Target recognition method and master-slave multi-terminal camera system

Abstract

The present application relates to a target recognition method and a master-slave multi-end camera system, wherein the target recognition method obtains an original image to be recognized, detects the target area where the target is located in the original image, and obtains the original data of the target area, wherein the original image contains at least one target area, and sends the original data to a slave camera with preset target recognition conditions in the master-slave multi-end camera system, and the slave camera performs target recognition processing based on the original data to obtain the target feature value of the original data, and processes the target feature value to obtain the target recognition result in each target area in the original image. It realizes the integration of multiple camera resources based on the master-slave multi-end camera system, thereby improving the chip processing capacity of the intelligent network camera, and then increasing the number of targets for the intelligent network camera to perform target recognition, thereby solving the problem of limited scenes for target recognition by the current intelligent network camera.

Description

Technical Field

[0001] The present application relates to the field of image processing, and in particular to a target recognition method and a master-slave multi-terminal camera system. Background Art

[0002] With the development of science and technology, smart network cameras that integrate image acquisition and intelligent detection functions have been widely used in the field of target recognition. However, in the process of using deep neural networks to recognize target images, due to the large number of parameters of deep neural networks, the model training and image recognition processes require a lot of calculations, so the requirements for AI chips are high. Current AI chips have limitations on the number of targets to be identified in the detection picture. For example, it can only support the recognition of 20 targets in the picture. Therefore, in the face of a scene with a large number of people, it is difficult for a smart network camera to meet the needs of identifying all targets in the scene.

[0003] There is currently no effective solution to the problem of limited recognition scenes when intelligent network cameras perform target recognition in related technologies. Summary of the invention

[0004] In this embodiment, a target recognition method and a master-slave multi-terminal camera system are provided to solve the problem of limited recognition scenes when intelligent network cameras perform target recognition in the related art.

[0005] In a first aspect, a target recognition method is provided in this embodiment, which is used for a master camera in a master-slave multi-terminal camera system, and the method includes:

[0006] Acquire an original image to be identified, and detect a target area where the target is located in the original image to obtain original data of the target area; wherein the original image contains at least one target area;

[0007] The raw data is sent to a slave camera in the master-slave multi-terminal camera system that has a preset target recognition condition, and the slave camera performs target recognition processing based on the raw data to obtain a target feature value of the raw data;

[0008] The target feature values ​​are processed to obtain target recognition results in each target area in the original image.

[0009] In some of the embodiments, detecting the target area where the target is located in the original image to obtain the original data of the target area includes:

[0010] According to the preset detection requirements, select the target area detection algorithm from the pre-built detection algorithm library;

[0011] The target region detection algorithm that has been selected is used to detect the target region where the target is located in the original image to obtain original data of the target region.

[0012] In some embodiments, the sending of the raw data to a slave camera having a preset target recognition condition in the master-slave multi-terminal camera system, and the slave camera performing target recognition processing based on the raw data, includes:

[0013] The raw data is filtered according to a preset filter template, and the filtered raw data is sent to a slave camera with preset target recognition conditions in the master-slave multi-terminal camera system, and the slave camera performs target recognition processing based on the raw data.

[0014] In some embodiments, the sending of the raw data to a slave camera having a preset target recognition condition in the master-slave multi-terminal camera system, and the slave camera performing target recognition processing based on the raw data, further includes:

[0015] Allocating raw data to the slave cameras according to the number of target areas in the original image and the processing priority of the slave cameras having preset target recognition conditions in the master-slave multi-terminal camera system;

[0016] According to the allocation result, the original data is sent to the allocated slave camera, and the allocated slave camera performs target recognition processing.

[0017] In some of the embodiments, in the master-slave multi-terminal camera system, the processing priority of the slave cameras that meet the preset target recognition conditions is determined by the ranking of the resource utilization rates of the slave cameras that meet the preset target recognition conditions in the master-slave multi-terminal camera system.

[0018] In some embodiments, the sending of the raw data to a slave camera having a preset target recognition condition in the master-slave multi-terminal camera system, and the slave camera performing target recognition processing based on the raw data to obtain a target feature value of the raw data, further includes:

[0019] Allocating raw data to the slave cameras according to the categories of the target areas and the feature extraction capabilities of the slave cameras with preset target recognition conditions for the target areas of different categories;

[0020] According to the allocation result, the original data is sent to the allocated slave camera, and the allocated slave camera performs target recognition processing to obtain the target feature value of the original data.

[0021] In some of the embodiments, the processing of the target feature values ​​to obtain target recognition results in each target area in the original image includes:

[0022] Performing target feature modeling on the multiple target feature values ​​of each target area to obtain a combined feature value of each target area in the original image;

[0023] The combined feature values ​​of each target area are compared with reference feature values ​​in a preset reference library to obtain target recognition results in each target area in the original image.

