] Figure 3a This is a schematic diagram of collecting streaming media data provided by an embodiment of the present application;

[0065] Figure 3b It is a schematic diagram of synthesizing and compressing multiple object images to obtain a target image provided by an embodiment of the present application;

[0066] Figure 3c is another schematic diagram of synthesizing and compressing multiple object images to obtain a target image provided by an embodiment of the present application;

[0067] Figure 3d is a schematic diagram of establishing a two-dimensional coordinate system on a target image provided by an embodiment of the present application;

[0068] Figure 4 This is a schematic diagram of a scenario of face payment based on face recognition provided by an embodiment of the present application;

[0069] Figure 5 This is a flow chart of a face-swiping payment based on face recognition provided by an embodiment of the present application;

[0070] Figure 6 is a structural schematic diagram of an object recognition device provided in an embodiment of the present application;

[0071] Figure 7 It is a structural diagram of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0072] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0073] The embodiments of the present application relate to AI (Artificial Intelligence) technology, and the so-called AI technology refers to the theory, method, technology and application system of using digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. In other words, AI is a comprehensive technology of computer science; it mainly produces a new intelligent machine that can respond in a similar way to human intelligence by understanding the essence of intelligence, so that the intelligent machine has multiple functions such as perception, reasoning and decision-making. Correspondingly, AI technology is a comprehensive discipline, which mainly includes computer vision technology (Computer Vision, CV), speech processing technology, natural language processing technology and machine learning (Machine Learning, ML) / deep learning and other major directions. Among them, computer vision technology is a science that studies how to make machines "see". To put it more concretely, it refers to machine vision such as using cameras and computers to replace human eyes to identify and measure targets, and further do graphic processing to make computer processing a technology that is more suitable for human eye observation or transmission to instrument detection.

[0074] In order to better realize object recognition, the embodiment of the present application proposes an object recognition system and a corresponding object recognition method based on the computer vision technology in AI technology. The so-called object recognition refers to the processing of the attribute information of the object in the object image. The object here can be the face of any creature (such as a human face, a cat face, etc.), a gesture, a fingerprint, or a whole animal (such as a cat, a dog), etc. The attribute information of the object refers to information that can be used to characterize some features of the object. For example, if the object is a human face, the attribute information may be the identity information corresponding to the human face; for example, if the object is a gesture, the attribute information may be the gesture information corresponding to the gesture (such as the shape of the gesture, the position information of each node in the hand, etc.); for example, if the object is a fingerprint, the attribute information may be the identity information corresponding to the fingerprint; for example, if the object is a whole animal, the attribute information may be the type of animal, and so on.

[0075] See also Figure 1 As shown, the object recognition system may include at least: multiple devices 11 and servers 12 located in the same physical space. A network connection can be established between each device 11 and the server 12, and network communication can be performed based on the established network connection; the so-called network connection refers to a communication connection that consumes network traffic, and correspondingly, network communication refers to communication that consumes network traffic. In addition, a local connection can be established between the devices 11 in the physical space, and local communication can be performed through the established local connection; the so-called local connection refers to a communication connection that does not consume network traffic, and correspondingly, local communication refers to communication that does not consume network traffic. It should be understood that Figure 1It is only an exemplary representation of the object recognition system, and does not limit the architecture of the object recognition system, such as not limiting the number of devices 11.

[0076] Among them, the device 11 refers to any device that can call a traditional camera component or a 3D (three-dimensional) camera component to collect an object image (i.e., an image containing an object). Specifically, the device 11 can be a smart phone, a computer (such as a tablet computer, a laptop computer, a desktop computer, etc.), a face payment device (a device that automatically deducts electronic resources from a resource account associated with the identity information after performing face recognition to obtain identity information), an automatic cash register device, etc. The server 12 refers to a service device that can provide cloud object recognition services and a series of other services; the so-called object recognition service refers to: a service that recognizes the attribute information of an object in an object image. Specifically, the server 12 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN (Content Delivery Network), and basic cloud computing services such as big data and artificial intelligence platforms, etc.

[0077] The traditional camera assembly mentioned above refers to a camera assembly for collecting color images (i.e., RGB images), while the 3D camera assembly refers to a camera assembly that adds relevant software and hardware for liveness detection on the basis of the traditional camera assembly to ensure information security. For example, the 3D camera assembly may also include: a depth camera for collecting depth images and an infrared camera for collecting infrared images, etc. on the basis of the traditional camera assembly. Among them, an RGB image refers to an image that uses the RGB values ​​of each point in the scene as pixel values, that is, an RGB image refers to an image with RGB values. A depth image, which can also be called a distance image frame, refers to an image that uses the distance of each point in the scene relative to the depth camera as a pixel value, that is, a depth image refers to an image with depth information. An infrared image refers to an image formed by an infrared camera by collecting the radiation of each point in the scene in the infrared band, that is, an infrared image refers to an image with infrared information. It can be seen that the object image mentioned in the embodiment of the present application can be an RGB image, a depth image, or an infrared image, without limitation.

[0078] In a specific implementation, the general principle of the object recognition method proposed based on the above-mentioned object recognition system is as follows: when multiple devices are deployed in the same physical space, a certain device in the physical space is supported to aggregate the object images simultaneously collected by multiple devices in the physical space through local communication, identify the source of each object image and synthesize and compress each object image into a target image, thereby sending the target image to the server, so that the server can cut the target image in the back end and identify the attributes of the objects in each cut object image, and then return the identified attribute recognition results to the device end, and the device will distribute the results uniformly according to the sources of the previously identified object images (that is, return each attribute recognition result to the corresponding device).

[0079] It can be seen that the object recognition method proposed in the embodiment of the present application can effectively reduce uplink requests by locally aggregating and synthesizing multiple object images to send the synthesized compressed target image to the server, thereby effectively reducing the number of requests to the server, reducing the object image's occupation of network traffic, and thus saving network traffic and reducing the server's operating pressure. In addition, by uniformly receiving the various attribute recognition results returned by the server and uniformly distributing the received attribute recognition results, the server does not need to communicate with each device, effectively reducing the server's downlink requests, thereby further saving network traffic and reducing the server's operating pressure.

[0080] It should be noted that in the specific implementation of the object recognition method embodiment proposed in the embodiment of the present application, relevant data such as object images are involved. When the above method embodiment is applied to specific products or technologies, these relevant data are collected with the user's permission or consent, and the collection, use and processing of the relevant data comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0081] Combine the following Figure 2 The flowchart shown in the figure illustrates the specific implementation of the object recognition method proposed in the embodiment of the present application; in the embodiment of the present application, the object recognition method can be executed by a target device, and the target object refers to one of the devices 11 in the object recognition system mentioned above. Figure 2 As shown, the object recognition method may include the following steps S201-S205:

[0082] S201, obtaining object images respectively collected by multiple devices in the physical space at the current moment through local connections between various devices located in the same physical space.

[0083] The same physical space mentioned here refers to the physical space where the target device is located, and the so-called physical space refers to a real three-dimensional space area. In different application scenarios, the physical space can be different three-dimensional space areas. For example, in the face payment scenario, the physical space can be a store or a scenic spot; in another example, in the face check-in scenario or fingerprint check-in scenario, the physical space can be the check-in area of ​​a certain office location or office building, and so on.

[0084] In a specific implementation, the local connection between the devices in the physical space may be a short-range wireless communication connection; the so-called short-range wireless communication connection refers to a communication connection that supports short-range communication, and short-range communication refers to communication that transmits data via radio waves and the transmission distance is less than a distance threshold. In this specific implementation, the specific implementation method of the target device executing step S201 may be: through the short-range wireless communication connection between the devices located in the same physical space, receiving image recognition requests sent by multiple devices in the physical space, and respectively parsing the image recognition requests sent by each device to obtain the object images collected by the corresponding devices at the current moment.

[0085] Exemplarily, the short-range wireless communication connection mentioned above may include: Bluetooth connection, Near Field Communication (NFC) connection, or wireless (Wi-Fi, WIFI) connection, etc., without limitation. Among them, Bluetooth is a radio communication technology that supports a communication distance of ten meters or tens of meters; it can establish a universal radio air interface (Radio Air Interface) for the communication environment between fixed devices or mobile devices, further combining communication technology with computer technology, so that various devices can communicate or operate with each other in a short range without being connected to each other by wires or cables. NFC is a high-frequency wireless communication technology that supports a short-range communication distance of several centimeters, more than ten centimeters or tens of centimeters; it is a technology developed on the basis of contactless radio frequency identification (RFID) technology and combined with wireless interconnection technology, and the "near field" in the Chinese name of NFC refers to radio waves near the electromagnetic field. WIFI is a wireless networking technology that supports a communication distance of more than ten meters or tens of meters; it can mainly achieve networking through radio waves. If a device transmits a WIFI signal, other devices within the effective range of the device's radio wave coverage can use WIFI connection to connect to the network.

[0086] In another specific implementation, the local connection between the devices in the physical space can be a connection established based on a removable storage medium, and the so-called removable storage medium refers to hardware that is removable and has storage functions. In this specific implementation, the user can manually copy the object images collected by multiple devices in the physical space at the current moment to the removable storage medium, and then insert the removable storage medium into the target device so that the removable storage medium and the target device are connected; accordingly, the specific implementation method of the target device executing step S201 can be: through the connection with the removable storage medium, read the object images collected by multiple devices in the physical space at the current moment from the removable storage medium.

