Image recognition test method, control device, test camera and test system

Abstract

The invention discloses an image recognition test method, a control device, a test camera and a test system, relates to the technical field of image recognition test, and discloses an image recognition test method which comprises the following steps: acquiring image data in a current simulation test scene; and inputting the image data to a preset artificial intelligence recognition algorithm, obtaining a recognition result of the preset artificial intelligence recognition algorithm for the current simulation test scene, determining the classification of the current simulation test scene, and classifying the image data to the storage partitions of the corresponding classes. And respectively executing the steps on various types of simulation test scenes. And then, according to whether the storage partitions classified by the image data in different simulation test scenes are matched with the preset partitions of the corresponding simulation test scenes, obtaining an image recognition result of a preset artificial intelligence recognition algorithm. The invention aims to solve the limitation of an existing method in the aspect of evaluating the accuracy of an artificial intelligence recognition algorithm in an image recognition system.

Description

Technical Field

[0001] The present application relates to the field of image recognition testing technology, and in particular to an image recognition testing method, a control device, a testing camera and a testing system. Background Art

[0002] With the rapid development of computer vision and artificial intelligence technology, image recognition technology has been widely used in many fields, such as autonomous driving, security monitoring, medical image analysis, etc. However, in practical applications, in order to ensure the accuracy and reliability of the image recognition system, the artificial intelligence recognition algorithm of the system must be strictly tested.

[0003] Traditional image recognition testing methods mainly rely on collecting multiple sample data in real-world environments for testing, including but not limited to target objects such as people, vehicles, and animals. However, this method not only consumes a lot of manpower and material resources, but also faces the problem of insufficient data diversity, especially for those targets that are difficult to capture or appear less frequently (such as specific species of animals). This makes it difficult for the test data set to fully cover all potential recognition objects, thereby limiting the evaluation of the accuracy of the system's artificial intelligence recognition algorithm.

[0004] The above contents are only used to assist in understanding the technical solution of the present application and do not constitute an admission that the above contents are prior art. Summary of the invention

[0005] The main purpose of this application is to provide an image recognition testing method, which aims to address the limitations of existing methods in evaluating the accuracy of artificial intelligence recognition algorithms in image recognition systems.

[0006] To achieve the above objectives, the present application proposes an image recognition test method, the steps of which include:

[0007] S10, obtaining image data in the current simulation test scenario;

[0008] S20, inputting the image data into a preset artificial intelligence recognition algorithm to obtain a recognition result of the preset artificial intelligence recognition algorithm for the current simulation test scene;

[0009] S30, determining the classification of the current simulation test scene according to the recognition result, and classifying the image data into a storage partition of the corresponding class;

[0010] S40, for various types of simulation test scenarios, respectively execute steps S10 to S30;

[0011] S50. Based on the preset partitions of multiple classified simulation test scenarios, the image recognition results of the preset artificial intelligence recognition algorithm are obtained according to whether the storage partitions classified by the image data under different simulation test scenarios match the preset partitions of the corresponding simulation test scenarios; wherein the preset partitions are used to indicate the storage partitions of the preset classification of the image data under different simulation test scenarios.

[0012] In one embodiment, the classification of simulation test scenes includes human test scenes, vehicle test scenes, animal test scenes and coexistence test scenes; wherein the human test scenes are used to test human images, the vehicle test scenes are used to test vehicle images, the animal test scenes are used to test animal images, and the coexistence test scenes are used to test at least two types of coexistence images of humans, vehicles and animals; the storage partitions include human image storage partitions, vehicle image storage partitions, animal image storage partitions and coexistence image storage partitions;

[0013] The step S30 comprises:

[0014] According to the recognition result, if it is determined that the current simulation test scene is classified as a human test scene, the image data is classified into a human image storage partition;

[0015] If it is determined that the current simulation test scene is classified as a vehicle test scene, the image data is classified into a vehicle image storage partition;

[0016] If it is determined that the current simulation test scene is classified as an animal test scene, the image data is classified into an animal image storage partition;

[0017] If it is determined that the current simulation test scene is classified as a coexistence test scene, the image data is classified into a coexistence image storage partition.

[0018] In one embodiment, the step S40 includes:

[0019] For the human test scene, the vehicle test scene, the animal test scene and the coexistence test scene, steps S10 to S30 are respectively performed.

[0020] In one embodiment, the image recognition result includes a person image recognition result, a vehicle image recognition result, an animal image recognition result, and a coexistence image recognition result; the step S50 includes:

[0021] Based on the preset partitions of the simulation test scenes with multiple classifications, the storage partitions classified according to the image data in the human test scene match the preset classified human image storage partitions, and the human image recognition results of the preset artificial intelligence recognition algorithm are obtained;

[0022] According to whether the storage partition of the image data in the vehicle test scenario matches the storage partition of the preset classified vehicle image, the vehicle image recognition result of the preset artificial intelligence recognition algorithm is obtained;

[0023] According to whether the storage partition of the image data in the animal test scene matches the preset classified animal image storage partition, the animal image recognition result of the preset artificial intelligence recognition algorithm is obtained;

[0024] According to whether the storage partition of the image data in the coexistence test scenario matches the storage partition of the coexistence image of the preset classification, the coexistence image recognition result of the preset artificial intelligence recognition algorithm is obtained.