[0024] In some embodiments, the comparing the combined feature values ​​of each target area with reference feature values ​​in a preset reference library to obtain target recognition results in each target area in the original image includes:

[0025] The combined feature values ​​of each target area are compared with the reference feature values ​​in a preset reference library to obtain the similarity between each target area and the reference area of ​​the preset reference library, and the target recognition results in each target area in the original image are obtained based on the similarity, a preset similarity threshold and the sensitivity of the target area.

[0026] In a second aspect, a target recognition method is provided in this embodiment, which is used for a slave camera of a master-slave multi-terminal camera system, and the method includes:

[0027] Obtaining raw data of a target area obtained after a master camera in the master-slave multi-terminal camera system detects a target area where a target is located in an original image; wherein the original image contains at least one target area;

[0028] Performing target recognition processing on the original data to obtain target feature values ​​of the original data;

[0029] The target feature value is sent to the main camera, and the main camera processes the target feature value to obtain the target recognition result in each target area in the original image.

[0030] In some embodiments, the performing target recognition processing on the original data to obtain the target feature value of the original data includes:

[0031] The original data is subjected to target tracking, target optimization, and recognition processing of target angle and target attribute to obtain target feature values ​​of the original data.

[0032] In a third aspect, a master-slave multi-terminal camera system is provided in this embodiment, including a master camera and a slave camera; wherein:

[0033] The main camera is used to perform the target recognition method of the first aspect;

[0034] The slave camera is used to perform the target recognition method of the second aspect;

[0035] The master camera and the slave camera are communicatively connected.

[0036] In some of the embodiments, the master camera is further configured to monitor the status of the slave cameras through a heartbeat mechanism, and determine a slave camera that meets a preset target recognition condition according to the status of the slave cameras and resource management information.

[0037] In some embodiments, the slave camera is further configured to send resource change information to the master camera when its own resources are changed, so that the master camera updates the resource management information of the slave camera.

[0038] Compared with the related art, the target recognition method for the master camera of the master-slave multi-terminal camera system described above obtains the original image to be recognized, detects the target area where the target is located in the original image, and obtains the original data of the target area, wherein the original image contains at least one target area, and sends the original data to the slave camera with preset target recognition conditions in the master-slave multi-terminal camera system, and the slave camera performs target recognition processing based on the original data to obtain the target feature value of the original data, and processes the target feature value to obtain the target recognition result in each target area in the original image. It realizes the integration of multiple camera resources based on the master-slave multi-terminal camera system, thereby improving the chip processing capability of the intelligent network camera, and then increasing the number of targets for target recognition by the intelligent network camera, thereby solving the problem of limited scenes for target recognition by the current intelligent network camera.

[0039] Details of one or more embodiments of the present application are set forth in the following drawings and description to make other features, objects, and advantages of the present application more readily apparent. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0041] Figure 1 It is a hardware structure block diagram of a terminal of a target recognition method of the related art;

[0042] Figure 2 is a flow chart of a target recognition method for a master camera of a master-slave multi-terminal camera system according to this embodiment;

[0043] Figure 3is a closed-loop flow chart of data transmission in the target recognition method of this embodiment;

[0044] Figure 4 is a business flow chart of the master-slave multi-terminal camera system of this embodiment;

[0045] Figure 5 is a flow chart of a target recognition method for a slave camera of a master-slave multi-terminal camera system according to this embodiment;

[0046] Figure 6 is a flow chart of the target recognition method of the preferred embodiment;

[0047] Figure 7 is a schematic diagram of the structure of the master-slave multi-terminal camera system of this embodiment;

[0048] Figure 8 is a flowchart of the master camera service processing of the master-slave multi-terminal camera system of this embodiment;

[0049] Fig. 9 This is a flowchart of slave camera business processing in the master-slave multi-terminal camera system of this embodiment. DETAILED DESCRIPTION

[0050] In order to more clearly understand the purpose, technical solutions and advantages of the present application, the present application is described and illustrated below in conjunction with the accompanying drawings and embodiments.

[0051] Unless otherwise defined, the technical terms or scientific terms involved in this application shall have the general meaning understood by people with ordinary skills in the technical field to which this application belongs. The words "one", "a", "a", "the", "these" and the like in this application do not represent quantitative restrictions, and they can be singular or plural. The terms "include", "comprise", "have" and any variants thereof involved in this application are intended to cover non-exclusive inclusions; for example, a process, method and system, product or device comprising a series of steps or modules (units) is not limited to the listed steps or modules (units), but may include unlisted steps or modules (units), or may include other steps or modules (units) inherent to these processes, methods, products or devices. The words "connect", "connected", "coupled" and the like involved in this application are not limited to physical or mechanical connections, but may include electrical connections, whether directly or indirectly. The "multiple" involved in this application refers to two or more. "And / or" describes the association relationship of associated objects, indicating that there can be three relationships. For example, "A and / or B" can mean: A exists alone, A and B exist at the same time, and B exists alone. Usually, the character " / " indicates that the objects associated with each other are in an "or" relationship. The terms "first", "second", "third", etc. involved in this application are only used to distinguish similar objects and do not represent a specific ordering of the objects.