[0087] Exemplarily, the removable storage media mentioned above may include: a USB flash drive (Universal Serial Bus, U disk), a mobile hard disk, a memory card or a floppy disk, etc., without limitation. Among them, a USB flash drive is a miniature high-capacity mobile storage that uses a USB (Universal Serial Bus) interface and does not require a physical driver. It can be connected to a device through a USB interface to achieve plug-and-play. A mobile hard disk refers to a small and portable hard disk storage that can transfer data with a device at a high speed. A memory card refers to an independent storage medium that can be used on a device and is in the form of a card. A floppy disk is the earliest removable medium used, and the reading and writing of a floppy disk can be completed through a floppy disk drive.

[0088] It should be noted that the current moment mentioned in the embodiments of the present application can be understood as: the latest moment displayed by the system clock of the device (i.e. the current moment); it should be understood that the moment displayed by the system clock of the device can be automatically updated as time passes. For example, if the latest moment displayed by the system clock of the device is 10:40, the current moment is the moment of 10:40; after 10 minutes, the system clock of the device will display 10:50, and the current moment at this time is the moment of 10:50. In addition, when the object mentioned in the embodiments of the present application is a face, the object image can be understood as a face image; when the object is a gesture, the object image can be understood as a gesture image; when the object is a fingerprint, the object image can be understood as a fingerprint image, and so on.

[0089] S202: Record the mapping relationship between each acquired object image and each device in the multiple devices.

[0090] In a specific implementation, the target device can determine the image identification of each acquired object image, and obtain the device identification of the corresponding device from each device that sent the object image through the local connection between each device in the corresponding physical space. The device identification of any device can be obtained together with the object image acquired by any device; or it can be obtained separately after the object image acquired by the corresponding device is acquired, and there is no limitation on this. Then, for any object image, a mapping relationship can be established between the image identification of any object image and the device identification of the device that acquired the corresponding object image, and the established mapping relationship can be stored in the local space of the target device.

[0091] Among them, the image identifier of any object image mentioned above can be an image name automatically generated when the corresponding object image is collected, or it can be an image number assigned to each object image by the target device after acquiring each object image, etc., without limitation. In addition, the device identifier mentioned above can be SN (Serial Number, string serial number), and SN can be understood as ID (Identity document, identity identification number) that can uniquely identify the device. Alternatively, the device identifier can also be other information that can uniquely identify the device, such as the device name, device number or device model, without limitation.

[0092] It should be noted that, in a specific implementation, if at the current moment, there is an object within the shooting range of the target device, the target device may also perform image acquisition on the corresponding object at the current moment to obtain an object image; in this case, the acquired object images mentioned in step S202 may include: the object image acquired by the target device at the current moment, and the object images acquired by the target device through a local connection at the current moment (i.e., the object image acquired through step S201). If at the current moment, there is no object within the shooting range of the target device, the target device does not perform image acquisition at the current moment; in this case, the acquired object images mentioned in step S202 may include: the object images acquired by the target device through a local connection at the current moment (i.e., the object image acquired through step S201).

[0093] Among them, the process of any device collecting the object image at the current moment may include: calling a camera component (such as a 3D camera component) at the current moment to collect an image of the object within the shooting range, and using the collected image as the object image collected by the corresponding device at the current moment. Alternatively, considering that the collected image may have low image quality due to various factors (such as object movement, lens shaking, etc.) during image collection, thereby affecting the subsequent recognition results, the embodiment of the present application can obtain streaming media data through continuous image collection, thereby selecting images with better quality from the streaming media data as object images to ensure the accuracy of subsequent recognition results. Based on this, the process of any device collecting the object image at the current moment includes the following steps s11-s13:

[0094] s11, capturing images of objects within the shooting range at the current moment to obtain streaming media data of the objects.

[0095] Specifically, any device can call a camera component (such as a 3D camera component) to continuously capture images of an object within the shooting range of the camera component at the current moment, thereby obtaining streaming media data of the object. The streaming media data includes multiple images, and each image includes an object. For example, a schematic diagram of streaming media data of a human face captured by any device through a camera component can be found in Figure 3a Optional, such as Figure 3a As shown, each time an image is acquired, the currently acquired image can be displayed on the display screen of any device.

[0096] s12, performing image quality assessment on each image in the streaming media data to obtain an image quality score for each image.

[0097] In a specific implementation, multiple preset evaluation indicators and evaluation rules corresponding to each evaluation indicator can be obtained. Secondly, under each evaluation indicator, the quality of each image in the streaming media data can be evaluated according to the corresponding evaluation rules to obtain the quality score of each image under each evaluation indicator. Then, the indicator weight value corresponding to each evaluation indicator can be dynamically determined according to the quality score of each image under each evaluation indicator. Alternatively, the indicator weight value corresponding to each evaluation indicator can also be preset. After obtaining the indicator weight value corresponding to each evaluation indicator, for any image in the streaming media data, the indicator weight value corresponding to each evaluation indicator is used to perform weighted summation on the quality score of any image under each evaluation indicator to obtain the image quality score of any image.

[0098] The following is a further description of some of the contents involved in this specific implementation:

[0099] (i) The multiple evaluation indicators mentioned above include at least two of the following: object size indicator, object angle indicator, image contrast indicator, image brightness indicator, and image clarity indicator, etc. It is worth emphasizing that the embodiment of the present application does not limit the evaluation rules under each evaluation indicator, and the specific evaluation method involved in each evaluation rule, which can be set according to business needs. Exemplary:

[0100] ① The evaluation rule corresponding to the object size index is used to indicate: the image quality is evaluated according to the size of the object in the image. Exemplarily, under the object size index, the method of performing quality evaluation on any image in the streaming media data according to the corresponding evaluation rule may include: according to the preset correspondence between the size of the object and the quality score, the quality score corresponding to the size of the object in any image is used as the quality score of the any image under the object size index. Alternatively, the size ratio between the size of the object in any image and the size of any image is calculated, and according to the preset correspondence between the size ratio and the quality score, the quality score corresponding to the calculated size ratio is used as the quality score of the any image under the object size index.

[0101] ② The evaluation rule corresponding to the object angle index is used to indicate: the image quality is evaluated according to the difference between the angle of the object in the image and the reference angle (set according to actual needs). Exemplarily, under the object angle index, the method of performing quality evaluation on any image in the streaming media data according to the corresponding evaluation rule may include: according to the principle that the angle difference and quality score are negatively correlated, the difference between the angle of the object in any image and the reference angle is scored for any image to obtain the quality score of the image under the object angle index.

[0102] ③ The evaluation rule corresponding to the image contrast index is used to indicate: the image quality is evaluated according to the contrast of the image. Alternatively, considering that there may be a situation where the contrast of the image as a whole is high, but the contrast of the image area where the object in the image is located is low, if the attribute information of the object is identified using such an image, the accuracy of the attribute identification result will be affected; it can be seen that in the scene of identifying attribute information, even if the contrast of the image is high, it does not mean that the image quality of the image is good. Based on this, in order to more accurately measure the image quality of the image, the embodiment of the present application proposes the concept of measuring the image quality by focusing on the contrast of the image area where the object is located; accordingly, the evaluation rule corresponding to the image contrast index is used to indicate: the image area where the object is located is cut out from the image, and the image quality is evaluated according to the contrast of the cut-out image area. Exemplarily, under the image contrast index, the method of performing quality evaluation on any image in the streaming media data according to the corresponding evaluation rule may include: cutting out the image area where the object is located from any image, and according to the preset correspondence between the contrast and the quality score, the quality score corresponding to the contrast of the cut-out image area is used as the quality score of the any image under the image contrast index.

[0103] ④ The evaluation rule corresponding to the image brightness index is used to indicate: perform quality evaluation on the image according to the brightness of the image. Alternatively, similar to the image contrast index, the evaluation rule corresponding to the image brightness index is used to indicate: extract the image area where the object is located from the image, and perform quality evaluation on the image according to the brightness of the extracted image area. Exemplarily, under the image brightness index, the method of performing quality evaluation on any image in the streaming media data according to the corresponding evaluation rules may include: extracting the image area where the object is located from any image, and according to the preset correspondence between brightness and quality score, using the quality score corresponding to the brightness of the extracted image area as the quality score of the any image under the image brightness index.

[0104] ⑤ The evaluation rule corresponding to the image clarity index is used to indicate: perform quality evaluation on the image according to the clarity of the image. Alternatively, similar to the image contrast index, the evaluation rule corresponding to the image clarity index is used to indicate: extract the image area where the object is located from the image, and perform quality evaluation on the image according to the clarity of the extracted image area. Exemplarily, under the image clarity index, the method of performing quality evaluation on any image in the streaming media data according to the corresponding evaluation rules may include: extracting the image area where the object is located from any image, and according to the preset correspondence between clarity and quality score, using the quality score corresponding to the clarity of the extracted image area as the quality score of the any image under the image clarity index.

[0105] (ii) The specific implementation methods of dynamically determining the indicator weight value corresponding to each evaluation indicator according to the quality score of each image under each evaluation indicator mentioned above may include any of the following:

[0106] In one embodiment, multiple evaluation indicators can be traversed, and the currently traversed evaluation indicator can be used as the current evaluation indicator; and according to the quality score of each image under the current evaluation indicator, the score distribution corresponding to the current evaluation indicator is statistically analyzed. If the score distribution is used to indicate that the corresponding quality scores are concentratedly distributed, the indicator weight value of the current evaluation indicator is determined as a first value; if the score distribution is used to indicate that the corresponding quality scores are sparsely distributed, the indicator weight value of the current evaluation indicator is determined as a second value; wherein the first value is less than the second value. Among them, the concentrated distribution of each quality score can be understood as: the difference between each quality score and the center point is less than or equal to the preset difference; on the contrary, the sparse distribution of each quality score can be understood as: there are one or more quality scores and the difference between the center point is greater than the preset difference. The center point mentioned here can be the mean between each quality score, or it can be the median of each quality score (that is, the quality score in the middle position after arranging each quality score in descending or ascending order). This implementation can achieve the following: when the difference between the quality scores of various images under the same evaluation index is small, a smaller index weight value is set for the evaluation index, so that when the quality scores under various evaluation indexes are weighted and summed to obtain the image quality score, the influence of the quality score under the evaluation index on the image quality score is weakened, so that the image quality scores of various images have certain differences, thereby facilitating the subsequent selection of object images.

[0107] In another implementation, multiple evaluation indicators can be traversed, and the currently traversed evaluation indicator is used as the current evaluation indicator. The initial weight value of the current evaluation indicator is obtained. If the quality score of each image under the current evaluation indicator is less than the target value, it means that the quality score under the current evaluation indicator has little effect on the image quality score. At this time, the initial weight value of the current evaluation indicator can be increased to obtain the indicator weight value of the current evaluation indicator; this can effectively increase the influence of the quality score under the current evaluation indicator on the image quality score, so that when determining the image index score of any image, the reference value of the quality score of the image under different evaluation indicators is the same. If the quality score of each image under the current evaluation indicator is greater than the target value, it means that the quality score under the current evaluation indicator has a greater influence on the image quality score. At this time, the initial weight value of the current evaluation indicator can be reduced to obtain the indicator weight value of the current evaluation indicator; this can effectively reduce the influence of the quality score under the current evaluation indicator on the image quality score, so that when determining the image index score of any image, the reference value of the quality score of the image under different evaluation indicators is the same.

[0108] s13, based on the image quality scores of each image, an image is selected from the streaming media data as the object image collected by any device at the current moment.

[0109] In a specific implementation, based on the image quality scores of each image, the image with the highest image quality score can be selected from the streaming media data as the object image collected by any device at the current moment. Alternatively, based on the image quality scores of each image, images with image quality scores greater than a score threshold can be screened from the streaming media data, and an image can be randomly selected from the screened images as the object image collected by any device at the current moment.

[0110] S203, performing synthesis compression processing on each object image to obtain a target image.

[0111] In a specific implementation, the target device can implement synthetic compression processing on each object image by means of splicing and compressing, so that the data volume of the generated target image is less than the sum of the data volumes of each object image. Specifically, the target device can first compress each object image, and then splice each compressed object image to obtain the target image. Alternatively, each object image can be spliced ​​first, and then the image obtained by the splicing process can be compressed to obtain the target image. In this specific implementation, the target image finally obtained can be understood as an image set composed of multiple object images; that is, the target image includes multiple independent object images, and different object images can be intuitively distinguished through the target image, such as Figure 3b shown.

[0112] It is worth emphasizing that in order to ensure that the object attributes can be recognized more accurately in the future, when compressing the image, it is necessary to ensure that the clarity of the compressed image is greater than the clarity threshold. Figure 3a It is only an example to characterize the form of the target image, and does not limit the actual stitching method used. For example, the corresponding object images can also be stitched in a horizontal stitching or a vertical stitching manner.

[0113] In another specific implementation, the target device can also extract the image content to perform synthetic compression processing on each object image, so that the data volume of the generated target image is much smaller than the sum of the data volumes of each object image. Specifically, the target device can first obtain a blank image, and the so-called blank image refers to an image that does not store image content. And, respectively extract the image content in each object image. Then, the extracted image content can be filled into the blank image in turn to obtain the target image. In this specific implementation, the target image finally obtained can be understood as an image composed of multiple image contents; that is, the target image includes multiple independent image contents, and different object images cannot be intuitively distinguished through the target image, such as Figure 3c shown.

[0114] It should be noted that the embodiments of the present application do not limit the method of extracting image content. For example, an image editing tool can be used to perform cutout processing on each object image to obtain the image content in each object image; or, the image content of each object image can be screenshotted to obtain the image content of each object image; or, a neural network model can be called to perform content extraction processing on each object image to obtain the image content in each object image. In addition, the embodiments of the present application do not limit the filling order of each object image and the filling logic of the blank image. For example, the filling logic can be in units of rows and start filling from top to bottom, or in units of columns and start filling from left to right.

[0115] In addition, the embodiment of the present application does not limit the execution order between step S202 and step S203. For example, step S202 may be executed first, and then step S203; step S203 may be executed first, and then step S202; step S202 and step S203 may also be executed simultaneously to improve processing efficiency.

[0116] S204, sending the target image to the server, so that the server cuts the target image, and respectively performs attribute recognition on the objects in each cut object image, and returns each attribute recognition result.

[0117] In a specific implementation, the target device can send the target image to the server through a network connection with the server. The network connection mentioned here can be, for example, a mobile communication connection based on the 4th Generation Mobile Communication Technology (4G), the 5th Generation Mobile Communication Technology (5G), Long Term Evolution (LTE), Worldwide Interoperability for Microwave Access (WiMAX), or a computer network communication connection based on the TCP / IP Protocol Suite (TCP / IP) or User Datagram Protocol (UDP).

[0118] As can be seen from the relevant description of the aforementioned step S203, the target image can be an image set composed of multiple object images, that is, different object images can be intuitively distinguished through the target image; in this case, the server can directly cut the target image according to the boundary lines between the various object images in the target image, so as to perform attribute recognition on the objects in the cut-out object images respectively. When performing attribute recognition on the object in any cut-out object image, the server can obtain a feature database, which includes multiple object features and attribute information corresponding to each object feature. A neural network is called to perform feature extraction on any object image to obtain the object features of the corresponding object. Then, the obtained object features are matched with the object features in the feature database to find the object features with the highest matching degree; specifically, the object features of the corresponding object can be scored according to the similarity between the object features of the corresponding object and any object features in the feature database, and the obtained score is used as the matching degree, so as to find the object features with the highest matching degree (i.e., the highest score). Then, the attribute information corresponding to the found object features in the feature database is added to the attribute recognition result of the object in the any object image.

[0119] Alternatively, the target image may be an image composed of multiple image contents, that is, different object images cannot be intuitively distinguished through the target image. In this case, in order to enable the server to accurately cut out each object image from the target image, the target device may also determine the indication information of each image content according to the arrangement of each image content in the target image; and send the indication information of each image content to the server, so that the server performs a cutting process on the target image according to the indication information of each image content, thereby performing attribute recognition on each object image cut out. The cutting process includes: according to the indication information of any image content, cutting out the corresponding image content from the target image, and using the cut out image content to generate an object image. The indication information of any image content is used to indicate: the position of the corresponding image content in the target image, and the size of the object image corresponding to the corresponding image content. Specifically, the server may determine a window for cutting out the image content, and scroll the target image according to the indication information of each image content, so that there is an image content in the target image that falls into the window, thereby cutting out the corresponding image content through the window and generating the corresponding object image. Then, the target image is continuously scrolled to update the image content displayed in the window, and then the corresponding image content is intercepted through the window and the corresponding object image is generated. This process is iterated until all the image content is intercepted.

[0120] It should be noted that the embodiments of the present application do not limit the method for determining the indication information of the image content. Exemplarily, a two-dimensional coordinate system can be constructed based on the reference point of the target image; the reference point of the target image mentioned here can be set according to business needs, for example, the reference point can be the upper right corner or upper left corner of the target image, etc. Then, based on the positional relationship between the center point of each image content and the reference point, the two-dimensional coordinates of the center point of each image content can be determined in the two-dimensional coordinate system to indicate the position of the corresponding image content in the target image through the two-dimensional coordinates. In this case, the indication information of any image content may include: the two-dimensional coordinates of the center point of the corresponding image content, and the image size of the corresponding image content, which image size is less than or equal to the size of the object image corresponding to the corresponding image content. When the size of the area occupied by the image content of any object image in the corresponding object image is much smaller than the size of the corresponding object image (that is, the difference between the size of the area and the size of the image is greater than a threshold), the image size in the indication information of the image content may be smaller than the size of the corresponding object image, and the image size in the indication information may be greater than or equal to the size of the area occupied by the image content in the corresponding object image. This ensures that the server can capture the complete image content according to the indication information, ensures the accuracy of subsequent attribute recognition, and avoids the existence of too much blank background in the object image generated by using the captured image content, thereby avoiding the server from wasting processing resources to identify the blank background and improving recognition efficiency.