[0052] The method embodiment provided in this embodiment can be executed in a terminal, a computer or a similar computing device. For example, running on a terminal, Figure 1 FIG. 1 is a hardware structure diagram of a terminal of the target recognition method of this embodiment. Figure 1 As shown, the terminal may include one or more ( Figure 1 Only one is shown in the figure) processor 102 and memory 104 for storing data, wherein processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA. The above terminal may also include a transmission device 106 and an input and output device 108 for communication functions. It can be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the above terminal. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations shown.

[0053] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to the target recognition method in this embodiment. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, to implement the above method. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to the terminal via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0054] The transmission device 106 is used to receive or send data via a network. The above network includes a wireless network provided by the communication provider of the terminal. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, referred to as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (Radio Frequency, referred to as RF) module, which is used to communicate with the Internet wirelessly.

[0055] In this embodiment, a target recognition method is provided for a master camera in a master-slave multi-terminal camera system. Figure 2 is a flow chart of the target recognition method of this embodiment, such as Figure 2 As shown, the process includes the following steps:

[0056] Step S210, obtaining an original image to be identified, detecting a target area where a target is located in the original image, and obtaining original data of the target area; wherein the original image contains at least one target area.

[0057] Specifically, the original image to be identified can be an image collected from a scene with a large number of people, containing more target areas. For example, the original image collected from scenes such as a self-study classroom, a conference site or an exhibition hall. The original image to be identified can also be an image collected from a monitoring scene containing more animals, such as an original image collected from a farm or the wild. The target in the original image can be determined according to the actual application scenario, and can specifically be a human face, a human body, a specific posture of a human body, a specific animal or other objects to be identified. In the process of target recognition of the original image, the characteristic value of the target area can be obtained by sequentially performing the detection of the target area, the tracking of the target area, the optimization of the target image, the identification of the target angle and the target attribute on the original image, and the identification of the target is completed based on the characteristic value. Among them, the target attribute can be specific information set corresponding to the target. For example, when the application scenario is face recognition, the target attribute can specifically include gender, glasses, age, expression, etc., and when the application scenario is animal detection, the target attribute can be animal type and size, etc.

[0058] The AI ​​chip is integrated in the smart network camera and can be used to identify the collected images. For example, the smart network camera can support functions such as image data collection, image feature value extraction, and image recognition. However, due to the limited resources occupied by a single smart network camera, the computing power is difficult to meet the target recognition in a multi-target scenario. The master-slave multi-end camera system can be specifically composed of several smart network cameras. Among them, one of the smart network cameras can be pre-selected as the master camera of the master-slave multi-end camera system, and the other smart network cameras can be used as slave cameras of the master-slave multi-end camera system. It should be noted that the master camera and the slave camera are relative concepts, and the resources of the master camera can also become part of the slave camera, thus forming a closed-loop resource pool. By allocating part of the target recognition business of the original image collected by the master camera to one or more slave cameras for processing, the occupation of the computer resources of the master camera can be reduced, thereby improving the performance of target recognition.

[0059] Specifically, after the main camera completes the acquisition of the original image to be identified, that is, after the original image to be identified is obtained, the target area where the target in the original image is located can be detected in the main camera. Specifically, an image recognition model with relatively less computing resources occupied during the detection process can be selected to realize the detection of the target area and the statistics of the number of targets in the original image. For example, in terms of business, different image recognition models can be graded according to the computing resources occupied, the fineness of detection and the computing power, and a large target detection model, a medium target detection model and a small target detection model can be obtained. Among them, the large target detection model requires relatively the most computing resources and has the highest detection fineness. The medium target detection model is second. The small target detection model requires relatively the least computing resources and has relatively low detection fineness. When the main camera performs target area detection processing on the original image, a small target detection model that meets the business needs can be selected for processing, so as to achieve the detection of the target area while minimizing the occupation of the main camera's computing resources. It should be noted that the target detection large model, target detection medium model and target detection small model in this embodiment are algorithm categories set to distinguish the detection precision of the algorithm, the occupied computing resources, etc. The specific algorithm adopted can be set accordingly according to the target recognition requirements in the actual application scenario and the camera hardware environment, and is not specifically limited here.

[0060] Alternatively, the original data may be a representation of key information of the original image, for example, the original data may include the size of the original image, the location information of the target area in the original image, grayscale feature values, and the like.

[0061] Step S220, the original data is sent to a slave camera with preset target recognition conditions in the master-slave multi-terminal camera system, and the slave camera performs target recognition processing based on the original data to obtain the target feature value of the original data.

[0062] In a master-slave multi-end camera system, the master camera can determine the slave camera that meets the target recognition conditions by querying the resource status and business usage of the slave camera. For example, the master camera can monitor the status of the slave camera through the heartbeat mechanism to determine the slave camera that is in an active state. The master camera allocates the raw data to be processed to the slave camera based on the resource status and business usage rate of each active slave camera, as well as the computing resources required to process the raw data. The slave camera can specifically use a target detection model with relatively more computing resources and higher detection precision to track the target area, optimize the target image, and identify the target angle and target attributes of the raw data. Among them, the extracted target feature value can specifically be any image feature in the field of image recognition, such as HOG features, LBP features, and Haar features.