[0121] Alternatively, a two-dimensional coordinate can be determined in a two-dimensional coordinate system according to the size of the object image corresponding to the image content, so as to indicate the position of the corresponding image content in the target image through the determined two-dimensional coordinate, and to indicate the size of the object image corresponding to the corresponding image content through the two-dimensional coordinate. In this case, the indication information of any image content may include: the two-dimensional coordinate determined according to the size of the corresponding object image. For example, the reference point is the upper right corner of the target image, and the two-dimensional coordinates are represented by x and y, with a space between x and y; then, the upper right corner of the target image is the origin (00), negative x values ​​indicate rightward, and negative y values ​​indicate downward, and the more negative the value, the greater the offset. See Figure 3dAs shown in the figure: Assuming that the size of the object image corresponding to each image content in the target image is 10 in length and 5 in width, then the two-dimensional coordinates corresponding to image content 31 in the target image are (-10 -5), the two-dimensional coordinates corresponding to image content 32 are (-20 -5), the two-dimensional coordinates corresponding to image content 33 are (-10 -10), and the two-dimensional coordinates corresponding to image content 34 are (-20, -20). Based on this, the server reversely infers that the size of the object image corresponding to each image content is 10 in length and 5 in width according to the differences between the two-dimensional coordinates in the indication information of each image content. Taking image content 33 as an example, since the x value of its corresponding two-dimensional coordinate is -10, and it is located at the rightmost side of the target image, it can be inferred that the length of the object image corresponding to the image content 33 is 10; since the y value of its corresponding two-dimensional coordinate is -10, and it is located in the second row of the target image, it can be inferred that the length of the object image corresponding to the image content 33 is 5 based on the difference between the y value (-5) of the two-dimensional coordinate corresponding to the image content located in the first row and the y value (-10) of the two-dimensional coordinate corresponding to the image content 33.

[0122] S205, after receiving each attribute recognition result returned by the server, dispatch (return) each attribute recognition result to a corresponding device according to the recorded mapping relationship.

[0123] The embodiment of the present application can establish a local connection between various devices in the same physical space, so that the object images collected by multiple devices in the same physical space at the same time can be aggregated through the established local connection, so that when the object images are aggregated, the network traffic of each device does not need to be consumed, and the network traffic occupied by the object images is saved. Secondly, the mapping relationship between the aggregated object images and each device can be recorded, and the aggregated object images can be synthesized and compressed, so that the target image obtained by the synthesis compression is sent to the server, so as to realize the cutting of the target image on the back-end server side, and the attribute recognition of the objects in the cut object images respectively; this can effectively reduce the uplink request, thereby effectively reducing the number of requests to the server, reducing the occupation of the network traffic by the object image, and further saving the network traffic and reducing the operating pressure of the server. Then, the server can return the identified attribute recognition results to the communicating device, so that the device can perform unified result distribution according to the recorded mapping relationship, so as to realize the dispatching of each attribute recognition result back to the corresponding device; this can make the server not need to communicate with each device, effectively reduce the downlink request of the server, thereby further saving the network traffic and reducing the operating pressure of the server.

[0124] From the above, we can see that the above Figure 2The method embodiment shown is executed by the target device in the physical space. The target device can obtain the object images collected by multiple devices at the current moment through step S201 after being elected as the main device; that is, the target device must be elected as the main device before it can be qualified to execute the above steps S201-S204. The so-called main device refers to a device that is qualified to aggregate the object images and synthesize and compress the aggregated object images.

[0125] In a specific implementation, a local connection may be established between various devices in the physical space, and a master device may be elected from various devices in the physical space through a local mode.

[0126] In this specific implementation, the target device may further perform the following steps s21-s24:

[0127] s21, obtaining network performance information of the target device, where the network performance information includes at least one of the following: network bandwidth, network delay, network jitter, and network packet loss rate.

[0128] s22, obtains network performance information of each device in the physical space except the target device through local connections between each device in the physical space.

[0129] s23, according to the network quality determination rules that have been agreed upon by each device in the physical space, determine the network quality of the corresponding device according to each network performance information obtained. The embodiment of the present application does not limit the network quality determination rules. For example, it can be to score the network quality of the corresponding device according to each information in the network performance information of any device, perform weighted summation on each score, obtain the network quality score of any device, and use the network quality score to indicate the network quality of any device. Alternatively, the rule can be to determine the network quality of the device according to the principle that the network delay is inversely proportional to the network quality. Alternatively, the rule can be to determine the network quality of the device according to the principle that the network packet loss rate is inversely proportional to the network quality, and so on. By using the network quality determination rules that have been agreed upon by each device to determine the network quality, the credibility of the determination result can be guaranteed, thereby ensuring the credibility of the subsequent main device.

[0130] s24, according to the network quality of the target device and the network quality of each other device, select the device with the highest network quality from each device in the physical space as the main device.

[0131] Furthermore, in order to prevent each device from tampering with or fabricating network performance information, the server can uniformly generate network performance information for each device. That is to say, the network performance information of any device is generated and signed by the server. In this case, each time the target device obtains network performance information of a device, it searches for the signature information of the server in the currently obtained network performance information; if the signature information of the server is found, it is determined that the currently obtained network performance information is credible; if the signature information of the server is not found, it is determined that the currently obtained network performance information is suspicious; after determining that each acquired network performance information is credible, the step of determining the network quality of the corresponding device according to each acquired network performance information is triggered.

[0132] Alternatively, the target device may further perform the following steps s31-s34:

[0133] s31, obtaining network performance information of the target device, where the network performance information includes at least one of the following: network bandwidth, network delay, network jitter, and network packet loss rate.

[0134] s32, through the local connection between various devices in the physical space, obtain the network performance information of each device other than the target device in the physical space, and send the network performance information of the target device to other devices.

[0135] s33, based on the obtained network performance information, vote for the main device among the devices in the physical space to obtain the election result; and obtain the election results obtained by other devices based on the network performance information through the local connection between the devices in the physical space.

[0136] s34, according to the election results, determine the master device from the devices in the physical space. Specifically, the device with the most votes can be selected as the master device. By determining the master device by voting on each device, the election process of the master device can be decentralized, thereby improving the credibility of the master device.

[0137] In another specific implementation, the master device election may be performed at the back end of the server, that is, the server elects the master device from various devices in the physical space.

[0138] In this specific implementation, the target device may further perform the following steps s41-s42:

[0139] s41, receives the heartbeat command sent by the server.

[0140] s42, responding to the heartbeat instruction and returning specified data to the server, so that the server determines the current network communication level of the target device according to the specified data returned by the target device.

[0141] It should be understood that in addition to sending a heartbeat command to the target device, the server will also send a heartbeat command to each other device in the physical space, so that each other device returns the specified data, thereby determining the current network communication level of each other device based on the specified data returned by each other device. Then, the server selects a master device from each device in the physical space based on the current network communication level of the target device and the current network communication level of each other device in the physical space except the target device; specifically, the device with the highest current network communication level can be selected as the master device.

[0142] Optionally, after selecting the master device, the server is also used to determine the level type of the current network communication level of the currently selected master device; if the level type is a fixed level type, the master device selection strategy is determined to be a strategy of continuously keeping the master device unchanged; if the level type is a non-fixed level type, the master device selection strategy is determined to be a strategy of reselecting the master device every preset time period. In this way, the flexibility of master device selection can be improved. Among them, the server can determine the level type of the current network communication level by comparing whether the historical network communication level of the master device is consistent with the current network communication level; if they are consistent, the level type is determined to be a fixed level type; if they are inconsistent, the level type is determined to be a non-fixed level type. Alternatively, the level type of the current network communication level is determined based on the current network volatility of the master device; if the current network volatility is greater than a threshold, the level type is determined to be a fixed level type; otherwise, the level type is determined to be a non-fixed level type.

[0143] In actual application, the object recognition methods mentioned above can be applied to different application scenarios according to business needs; for example, face-swiping payment scenarios based on face recognition, sign-in scenarios based on gesture recognition, etc. The so-called face recognition refers to the technology of exchanging facial multimedia information (such as images containing faces) for identity information corresponding to faces; and the face-swiping payment scenario based on face recognition refers to: collecting images containing user faces for identity recognition, and automatically deducting corresponding electronic resources from the resource account associated with the recognized identity information according to the items purchased by the user; the payment scenario can be realized without the user having to operate his own mobile terminal (such as a mobile phone or tablet).

[0144] The following uses the object recognition method in a face-swiping payment scenario based on face recognition as an example to illustrate the specific application process of the object recognition method.