[0063] Before sending the raw data to the slave camera, the master camera can also filter the raw data based on image features such as grayscale feature values, and send the filtered raw data to the slave camera, thereby reducing the processing pressure of the slave camera. Specifically, the raw data can be filtered using a filter template corresponding to the image features to obtain the filtered raw data. In addition, after filtering the raw data, a unique identifier can be generated for the raw data based on the size, location information, grayscale feature values ​​and other features of the raw data. Preferably, the unique identifier can be a uuid. The unique identifier is associated with the raw data and sent to the slave camera, thereby realizing a closed-loop operation of target recognition in the master-slave multi-end camera system. Figure 3 This is a closed-loop process for target identification and data transmission in this embodiment. Figure 3 As shown in the figure, after obtaining the raw data of the original image, the master camera uses the relevant filter to filter the image frame size, the location information of the target area, the local grayscale features, and the weighted average results of the image features, and then binds the filtered raw data to a unique uuid and outputs it to the slave camera. The slave camera processes the raw data into a two-dimensional array and performs the above-mentioned target recognition processing on the raw data to obtain the feature value of the raw data. Figure 3 It can also be known that the target image after eigenvalue processing can be regularized from the camera, specifically, light compensation, grayscale transformation, histogram equalization and other processing can be performed on the target image to reduce noise and interference in the target image for subsequent display and application of the target image.

[0064] Step S230, processing the target feature values ​​to obtain target recognition results in each target area in the original image.

[0065] After receiving the target feature value, the main camera can perform target feature modeling on the feature value, further process the target feature value extracted from the camera, and perform image visual feature processing, target pixel statistical feature processing, image transformation coefficient feature processing, and image algebraic feature processing on the basis of the target feature value extracted from the camera, so as to obtain a combined feature value for characterizing the original image that meets the requirements of the main camera. Figure 3As shown, the main camera can also combine the similarity threshold between the combined feature value and the reference feature value of the preset reference library, as well as the sensitivity of the target area, to obtain the recognition results of each target area in the original image. Among them, the sensitivity of the target area is specifically an indicator for measuring the detection quality of the target area, which may specifically include the clarity of the target area, the angle of the target, and the light intensity of the target. When the sensitivity reaches the preset sensitivity threshold, and the similarity between the combined feature value and the reference feature value of the reference library meets the similarity threshold, the recognition result of the corresponding target area in the original image can be determined according to the identity information of the corresponding target in the reference library. Among them, the reference feature value can be the value of the same type of feature as the combined feature value obtained after feature extraction and identification of the target area in the reference library in advance. In addition, the main camera compares the uuid sent from the camera with the uuid of the original data to determine the original data that has been identified, avoiding repeated processing of the same original data.

[0066] Figure 4 FIG. 1 is a business flow chart of the master-slave multi-terminal camera system of this embodiment. Figure 4 As shown, the main camera detects the target area of ​​the original image and determines the number of detected target areas. The main camera allocates a corresponding number of raw data to each slave camera. For example, each slave camera can process image data of 20 target areas at most. When the number of target areas is N times 20, N slave cameras can be selected, and each of the N slave cameras is allocated with raw data of 20 target areas. The N slave cameras perform feature value processing on the raw data to obtain the target feature value of the raw data, and perform target regularization processing on the image of the target area. After receiving the target feature value of the N slave cameras, the slave camera processes the target feature value to obtain the recognition result of each target area in the original image. Among them, when the target recognition method is applied to face recognition, the recognition result can specifically include time, personnel ID (Identity document), personnel external business ID, personnel code, personnel name, birthday, household registration address and other information.

[0067] In the above steps S210 to S230, the original image to be identified is obtained, the target area where the target is located is detected in the original image, and the original data of the target area is obtained, wherein the original image contains at least one target area, and the original data is sent to the slave camera with preset target recognition conditions in the master-slave multi-end camera system, and the slave camera performs target recognition processing based on the original data to obtain the target feature value of the original data, and processes the target feature value to obtain the target recognition result in each target area in the original image. It realizes the integration of multiple camera resources based on the master-slave multi-end camera system, thereby improving the chip processing capacity of the intelligent network camera, and then increasing the number of targets for target recognition by the intelligent network camera, thereby solving the problem of limited scenes for target recognition by the current intelligent network camera.

[0068] In one embodiment, based on the above step S210, in the original image, the target area where the target is located is detected to obtain the original data of the target area, including the following steps:

[0069] Step S211 , selecting a target area detection algorithm from a pre-built detection algorithm library according to preset detection requirements.

[0070] Specifically, the pre-built detection algorithm library can be combined with the actual application scenario, and the target detection algorithm suitable for the actual application scenario can be classified and managed according to indicators such as the accuracy, detection precision, computing resource occupancy rate, and detection efficiency (for example, 100 frames / second) of the algorithm. The detection requirement can be input by the user, or it can be pre-set in combination with the application scenario, which can specifically be the requirement for the above-mentioned accuracy, detection precision, computing resource occupancy rate, and detection efficiency of the target detection algorithm. In the case where the detection algorithm library contains a variety of different categories of target detection algorithms, the target detection algorithm that meets the detection requirement can be automatically matched from the detection algorithm library to process the original image. In the case where only a single target detection algorithm is managed in the detection algorithm library, the single target detection algorithm can be directly used to process the original image.

[0071] Step S212, using the selected target region detection algorithm, detects the target region where the target is located in the original image to obtain original data of the target region.

[0072] In addition, in one embodiment, based on the above step S220, the raw data is sent to a slave camera in the master-slave multi-terminal camera system that has a preset target recognition condition, and the slave camera performs target recognition processing based on the raw data, specifically including the following steps:

[0073] Step S221, the raw data is filtered according to a preset filter template, and the filtered raw data is sent to a slave camera with preset target recognition conditions in the master-slave multi-terminal camera system, and the slave camera performs target recognition processing based on the raw data.

[0074] Among them, the preset filter template can be a template corresponding to the image features to be filtered. After the raw data is filtered according to the preset filter template, a unique identifier can also be generated for the filtered raw data to achieve data consistency in the master-slave multi-terminal system. The filtered raw data is sent to the slave camera with preset target recognition conditions in the master-slave multi-terminal camera system for processing. The slave camera can share the processing pressure of the master camera. In addition, filtering the raw data can also reduce the pressure of the slave camera to process data and reduce the consumption of computing resources of the slave camera.

[0075] In addition, in one embodiment, based on the above step S220, the raw data is sent to a slave camera with a preset target recognition condition in the master-slave multi-terminal camera system, and the slave camera performs target recognition processing based on the raw data, and the following steps may also be included:

[0076] Step S222, distributes raw data to the slave cameras according to the number of target areas in the original image and the processing priority of the slave cameras that meet the preset target recognition conditions in the master-slave multi-terminal camera system.

[0077] Step S223: according to the allocation result, the original data is sent to the allocated slave camera, and the allocated slave camera performs target recognition processing.

[0078] Further, in one embodiment, based on the above step S222, in the master-slave multi-terminal camera system, the processing priority of the slave cameras that meet the preset target recognition conditions is determined by the ranking of the resource utilization rates of the slave cameras that meet the preset target recognition conditions in the master-slave multi-terminal camera system.

[0079] Among them, in the master-slave multi-end camera system, the master camera can obtain the CPU, memory and flash (flash memory) module usage of each slave camera through query. And according to the above hardware conditions of each slave camera, combined with the computing resources of the slave camera, the resource utilization rate of the slave camera is obtained, and the slave cameras in the master-slave multi-end camera system are sorted according to the resource utilization rate, so as to obtain the processing priority of the slave camera.

[0080] In one embodiment, based on the above step S220, the raw data is sent to a slave camera with preset target recognition conditions in the master-slave multi-terminal camera system, and the slave camera performs target recognition processing based on the raw data to obtain the target feature value of the raw data, and further includes the following steps:

[0081] Step S224 , allocating raw data to the slave cameras according to the categories of the target areas and the feature extraction capabilities of the slave cameras with preset target recognition conditions for target areas of different categories.

[0082] Among them, the category of the target area can be specifically determined by the category to which the target to be identified belongs. For example, if the target to be identified is a vehicle, the category of the target area is a vehicle area; if the target to be identified is a face, the category of the target area is a face area. Specifically, according to the category of the target area, the raw data can be assigned to a slave camera with the feature extraction capability of the category for processing. For example, in a master-slave multi-end camera system, slave camera A has a feature extraction capability corresponding to face recognition, slave camera B has a feature extraction capability corresponding to vehicle recognition, and slave camera C has a feature extraction capability for dog recognition. The master camera can assign the target area classified as a face area to slave camera A when slave camera A has a preset target recognition condition, that is, the hardware computing resources of the slave camera can process the target area detection. Similarly, the target area classified as a vehicle area is assigned to slave camera B, and the target area classified as a dog area is assigned to slave camera C.

[0083] Step S225 , according to the allocation result, the original data is sent to the allocated slave camera, and the allocated slave camera performs target recognition processing to obtain the target feature value of the original data.

[0084] When the master camera obtains original images containing different target areas respectively, different target areas can be assigned to corresponding slave cameras, so that the master-slave multi-end camera system can be compatible with multiple categories of target recognition and facilitate the deployment and upgrade of algorithm models.

[0085] Additionally, in one embodiment, based on the above step S230, the target feature value is processed to obtain the target recognition result in each target area in the original image, which specifically includes the following steps:

[0086] Step S231 , performing target feature modeling on the multiple target feature values ​​of each target area to obtain a combined feature value of each target area in the original image.

[0087] Step S232, comparing the combined feature values ​​of each target area with the reference feature values ​​in a preset reference library to obtain the recognition results of each target area in the original image.