[0145] See also Figure 4As shown: Assume that there are three face-scanning payment devices in the store, namely face-scanning payment device A, face-scanning payment device B and face-scanning payment device C; local connections can be established between each face-scanning payment device, and a network connection can be established between each face-scanning payment device and the server. For further information, see Figure 5 As shown: these three face-scanning payment devices can perform local connection detection with each other to determine the network quality of each face-scanning payment device, so as to elect and confirm face-scanning payment device C as the main device. When face-scanning payment device A and face-scanning payment device B receive face-scanning payment requests from different users, they can send the corresponding face-scanning payment requests to face-scanning payment device C, so that face-scanning payment device C associates and records the sequence number (seq) of the face image and the corresponding device identifier (SN) in each face-scanning payment request, and merges (i.e., synthesizes and compresses) each face image, and sends the merged image to the server, which splits and recognizes the image and returns the recognition results of each face. After receiving the face recognition results returned by the server, face-scanning payment device C can dispatch each face recognition result to face-scanning payment device A and face-scanning payment device B according to the mapping relationship between the recorded sequence number of the face image and the corresponding device identifier, so that the corresponding face-scanning payment device performs subsequent payment processing based on the face recognition result.

[0146] Among them, each of the face-scanning payment devices mentioned above can be an IoT (Internet of Things) face terminal device; for the sake of convenience, the face-scanning payment device will be referred to as a device in the following. The core of each device runs a face recognition APP (application, client), and the core of the APP includes a face collection module, a face recognition LOADING module, a request merging and request scheduling module, an identity confirmation module, a network module, a local communication module, etc., as well as a result page module. Among them, the working content of each module is as follows:

[0147] ① Network module, whose core is responsible for connecting the device with the back-end server through network capabilities to achieve data uplink and download. In addition, the server can regularly send heartbeat instructions to the network module of each device for network detection; correspondingly, the network module of each device can regularly receive the heartbeat instructions sent by the server, respond to the heartbeat instructions, and return the specified data to the server, so that the server can select the device with the best network according to the packet return delay of each device as the main device for subsequent merging requests and distribution requests.

[0148] ② Local communication module, which is mainly responsible for establishing local connections (such as Bluetooth or wired USB, etc.) between devices in offline stores, so that each device can elect a master device based on the local connection; and after the master device is elected, the local communication module is responsible for supporting the communication between devices and the master device through the local connection, confirming the current status of the master device (such as working status, shutdown status, loading status, etc.) and sending requests with collected facial images. The master device can summarize the requests of other devices through the local communication module, and merge and dispatch the facial images in each request. Optionally, each device in the offline store can include a network detection module, which can continuously detect and evaluate the network communication level of the current device. The assessed network communication level is not a fixed level and can be set to elect the master device on an hourly basis. After a device is elected as the master device, it will be the master device within the time window (i.e. within one hour), that is, the relevant requests of other devices will be merged by the device, and the server will recognize the facial recognition results corresponding to each facial image before distribution.

[0149] ③ Face acquisition module, which is mainly responsible for calling the 3D camera component to collect the user's streaming media data at the current moment. After the face acquisition module obtains the streaming media data, the device can optimize the streaming media data. The optimal face image is selected by comprehensive evaluation of coefficient indicators such as face size, face angle, image contrast, image brightness and clarity. After the face image is collected, the face image and the device's SN are transmitted to the main device through the local communication module, so that the main device records the mapping relationship between the face image and the device SN, so that after the back-end server returns the face recognition result of the face image, the face recognition result is dispatched back to the device of the SN according to the recorded mapping relationship.

[0150] ④ Request merging and request scheduling module, which is mainly responsible for merging face images from other devices and sending network requests to the server together, so that the server can perform face recognition processing. Merging face images means splicing and marking each face image, using seq (sequence number) to identify the order of each face image in the current merged image, and using SN to indicate which device the merged image comes from. After the background processing is completed, the server can return the identity information corresponding to each recognized face image to the device indicated by the SN carried by the merged image.

[0151] ⑤ Face recognition LOADING module, which is mainly responsible for controlling the relevant devices to enter the LOADING state (loading state) after the device transmits the face image and the device's SN to the main device through the local communication module, in order to wait for the return of the final face recognition result. It is understandable that the device in the loading state can refuse to respond to other operations.

[0152] ⑥ Identity confirmation module, which is mainly responsible for displaying the user's account information through the device according to the face recognition result after obtaining the face recognition result corresponding to the corresponding face image, and deducting the corresponding electronic resources in the account after the user confirms the account information.

[0153] ⑦Result page module, which is mainly responsible for displaying the interface of the user's current payment status, including success, failure, or uncertainty.

[0154] The server core mentioned above provides backend services such as cloud face recognition service, basic account service, basic payment service and strategy election service. The work content of each service is as follows:

[0155] ① Strategy election service, which is mainly used to elect the main device at the back end. Specifically, the current network quality level of each device can be determined according to the network heartbeat of the device, so as to elect the device with the current high network quality level as the main device of the store, so that other devices in the subsequent store can send data such as face images to the main device through local connection for unified processing. It should be noted that the strategy mentioned here is to elect the main device at the back end, but it can be seen from the above that the embodiment of the present application supports the election of the main device through the local mode; both strategies can be satisfied, but the strategy of electing the main device at the back end can be preferred. The reason is that the server at the back end can perform more refined management and analysis. For example, the network quality level of the device can be classified according to the network performance of different devices in different time periods. After a long period of classification, if it is determined that the network quality level of the currently selected main device is a non-fixed level, the election strategy can be adjusted more proactively, and the local election has greater volatility.

[0156] ② Cloud-based face recognition service, which is mainly used to receive images uploaded from devices in stores, and then cut and extract features from the currently uploaded images. Cutting means cutting the currently uploaded image, and after cutting, each face image cut out can be sent to the recognition module. During recognition, the recognition module can extract facial features, and compare the extracted facial features with the facial features in the feature database to find the facial features with the highest scores, so as to add the identity information corresponding to the facial features with the highest scores to the facial recognition results corresponding to the corresponding facial images. Optionally, in the payment scenario, relevant information such as the user's account number or payment code indicated by the identity information can also be added to the facial recognition results.

[0157] ③ Basic account service, which is mainly used to obtain the account and other information of the user indicated by the identified identity information after successful face recognition.

[0158] ④Basic payment services, which are mainly used to deduct corresponding electronic resources from the user's account.

[0159] Based on the above description, it can be known that the embodiment of the present application, in combination with the case of multiple devices in the store, selects the high network quality as the main device, and the main device aggregates the facial image information of multiple devices in the store that are simultaneously performing face payment through local communication, performs facial image synthesis compression, and identifies the request source of each facial image. Subsequently, the merged image is cut and face recognition is performed at the back end, and the recognized user information is returned to the main device end, and the main device end performs unified result distribution, that is, returns the face recognition result to other devices. By merging and compressing facial images, uplink requests can be effectively reduced, the image resources occupying network traffic can be reduced, and network traffic can be effectively saved.

[0160] Based on the description of the above-mentioned object recognition method embodiment, the present application embodiment further discloses an object recognition device; the object recognition device may be a computer program (including one or more instructions) running in a computer device, and the object recognition device may execute Figure 2 The steps in the method flow are shown. Figure 6 , the object recognition device can run the following units:

[0161] The acquisition unit 601 is used to acquire the object images respectively collected by multiple devices in the physical space at the current moment through the local connection between the devices in the same physical space; wherein the local connection refers to a communication connection that does not consume network traffic;

[0162] The processing unit 602 is used to record the mapping relationship between each acquired object image and each device in the plurality of devices; and perform synthesis compression processing on each object image to obtain a target image;

[0163] The communication unit 603 is used to send the target image to the server, so that the server cuts the target image, and respectively identifies the attributes of the objects in each of the cut object images, and returns the identification results of each attribute.

[0164] The communication unit 603 is further configured to dispatch each attribute recognition result to a corresponding device according to the recorded mapping relationship after receiving each attribute recognition result returned by the server.

[0165] In one implementation, each acquired object image includes an object image acquired by the target device at the current moment; accordingly, the processing unit 602 may be specifically configured to:

[0166] At the current moment, an image of an object located within a shooting range is collected to obtain streaming media data of the object; the streaming media data includes a plurality of images, and each image includes the object;

[0167] Performing image quality evaluation on each image in the streaming media data respectively to obtain an image quality score of each image;

[0168] Based on the image quality scores of the respective images, an image is selected from the streaming media data as an object image collected by the target device at the current moment.

[0169] In another implementation manner, when the processing unit 602 is used to respectively evaluate the image quality of each image in the streaming media data to obtain the image quality score of each image, it can be specifically used to:

[0170] Obtain multiple preset evaluation indicators and the evaluation rules corresponding to each evaluation indicator;

[0171] Under each evaluation index, quality evaluation is performed on each image in the streaming media data according to corresponding evaluation rules to obtain a quality score of each image under each evaluation index;

[0172] Dynamically determine the indicator weight value corresponding to each evaluation indicator according to the quality score of each image under each evaluation indicator;

[0173] For any image in the streaming media data, the indicator weight values ​​corresponding to the various evaluation indicators are used to perform weighted summation on the quality score of the any image under each evaluation indicator to obtain the image quality score of the any image.