[0088] Further, based on the above step S232, the combined feature value of each target area is compared with the reference feature value in the preset reference library to obtain the recognition result of each target area in the original image, which specifically includes the following steps:

[0089] Step S2321, compare the combined feature values ​​of each target area with the reference feature values ​​in a preset reference library to obtain the similarity between each target area and the reference area of ​​the preset reference library, and obtain the recognition result of each target area in the original image based on the similarity, similarity threshold and sensitivity of the target area.

[0090] The preset reference library is composed of images of objects whose identity information is determined in advance. The reference area is the area in the image of the reference library where the object is located. The reference feature value is the image feature of the image in the reference library. By comparing the combined feature value of each target area with the reference feature value in the preset reference library, the similarity after comparison is obtained. When the similarity reaches the similarity threshold and the sensitivity of the target area reaches the sensitivity threshold, the identity information corresponding to the target area in the reference library can be determined.

[0091] In the above steps S210 to S232, according to the preset detection requirements, a target area detection algorithm is selected from a pre-built detection algorithm library, and the target area where the target is located in the original image is detected using the selected target area detection algorithm to obtain raw data, and the processed raw data is sent to the slave camera for processing, thereby reducing the processing pressure of the master camera; according to the preset filtering template, the raw data is filtered, and the filtered raw data is sent to the slave camera with preset target recognition conditions in the master-slave multi-terminal camera system, and the slave camera performs target recognition processing based on the raw data, thereby reducing the computing resources consumed by the slave camera for target recognition processing and improving the processing performance of the master-slave multi-terminal camera system; finally, based on the master-slave multi-terminal camera system, the integration of multiple camera resources is realized, thereby improving the chip processing capability of the intelligent network camera, and then increasing the number of targets for target recognition by the intelligent network camera, thereby solving the problem of limited scenes for target recognition by the current intelligent network camera.

[0092] In this embodiment, another target recognition method is provided for a slave camera of a master-slave multi-terminal camera system. Figure 5 is a flow chart of another target recognition method of this embodiment, such as Figure 5 As shown, the process includes the following steps:

[0093] Step S510, obtaining raw data of a target area after a master camera in a master-slave multi-terminal camera system detects a target area where a target is located in an original image; wherein the original image contains at least one target area.

[0094] Step S520, performing target recognition processing on the original data to obtain the target feature value of the original data.

[0095] Step S530, sending the target feature value to the main camera, which processes the target feature value to obtain the recognition result of each target area in the original image.

[0096] In the above steps S510 to S530, the original data of the target area is obtained after the main camera in the master-slave multi-end camera system detects the target area where the target is located in the original image; wherein the original image contains at least one target area, the original data is processed for target recognition, the target feature value of the original data is obtained, the target feature value is sent to the main camera, and the main camera processes the target feature value to obtain the recognition result of each target area in the original image. It realizes the integration of multiple camera resources based on the master-slave multi-end camera system, thereby improving the chip processing capacity of the intelligent network camera, and then increasing the number of targets for target recognition by the intelligent network camera, thereby solving the problem of limited scenes for target recognition by the current intelligent network camera.

[0097] Further, in one embodiment, based on the above step S520, target recognition processing is performed on the original data to obtain the target feature value of the original data, which specifically includes the following steps:

[0098] Step S521, target area tracking, target optimization, and target angle and target attribute recognition processing are performed on the original data to obtain the target feature value of the original data.

[0099] The present embodiment is described and illustrated below through preferred embodiments.

[0100] Figure 6 FIG. 1 is a flow chart of the target recognition method of the preferred embodiment. Figure 6 As shown, the following steps are included:

[0101] Step S610, the main camera collects the original image, including image acquisition and data decoding, to obtain the original image in YUV format, and performs preprocessing such as size adjustment and format conversion on the original image in YUV format to obtain a preprocessed image;

[0102] Step S620, the master camera performs target area detection on the pre-processed image, determines the number of targets in the entire pre-processed image based on the detected target area, and sends the raw data of the detected target area to the corresponding slave camera for processing based on the number of targets;

[0103] Step S630, tracking, optimizing, angle, and target attribute recognition are performed on the raw data from the camera, and feature value preprocessing is performed;

[0104] Step S640, performing image processing such as grayscale correction and noise filtering on the target area after the eigenvalue preprocessing from the camera, specifically including light compensation, grayscale conversion, histogram equalization, normalization, geometric correction, filtering and sharpening, etc., to preferentially process the target area;

[0105] Step S650, performing target feature modeling based on relevant feature values ​​extracted from the camera to obtain combined feature values;

[0106] In step S660, the main camera compares the combined feature value with the reference feature value of the reference library and outputs a recognition result according to a similarity threshold.

[0107] It should be noted that the steps shown in the above process or the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0108] In this embodiment, a master-slave multi-terminal camera system 70 is also provided. Figure 7 As shown, it includes a main camera 72 and a slave camera 74, wherein: the main camera 72 is used to execute the target recognition method for the main camera in the above embodiment; the slave camera 74 is used to execute the target recognition method for the slave camera in the above embodiment; the main camera 72 and the slave camera 74 are communicatively connected.

[0109] Figure 8 7 is a flowchart of the business processing of the master camera 72 in the master-slave multi-terminal camera system 70. Figure 7 and Figure 8 As shown, on the one hand, the master camera 72 can maintain a list of slave cameras 74 and monitor the status of the slave cameras 74 through heartbeats. In addition, the master camera 72 will implement resource management of the slave cameras 74 through a three-dimensional array consisting of camera serial numbers, camera resource numbers, and camera usage rates. The three-dimensional array is the resource management information of the slave cameras 74. The list of the three-dimensional data is updated in real time and sorted according to the business usage rate of each slave camera 74. A heartbeat mechanism can also be established between the master and slave cameras. Specifically, the slave camera 74 can set a certain time interval to send a heartbeat to the master camera. If the master camera 72 receives the heartbeat of a slave camera, it means that the slave camera 74 is still active. If the master camera 72 does not receive the heartbeat of a slave camera 74, it means that the slave camera 74 is not in working state. The master camera 72 can determine which resources in the slave cameras 74 are available for selection through the list of active slave cameras 74 and the usage of resources.

[0110] On the other hand, the master camera 72 can control the addition and deletion of the slave cameras 74. If some slave cameras 74 need to be offline or new slave cameras 72 need to be added, they can be flexibly deleted and added through the master camera 72. Before adding or deleting the slave cameras 74, it is necessary to analyze and determine the status of the slave cameras 74 to determine whether the resources in the slave cameras 74 are being used. If so, the master camera 72 first needs to notify all slave cameras 74 not to add new services to the slave cameras 74 to be deleted, and wait until the slave cameras 74 to be deleted have completed processing the running services before removing the slave cameras 74 from the master-slave multi-end camera system 70.

[0111] Fig. 9 FIG. 7 is a flow chart of the business processing of the slave camera 74 in the master-slave multi-terminal camera system 70. Figure 7 and Fig. 9 As shown, the slave camera 74 includes a configuration module, an algorithm module, a receiving module, a sending module and an algorithm manager, wherein the above modules and the algorithm manager can be composed of a combination of hardware modules and software modules. When the receiving module of the slave camera receives the target recognition service request sent by the master camera, the receiving module of the slave camera 74 will first flow to the configuration module, and combine the media data and address information of the configuration module to form the data required for the service of processing the target recognition, so that the receiving module can return it to the master camera 72. The configuration module will make a judgment based on the performance of the slave camera 74. If the performance of the slave camera 74 does not match the service processing, it will inform the sending module that the resources of the slave camera 74 are currently insufficient and cannot be processed, otherwise the target recognition service will be received.

[0112] When performing target recognition processing, the algorithm module of the slave camera 74 will perform analysis and processing after receiving the complete raw data. The algorithm module performs feature value processing on the raw data and calls back the obtained feature value to the algorithm manager. The algorithm manager will regularize the target according to the feature value, and after the processing is completed, it will tell the main camera 72 the current resource status of the slave camera 74 through the sending module. The algorithm module of the slave camera 74 outputs the media information including the image frame after the target regularization, and the feature value after the target regularization will be combined into the format required in the reference library. After receiving the feature value and media information, the sending module of the slave camera 74 combines them into the form required by the main camera 72 and sends them to the main camera 72.

[0113] The above-mentioned master-slave multi-terminal camera system 70 can realize the integration of multiple camera resources by utilizing the master camera and the slave camera to respectively execute the corresponding target recognition methods, thereby improving the chip processing capability of the smart network camera, and further increasing the number of targets that the smart network camera can recognize, thereby solving the problem of limited scenarios for target recognition by the current smart network camera.

[0114] Additionally, in one embodiment, the master camera 72 is further configured to monitor the status of the slave cameras 74 through a heartbeat mechanism, and determine the slave cameras 74 that meet the preset target recognition conditions according to the status of the slave cameras 74 and resource management information.

[0115] In one embodiment, the slave camera 74 is further configured to send resource change information to the master camera 72 when its own resources are changed, so that the master camera 72 updates the resource management information of the slave camera 74 .

[0116] It should be understood that the specific embodiments described herein are only used to explain the application, rather than to limit it. Based on the embodiments provided in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the protection scope of this application.

[0117] Obviously, the drawings are only some examples or embodiments of the present application. For ordinary technicians in the field, the present application can also be applied to other similar situations based on these drawings without creative work. In addition, it is understandable that although the work done in this development process may be complicated and lengthy, for ordinary technicians in the field, certain changes in design, manufacturing or production based on the technical content disclosed in this application are only conventional technical means and should not be regarded as insufficient content disclosed in this application.

[0118] The term "embodiment" in this application refers to a specific feature, structure or characteristic described in conjunction with the embodiment that can be included in at least one embodiment of the present application. The appearance of this phrase in various locations in the specification does not necessarily mean the same embodiment, nor does it mean that it is mutually exclusive with other embodiments and is independent or optional. It is clearly or implicitly understood by those of ordinary skill in the art that the embodiments described in this application can be combined with other embodiments without conflict.