[0174] In another embodiment, the plurality of evaluation indicators include at least two of the following: an object size indicator, an object angle indicator, an image contrast indicator, an image brightness indicator, and an image clarity indicator;

[0175] The evaluation rule corresponding to the object size indicator is used to indicate: performing quality evaluation on the corresponding object image according to the size of the object in the image;

[0176] The evaluation rule corresponding to the object angle index is used to indicate: performing quality evaluation on the image according to the difference between the angle of the object in the image and the reference angle;

[0177] The evaluation rule corresponding to the image contrast index is used to indicate: extracting the image region where the object is located from the image, and performing quality evaluation on the image according to the contrast of the extracted image region;

[0178] The evaluation rule corresponding to the image brightness index is used to indicate: extracting the image region where the object is located from the image, and performing quality evaluation on the image according to the brightness of the extracted image region;

[0179] The evaluation rule corresponding to the image clarity index is used to indicate: extracting an image region where an object is located from an image, and performing quality evaluation on the image according to the clarity of the extracted image region.

[0180] In another implementation manner, when the processing unit 602 is used to dynamically determine the indicator weight value corresponding to each evaluation indicator according to the quality score of each image under each evaluation indicator, it can be specifically used to:

[0181] Traversing the multiple evaluation indicators, and taking the currently traversed evaluation indicator as the current evaluation indicator;

[0182] According to the quality scores of the respective images under the current evaluation index, statistically analyzing the score distribution corresponding to the current evaluation index;

[0183] If the score distribution is used to indicate that the corresponding quality scores are concentratedly distributed, the indicator weight value of the current evaluation indicator is determined as a first value;

[0184] If the score distribution is used to indicate that the corresponding quality scores are sparsely distributed, the indicator weight value of the current evaluation indicator is determined as a second value;

[0185] The first value is smaller than the second value.

[0186] In another implementation manner, when the processing unit 602 is used to perform synthesis compression processing on the respective object images to obtain the target image, it can be specifically used to:

[0187] Acquire a blank image, where the blank image refers to an image without storing image content;

[0188] The image contents in the respective object images are extracted respectively, and the extracted image contents are sequentially filled into the blank images to obtain the target image.

[0189] In another implementation, the processing unit 602 may also be used to: determine the indication information of each image content according to the arrangement of each image content in the target image; the indication information of any image content is used to indicate: the position of the corresponding image content in the target image, and the size of the object image corresponding to the corresponding image content;

[0190] The communication unit 603 may also be used to: send the indication information of the contents of each image to the server, so that the server performs segmentation processing on the target image according to the indication information of the contents of each image;

[0191] The cutting process includes: extracting corresponding image content from the target image according to the indication information of any image content, and generating an object image by using the extracted image content.

[0192] In another implementation manner, the acquisition unit acquires the object images respectively acquired by the multiple devices at the current moment after the target device is elected as the master device; accordingly, the processing unit 602 may also be used to:

[0193] Acquire network performance information of the target device, the network performance information including at least one of the following: network bandwidth, network delay, network jitter, and network packet loss rate;

[0194] Acquire network performance information of each device other than the target device in the physical space through local connections between the devices in the physical space;

[0195] According to the network quality determination rule that has been agreed upon by each device in the physical space, the network quality of the corresponding device is determined according to each acquired network performance information;

[0196] According to the network quality of the target device and the network quality of each other device, a device with the highest network quality is selected from each device in the physical space as the main device.

[0197] In another implementation manner, the network performance information of any device is generated and signed by the server; accordingly, the processing unit 602 may also be used to:

[0198] Whenever network performance information of a device is obtained, the signature information of the server is searched in the currently obtained network performance information;

[0199] If the signature information of the server is found, it is determined that the currently acquired network performance information is credible information; if the signature information of the server is not found, it is determined that the currently acquired network performance information is suspicious information;

[0200] After determining that each acquired network performance information is credible information, the step of triggering the execution of the network quality determination rule that has been agreed upon by each device in the physical space, and determining the network quality of the corresponding device according to each acquired network performance information.

[0201] In another implementation manner, the acquisition unit acquires the object images respectively acquired by the multiple devices at the current moment after the target device is elected as the master device; accordingly, the processing unit 602 may also be used to:

[0202] Acquire network performance information of the target device, the network performance information including at least one of the following: network bandwidth, network delay, network jitter, and network packet loss rate;

[0203] Acquiring network performance information of each device other than the target device in the physical space through local connections between the devices in the physical space, and sending the network performance information of the target device to each of the other devices;

[0204] According to the obtained network performance information, voting is performed among the devices in the physical space to elect a master device to obtain an election result; and through the local connection between the devices in the physical space, the election results obtained by other devices based on the network performance information are obtained;

[0205] According to each election result, a master device is determined from each device in the physical space.

[0206] In another implementation manner, the acquisition unit acquires the object images respectively acquired by the multiple devices at the current moment after the target device is elected as the master device; accordingly, the processing unit 602 may also be used to:

[0207] Receiving a heartbeat instruction sent by the server;

[0208] Returning specified data to the server in response to the heartbeat instruction, so that the server determines the current network communication level of the target device according to the specified data returned by the target device;

[0209] The server selects a master device from the devices in the physical space according to the current network communication level of the target device and the current network communication level of other devices in the physical space except the target device.

[0210] In another embodiment, after selecting the master device, the server is further used to determine the level type of the current network communication level of the currently selected master device;

[0211] If the level type is a fixed level type, the master device selection strategy is determined to be a strategy of continuously keeping the master device unchanged; if the level type is a non-fixed level type, the master device selection strategy is determined to be a strategy of reselecting the master device every preset time period.

[0212] According to another embodiment of the present application, Figure 6 The various units in the object recognition device shown can be separately or all combined into one or several other units to constitute, or one (some) of the units can also be split into multiple smaller units in function to constitute, which can achieve the same operation without affecting the realization of the technical effects of the embodiments of the present application. The above-mentioned units are divided based on logical functions. In practical applications, the function of one unit can also be implemented by multiple units, or the functions of multiple units can be implemented by one unit. In other embodiments of the present application, other units can also be included based on the object recognition device. In practical applications, these functions can also be implemented with the assistance of other units, and can be implemented by the collaboration of multiple units.

[0213] According to another embodiment of the present application, the program can be executed by running on a general computing device such as a computer including a central processing unit (CPU), a random access memory medium (RAM), a read-only memory medium (ROM), and other processing elements and storage elements. Figure 2 A computer program (including one or more instructions) for each step involved in the corresponding method shown in , to construct Figure 6 The object recognition device shown in the embodiment of the present application is used to implement the object recognition method of the embodiment of the present application. The computer program can be recorded on a computer-readable storage medium, for example, and loaded into the above-mentioned computing device through the computer-readable storage medium and run therein.

[0214] The embodiment of the present application can establish a local connection between various devices in the same physical space, so that the object images collected by multiple devices in the same physical space at the same time can be aggregated through the established local connection, so that when the object images are aggregated, the network traffic of each device does not need to be consumed, and the network traffic occupied by the object images is saved. Secondly, the mapping relationship between the aggregated object images and each device can be recorded, and the aggregated object images can be synthesized and compressed, so that the target image obtained by the synthesis compression is sent to the server, so as to realize the cutting of the target image on the back-end server side, and the attribute recognition of the objects in the cut object images respectively; this can effectively reduce the uplink request, thereby effectively reducing the number of requests to the server, reducing the occupation of the network traffic by the object image, and further saving the network traffic and reducing the operating pressure of the server. Then, the server can return the identified attribute recognition results to the communicating device, so that the device can perform unified result distribution according to the recorded mapping relationship, so as to realize the dispatching of each attribute recognition result back to the corresponding device; this can make the server not need to communicate with each device, effectively reduce the downlink request of the server, thereby further saving the network traffic and reducing the operating pressure of the server.

[0215] Based on the description of the above method embodiment and the apparatus embodiment, the present application embodiment further provides a computer device, which may be the target device mentioned above. Figure 7 , the computer device at least includes a processor 701, an input interface 702, an output interface 703 and a computer storage medium 704. Among them, the processor 701, the input interface 702, the output interface 703 and the computer storage medium 704 in the computer device can be connected via a bus or other means. The computer storage medium 704 can be stored in the memory of the computer device, the computer storage medium 704 is used to store a computer program, the computer program includes one or more instructions, and the processor 701 is used to execute one or more instructions in the computer program stored in the computer storage medium 704. The processor 701 (or CPU (Central Processing Unit)) is the computing core and control core of the computer device, which is suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions to realize the corresponding method flow or corresponding function.

[0216] In one embodiment, the processor 701 described in the embodiment of the present application can be used to perform a series of object recognition processing, specifically including: obtaining object images respectively collected by multiple devices in the physical space at the current moment through local connections between devices located in the same physical space; wherein the local connection refers to a communication connection that does not consume network traffic; recording the mapping relationship between each acquired object image and each device in the multiple devices; and performing synthetic compression processing on the each object image to obtain a target image; sending the target image to a server, so that the server cuts the target image, and respectively performs attribute recognition on the objects in each cut object image, and returns each attribute recognition result; after receiving each attribute recognition result returned by the server, dispatching each attribute recognition result to the corresponding device according to the recorded mapping relationship, and so on.

[0217] The embodiment of the present application also provides a computer storage medium (Memory), which is a memory device in a computer device for storing computer programs and data. It is understandable that the computer storage medium here can include both built-in storage media in the computer device and, of course, extended storage media supported by the computer device. The computer storage medium provides a storage space, which stores the operating system of the computer device. In addition, a computer program is also stored in the storage space, which includes one or more instructions suitable for being loaded and executed by the processor 701, and these instructions can be one or more program codes. It should be noted that the computer storage medium here can be a high-speed RAM memory, or a non-volatile memory (non-volatile memory), such as at least one disk storage; optionally, it can also be at least one computer storage medium located away from the aforementioned processor.

[0218] In one embodiment, the processor may load and execute one or more instructions stored in the computer storage medium to implement the above-mentioned Figure 2 Corresponding steps in the method embodiment shown; in a specific implementation, one or more instructions in the computer storage medium can be loaded by the processor and execute the following steps:

[0219] Obtaining, through local connections between devices located in the same physical space, object images respectively captured by multiple devices in the physical space at the current moment; wherein the local connection refers to a communication connection that does not consume network traffic;

[0220] Recording the mapping relationship between each acquired object image and each device in the plurality of devices; and performing synthesis compression processing on each object image to obtain a target image;

[0221] Sending the target image to a server, causing the server to cut the target image, and respectively identify the attributes of the objects in each of the cut object images, and returning each attribute recognition result;

[0222] After receiving each attribute recognition result returned by the server, each attribute recognition result is dispatched to a corresponding device according to the recorded mapping relationship.

[0223] In one implementation, each acquired object image includes an object image acquired by the target device at the current moment; accordingly, the one or more instructions may be loaded and specifically executed by the processor:

[0224] At the current moment, an image of an object located within a shooting range is collected to obtain streaming media data of the object; the streaming media data includes a plurality of images, and each image includes the object;

[0225] Performing image quality evaluation on each image in the streaming media data respectively to obtain an image quality score of each image;

[0226] Based on the image quality scores of the respective images, an image is selected from the streaming media data as an object image collected by the target device at the current moment.

[0227] In another implementation, when performing image quality evaluation on each image in the streaming media data to obtain the image quality score of each image, the one or more instructions may be loaded and specifically executed by the processor:

[0228] Obtain multiple preset evaluation indicators and the evaluation rules corresponding to each evaluation indicator;

[0229] Under each evaluation index, quality evaluation is performed on each image in the streaming media data according to corresponding evaluation rules to obtain a quality score of each image under each evaluation index;

[0230] Dynamically determine the indicator weight value corresponding to each evaluation indicator according to the quality score of each image under each evaluation indicator;

[0231] For any image in the streaming media data, the indicator weight values ​​corresponding to the various evaluation indicators are used to perform weighted summation on the quality score of the any image under each evaluation indicator to obtain the image quality score of the any image.

[0232] In another embodiment, the plurality of evaluation indicators include at least two of the following: an object size indicator, an object angle indicator, an image contrast indicator, an image brightness indicator, and an image clarity indicator;

[0233] The evaluation rule corresponding to the object size indicator is used to indicate: performing quality evaluation on the corresponding object image according to the size of the object in the image;

[0234] The evaluation rule corresponding to the object angle index is used to indicate: performing quality evaluation on the image according to the difference between the angle of the object in the image and the reference angle;

[0235] The evaluation rule corresponding to the image contrast index is used to indicate: extracting the image region where the object is located from the image, and performing quality evaluation on the image according to the contrast of the extracted image region;

[0236] The evaluation rule corresponding to the image brightness index is used to indicate: extracting the image region where the object is located from the image, and performing quality evaluation on the image according to the brightness of the extracted image region;

[0237] The evaluation rule corresponding to the image clarity index is used to indicate: extracting an image region where an object is located from an image, and performing quality evaluation on the image according to the clarity of the extracted image region.

[0238] In another implementation, when dynamically determining the indicator weight value corresponding to each evaluation indicator according to the quality score of each image under each evaluation indicator, the one or more instructions may be loaded and specifically executed by the processor:

[0239] Traversing the multiple evaluation indicators, and taking the currently traversed evaluation indicator as the current evaluation indicator;

[0240] According to the quality scores of the respective images under the current evaluation index, statistically analyzing the score distribution corresponding to the current evaluation index;

[0241] If the score distribution is used to indicate that the corresponding quality scores are concentratedly distributed, the indicator weight value of the current evaluation indicator is determined as a first value;

[0242] If the score distribution is used to indicate that the corresponding quality scores are sparsely distributed, the indicator weight value of the current evaluation indicator is determined as a second value;

[0243] The first value is smaller than the second value.

[0244] In another implementation, when the target image is obtained by synthesizing and compressing the respective object images, the one or more instructions may be loaded and specifically executed by the processor:

[0245] Acquire a blank image, where the blank image refers to an image without storing image content;

[0246] The image contents in the respective object images are extracted respectively, and the extracted image contents are sequentially filled into the blank images to obtain the target image.

[0247] In another implementation, the one or more instructions may be loaded and specifically executed by a processor:

[0248] Determine the indication information of each image content according to the arrangement of each image content in the target image; the indication information of any image content is used to indicate: the position of the corresponding image content in the target image, and the size of the object image corresponding to the corresponding image content;

[0249] Sending the indication information of the contents of each image to the server, so that the server performs segmentation processing on the target image according to the indication information of the contents of each image;

[0250] The cutting process includes: extracting corresponding image content from the target image according to the indication information of any image content, and generating an object image by using the extracted image content.

[0251] In another implementation manner, after the target device is selected as the master device, the one or more instructions are loaded and specifically executed by the processor: obtaining the object images respectively collected by the multiple devices at the current moment; accordingly, the one or more instructions can be loaded and specifically executed by the processor:

[0252] Acquire network performance information of the target device, the network performance information including at least one of the following: network bandwidth, network delay, network jitter, and network packet loss rate;

[0253] Acquire network performance information of each device other than the target device in the physical space through local connections between the devices in the physical space;

[0254] According to the network quality determination rule that has been agreed upon by each device in the physical space, the network quality of the corresponding device is determined according to each acquired network performance information;

[0255] According to the network quality of the target device and the network quality of each other device, a device with the highest network quality is selected from each device in the physical space as the main device.

[0256] In another implementation manner, the network performance information of any device is generated and signed by the server; accordingly, the one or more instructions may be loaded and specifically executed by the processor:

[0257] Whenever network performance information of a device is obtained, the signature information of the server is searched in the currently obtained network performance information;

[0258] If the signature information of the server is found, it is determined that the currently acquired network performance information is credible information; if the signature information of the server is not found, it is determined that the currently acquired network performance information is suspicious information;

[0259] After determining that each acquired network performance information is credible information, the step of triggering the execution of the network quality determination rule that has been agreed upon by each device in the physical space, and determining the network quality of the corresponding device according to each acquired network performance information.

[0260] In another implementation manner, after the target device is selected as the master device, the one or more instructions are loaded and specifically executed by the processor: obtaining the object images respectively collected by the multiple devices at the current moment; accordingly, the one or more instructions can be loaded and specifically executed by the processor:

[0261] Acquire network performance information of the target device, the network performance information including at least one of the following: network bandwidth, network delay, network jitter, and network packet loss rate;

[0262] Acquiring network performance information of each device other than the target device in the physical space through local connections between the devices in the physical space, and sending the network performance information of the target device to each of the other devices;

[0263] According to the obtained network performance information, voting is performed among the devices in the physical space to elect a master device to obtain an election result; and through the local connection between the devices in the physical space, the election results obtained by other devices based on the network performance information are obtained;

[0264] According to each election result, a master device is determined from each device in the physical space.

[0265] In another implementation manner, after the target device is selected as the master device, the one or more instructions are loaded and specifically executed by the processor: obtaining the object images respectively collected by the multiple devices at the current moment; accordingly, the one or more instructions can be loaded and specifically executed by the processor:

[0266] Receiving a heartbeat instruction sent by the server;

[0267] Returning specified data to the server in response to the heartbeat instruction, so that the server determines the current network communication level of the target device according to the specified data returned by the target device;

[0268] The server selects a master device from the devices in the physical space according to the current network communication level of the target device and the current network communication level of other devices in the physical space except the target device.

[0269] The embodiment of the present application can establish a local connection between various devices in the same physical space, so that the object images collected by multiple devices in the same physical space at the same time can be aggregated through the established local connection, so that when the object images are aggregated, the network traffic of each device does not need to be consumed, and the network traffic occupied by the object images is saved. Secondly, the mapping relationship between the aggregated object images and each device can be recorded, and the aggregated object images can be synthesized and compressed, so that the target image obtained by the synthesis compression is sent to the server, so as to realize the cutting of the target image on the back-end server side, and the attribute recognition of the objects in the cut object images respectively; this can effectively reduce the uplink request, thereby effectively reducing the number of requests to the server, reducing the occupation of the network traffic by the object image, and further saving the network traffic and reducing the operating pressure of the server. Then, the server can return the identified attribute recognition results to the communicating device, so that the device can perform unified result distribution according to the recorded mapping relationship, so as to realize the dispatching of each attribute recognition result back to the corresponding device; this can make the server not need to communicate with each device, effectively reduce the downlink request of the server, thereby further saving the network traffic and reducing the operating pressure of the server.