[0119] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of patent protection. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the scope of protection of the present application. Therefore, the scope of protection of the present application shall be subject to the attached claims.

Claims

1. A target recognition method for a master camera in a master-slave multi-terminal camera system, characterized in that: The method comprises: Acquire an original image to be identified, and detect a target area where a target is located in the original image to obtain original data of the target area; wherein the original image contains at least one target area; The raw data is sent to a slave camera in the master-slave multi-terminal camera system that has a preset target recognition condition, and the slave camera performs target recognition processing based on the raw data to obtain a target feature value of the raw data; After receiving the target feature value, the target feature value is processed to obtain the target recognition result in each target area in the original image.

2. The target recognition method according to claim 1, characterized in that: The detecting of the target area where the target is located in the original image to obtain the original data of the target area includes: According to the preset detection requirements, select the target area detection algorithm from the pre-built detection algorithm library; The target region detection algorithm that has been selected is used to detect the target region where the target is located in the original image to obtain original data of the target region.

3. The target recognition method according to claim 1, characterized in that: The sending of the raw data to a slave camera having a preset target recognition condition in the master-slave multi-terminal camera system, and the slave camera performing target recognition processing based on the raw data, includes: The raw data is filtered according to a preset filter template, and the filtered raw data is sent to a slave camera with preset target recognition conditions in the master-slave multi-terminal camera system, and the slave camera performs target recognition processing based on the raw data.

4. The target recognition method according to claim 1, characterized in that: The sending of the raw data to a slave camera having a preset target recognition condition in the master-slave multi-terminal camera system, and the slave camera performing target recognition processing based on the raw data, further includes: Allocating raw data to the slave cameras according to the number of target areas in the original image and the processing priority of the slave cameras having preset target recognition conditions in the master-slave multi-terminal camera system; According to the allocation result, the original data is sent to the allocated slave camera, and the allocated slave camera performs target recognition processing.

5. The target recognition method according to claim 4, characterized in that: In the master-slave multi-terminal camera system, the processing priority of the slave cameras that meet the preset target recognition conditions is determined by the order of resource usage rates of the slave cameras that meet the preset target recognition conditions in the master-slave multi-terminal camera system.

6. The target recognition method according to claim 1, characterized in that: The sending of the raw data to a slave camera having a preset target recognition condition in the master-slave multi-terminal camera system, wherein the slave camera performs target recognition processing based on the raw data to obtain a target feature value of the raw data, further includes: Allocating raw data to the slave cameras according to the categories of the target areas and the feature extraction capabilities of the slave cameras with preset target recognition conditions for the target areas of different categories; According to the allocation result, the original data is sent to the allocated slave camera, and the allocated slave camera performs target recognition processing to obtain the target feature value of the original data.

7. The target recognition method according to claim 1, characterized in that: The processing of the target feature values ​​to obtain target recognition results in each target area in the original image includes: Performing target feature modeling on the multiple target feature values ​​of each target area to obtain a combined feature value of each target area in the original image; The combined feature values ​​of each target area are compared with reference feature values ​​in a preset reference library to obtain target recognition results in each target area in the original image.

8. The target recognition method according to claim 7, characterized in that: The step of comparing the combined feature values ​​of each target area with reference feature values ​​in a preset reference library to obtain target recognition results in each target area in the original image includes: The combined feature values ​​of each target area are compared with the reference feature values ​​in a preset reference library to obtain the similarity between each target area and the reference area of ​​the preset reference library, and the target recognition results in each target area in the original image are obtained based on the similarity, a preset similarity threshold and the sensitivity of the target area.

9. A target recognition method for a slave camera of a master-slave multi-terminal camera system, characterized in that: The method comprises: Obtaining raw data of a target area obtained after a master camera in the master-slave multi-terminal camera system detects a target area where a target is located in an original image; wherein the original image contains at least one target area; Performing target recognition processing on the original data to obtain target feature values ​​of the original data; The target feature value is sent to the main camera, and after receiving the target feature value, the main camera processes the target feature value to obtain the target recognition result in each target area in the original image.

10. The target recognition method according to claim 9, characterized in that: The performing target recognition processing on the original data to obtain the target feature value of the original data includes: The original data is subjected to target tracking, target optimization, and recognition processing of target angle and target attribute to obtain target feature values ​​of the original data.

11. A master-slave multi-end camera system, comprising a master camera and a slave camera; wherein: The main camera is used to perform the target recognition method according to any one of claims 1 to 8; The slave camera is used to perform the target recognition method according to any one of claims 9 to 10; The master camera and the slave camera are communicatively connected.

12. The master-slave multi-terminal camera system according to claim 11, characterized in that: The master camera is further configured to monitor the status of the slave cameras through a heartbeat mechanism, and determine a slave camera that meets the preset target recognition condition according to the status of the slave cameras and resource management information.

13. The master-slave multi-terminal camera system according to claim 11, characterized in that: The slave camera is further configured to send resource change information to the master camera when its own resources are changed, so that the master camera updates the resource management information of the slave camera.

Images