[0270] It should be noted that, according to one aspect of the present application, a computer program product or a computer program is also provided, the computer program product or the computer program includes one or more instructions, the one or more instructions are stored in a computer storage medium. The processor of the computer device reads the one or more instructions from the computer storage medium, and the processor executes the one or more instructions, so that the computer device performs the above Figure 2 The method is provided in various optional aspects of the method embodiment shown.

[0271] It should be understood that what is disclosed above is only a preferred embodiment of the present application, and certainly cannot be used to limit the scope of rights of the present application. Therefore, equivalent changes made according to the claims of the present application are still within the scope covered by the present application.

Claims

1. An object recognition method, characterized in that: The method comprises: Obtaining, through local connections between devices located in the same physical space, object images respectively captured by multiple devices in the physical space at the current moment; wherein the local connection refers to a communication connection that does not consume network traffic; Recording the mapping relationship between each acquired object image and each device in the plurality of devices; and performing synthesis compression processing on each object image to obtain a target image; Sending the target image to a server, causing the server to cut the target image, and respectively identify the attributes of the objects in each of the cut object images, and returning each attribute recognition result; After receiving each attribute recognition result returned by the server, each attribute recognition result is dispatched to a corresponding device according to the recorded mapping relationship.

2. The method according to claim 1, characterized in that The local connection includes: a short-range wireless communication connection, or a connection established based on a removable hardware medium; Among them, the short-range wireless communication connection includes: Bluetooth connection, near-field communication connection, or wireless connection; the removable hardware medium includes: a USB flash drive, a mobile hard disk, or a memory card.

3. The method according to claim 1, characterized in that The process of any device collecting an image of an object at the current moment includes: At the current moment, an image of an object located within a shooting range is collected to obtain streaming media data of the object; the streaming media data includes a plurality of images, and each image includes the object; Performing image quality evaluation on each image in the streaming media data respectively to obtain an image quality score of each image; Based on the image quality scores of the respective images, an image is selected from the streaming media data as the object image collected by any one of the devices at the current moment.

4. The method according to claim 3, characterized in that The performing image quality evaluation on each image in the streaming media data to obtain an image quality score of each image includes: Obtain multiple preset evaluation indicators and the evaluation rules corresponding to each evaluation indicator; Under each evaluation index, quality evaluation is performed on each image in the streaming media data according to corresponding evaluation rules to obtain a quality score of each image under each evaluation index; Dynamically determine the indicator weight value corresponding to each evaluation indicator according to the quality score of each image under each evaluation indicator; For any image in the streaming media data, the indicator weight values ​​corresponding to the various evaluation indicators are used to perform weighted summation on the quality score of the any image under each evaluation indicator to obtain the image quality score of the any image.

5. The method according to claim 4, characterized in that The plurality of evaluation indices include at least two of the following: an object size index, an object angle index, an image contrast index, an image brightness index, and an image clarity index; The evaluation rule corresponding to the object size indicator is used to indicate: performing quality evaluation on the corresponding object image according to the size of the object in the image; The evaluation rule corresponding to the object angle index is used to indicate: performing quality evaluation on the image according to the difference between the angle of the object in the image and the reference angle; The evaluation rule corresponding to the image contrast index is used to indicate: extracting the image region where the object is located from the image, and performing quality evaluation on the image according to the contrast of the extracted image region; The evaluation rule corresponding to the image brightness index is used to indicate: extracting the image region where the object is located from the image, and performing quality evaluation on the image according to the brightness of the extracted image region; The evaluation rule corresponding to the image clarity index is used to indicate: extracting an image region where an object is located from an image, and performing quality evaluation on the image according to the clarity of the extracted image region.

6. The method according to claim 4, characterized in that The step of dynamically determining the indicator weight value corresponding to each evaluation indicator according to the quality score of each image under each evaluation indicator comprises: Traversing the multiple evaluation indicators, and taking the currently traversed evaluation indicator as the current evaluation indicator; According to the quality scores of the respective images under the current evaluation index, statistically analyzing the score distribution corresponding to the current evaluation index; If the score distribution is used to indicate that the corresponding quality scores are concentratedly distributed, the indicator weight value of the current evaluation indicator is determined as a first value; If the score distribution is used to indicate that the corresponding quality scores are sparsely distributed, the indicator weight value of the current evaluation indicator is determined as a second value; The first value is smaller than the second value.

7. The method according to claim 1, characterized in that The step of performing synthesis compression processing on each of the object images to obtain a target image comprises: Acquire a blank image, where the blank image refers to an image without storing image content; The image contents in the respective object images are extracted respectively, and the extracted image contents are sequentially filled into the blank images to obtain the target image.

8. The method according to claim 7, characterized in that The method further comprises: Determine the indication information of each image content according to the arrangement of each image content in the target image; the indication information of any image content is used to indicate: the position of the corresponding image content in the target image, and the size of the object image corresponding to the corresponding image content; Sending the indication information of the contents of each image to the server, so that the server performs segmentation processing on the target image according to the indication information of the contents of each image; The cutting process includes: extracting corresponding image content from the target image according to the indication information of any image content, and generating an object image by using the extracted image content.

9. The method according to claim 1, characterized in that The method is performed by a target device in the physical space, and the target device obtains the object images respectively collected by the multiple devices at the current moment after being elected as the main device; The method further comprises: Acquire network performance information of the target device, the network performance information including at least one of the following: network bandwidth, network delay, network jitter, and network packet loss rate; Acquire network performance information of each device other than the target device in the physical space through local connections between the devices in the physical space; According to the network quality determination rule that has been agreed upon by each device in the physical space, the network quality of the corresponding device is determined according to each acquired network performance information; According to the network quality of the target device and the network quality of each other device, a device with the highest network quality is selected from each device in the physical space as the main device.

10. The method according to claim 9, characterized in that The network performance information of any device is generated and signed by the server; the method further includes: Whenever network performance information of a device is obtained, the signature information of the server is searched in the currently obtained network performance information; If the signature information of the server is found, it is determined that the currently acquired network performance information is credible information; if the signature information of the server is not found, it is determined that the currently acquired network performance information is suspicious information; After determining that each acquired network performance information is credible information, the step of triggering the execution of the network quality determination rule that has been agreed upon by each device in the physical space, and determining the network quality of the corresponding device according to each acquired network performance information.

11. The method according to claim 1, characterized in that The method is performed by a target device in the physical space, and the target device obtains the object images respectively collected by the multiple devices at the current moment after being elected as the main device; The method further comprises: Acquire network performance information of the target device, the network performance information including at least one of the following: network bandwidth, network delay, network jitter, and network packet loss rate; Acquiring network performance information of each device other than the target device in the physical space through local connections between the devices in the physical space, and sending the network performance information of the target device to each of the other devices; According to the obtained network performance information, voting is performed among the devices in the physical space to elect a master device to obtain an election result; and through the local connection between the devices in the physical space, the election results obtained by other devices based on the network performance information are obtained; According to each election result, a master device is determined from each device in the physical space.

12. The method according to claim 1, characterized in that The method is performed by a target device in the physical space, and the target device obtains the object images respectively collected by the multiple devices at the current moment after being elected as the main device; The method further comprises: Receiving a heartbeat instruction sent by the server; Returning specified data to the server in response to the heartbeat instruction, so that the server determines the current network communication level of the target device according to the specified data returned by the target device; The server selects a master device from the devices in the physical space according to the current network communication level of the target device and the current network communication level of other devices in the physical space except the target device.

13. The method according to claim 12, characterized in that After selecting the master device, the server is further used to determine the level type of the current network communication level of the currently selected master device; If the level type is a fixed level type, the master device selection strategy is determined to be a strategy of continuously keeping the master device unchanged; if the level type is a non-fixed level type, the master device selection strategy is determined to be a strategy of reselecting the master device every preset time period.

14. An object recognition device, characterized in that: include: An acquisition unit, configured to acquire, through a local connection between devices located in the same physical space, the object images respectively acquired by multiple devices in the physical space at the current moment; wherein the local connection refers to a communication connection that does not consume network traffic; A processing unit, used to record the mapping relationship between each acquired object image and each device in the plurality of devices; and perform synthesis compression processing on each object image to obtain a target image; A communication unit, used to send the target image to a server, so that the server cuts the target image, and respectively identifies the attributes of the objects in each of the cut object images, and returns each attribute identification result; The communication unit is further configured to dispatch each attribute recognition result to a corresponding device according to the recorded mapping relationship after receiving each attribute recognition result returned by the server.

15. A computer device comprising an input interface and an output interface, characterized in that: Also includes: processor and computer storage media; The processor is suitable for implementing one or more instructions, the computer storage medium stores one or more instructions, and the one or more instructions are suitable for being loaded by the processor and executing the object recognition method according to any one of claims 1-13.

16. A computer storage medium, characterized in that: The computer storage medium stores one or more instructions, and the one or more instructions are suitable for being loaded by a processor and executing the object recognition method according to any one of claims 1 to 13.

17. A computer program product, characterized in that The computer program product includes one or more instructions; when the one or more instructions in the computer program are executed by a processor, the object recognition method according to any one of claims 1 to 13 is implemented.

Images