[0025] In addition, to achieve the above-mentioned purpose, the present application also proposes a control device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the image recognition test method described above.

[0026] In addition, to achieve the above-mentioned purpose, the present application also proposes a test camera, which includes the control device as described above, and also includes a camera module, which is used to collect image data of simulation test scenes of different categories and input the image data into the control device.

[0027] In one embodiment, the camera module includes:

[0028] A passive infrared sensor, the passive infrared sensor is used to obtain temperature change information in a simulation test scenario and input the temperature change information into the control device;

[0029] A camera assembly, the camera assembly is used to collect image data of the simulation test scene and input the image data to the control device;

[0030] The control device is also used to control the camera module to work when it is determined that there is an object to be identified in the simulation test scene according to the temperature change information; wherein the object to be identified includes at least one of a person, a vehicle and an animal.

[0031] In addition, to achieve the above-mentioned purpose, the present application also proposes a test system, which includes the test camera as described above and also includes a simulation test component, and the simulation test component is used to construct multiple classifications of simulation test scenarios.

[0032] In one embodiment, the simulation test component includes at least one of a simulation animal component, a simulation human component and a simulation vehicle component.

[0033] In one embodiment, the simulation test component includes:

[0034] The simulation model body is used to simulate the morphology of the object to be identified;

[0035] A temperature control module, disposed on the surface of the simulation model body, for simulating the temperature of the object to be identified;

[0036] A driving module, arranged at an actionable part of the simulation model body, for simulating the action of the object to be identified;

[0037] The remote control module comprises a control unit and a plurality of controlled units; the plurality of controlled units of the remote control module are electrically connected to the temperature control module and the driving module respectively; the control unit of the remote control module is communicatively connected to the plurality of controlled units of the remote control module; the control unit of the remote control module is used to output control instructions when triggered by a user; the control instructions are used to instruct the operation of the temperature control module and / or the operation of the driving module.

[0038] One or more technical solutions proposed in this application have at least the following technical effects:

[0039] Step S10, obtaining image data in the current simulation test scenario. Step S20, inputting the image data into a preset artificial intelligence recognition algorithm, and obtaining the recognition result of the preset artificial intelligence recognition algorithm for the current simulation test scenario. Step S30, determining the classification of the current simulation test scenario based on the recognition result, and classifying the image data into storage partitions of corresponding categories. Step S40, executing steps S10 to S30 respectively for simulation test scenarios of multiple classifications. Step S50, based on preset partitions of simulation test scenarios of multiple classifications, and according to whether the storage partitions classified by the image data in different simulation test scenarios match the preset partitions of the corresponding simulation test scenarios, obtaining the image recognition result of the preset artificial intelligence recognition algorithm; wherein the preset partitions are used to indicate the storage partitions of preset classifications of the image data in different simulation test scenarios. In this way, the present application can simulate high-cost objects or targets with low frequency of occurrence in simulation test scenarios by using simulation models such as simulated vehicles and simulated animals, thereby solving the problem of insufficient test data diversity in existing testing methods. Image recognition results of preset artificial intelligence recognition algorithms can be obtained in multiple classification simulation test scenarios, thereby improving the ability to evaluate the accuracy of artificial intelligence recognition algorithms in image recognition systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The accompanying drawings herein are incorporated in and constitute a part of the specification, illustrate embodiments matching the present application, and together with the description, are used to explain the principles of the present application.

[0041] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0042] Figure 1 A flowchart of an embodiment of the image recognition test method of the present application is provided;

[0043] Figure 2 A flowchart diagram of another embodiment of the image recognition test method of the present application is provided;

[0044] Figure 3 A schematic diagram of the structure of an embodiment of a control device of the present application;

[0045] Figure 4 A schematic diagram of the structure provided for an embodiment of a test system of the present application;

[0046] Figure 5 A schematic diagram of the structure of a simulation test component provided for an embodiment of the test system of the present application.

[0047] The purpose, features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0048] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0049] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0050] The main solution of the embodiment of the present application is: Step S10, obtaining the image data in the current simulation test scene. Step S20, inputting the image data into the preset artificial intelligence recognition algorithm, and obtaining the recognition result of the preset artificial intelligence recognition algorithm for the current simulation test scene. Step S30, determining the classification of the current simulation test scene according to the recognition result, and classifying the image data into the storage partition of the corresponding class. Step S40, for simulation test scenes of multiple classifications, respectively execute steps S10 to S30. Step S50, based on the preset partitions of simulation test scenes of multiple classifications, according to whether the storage partitions classified by the image data in different simulation test scenes match the preset partitions of the corresponding simulation test scenes, obtain the image recognition result of the preset artificial intelligence recognition algorithm; wherein, the preset partitions are used to indicate the storage partitions of the preset classification of the image data in different simulation test scenes.

[0051] Technical terms involved in the embodiments of this application:

[0052] CNN (Convolutional Neural Network): It extracts feature levels in images through multi-layer convolution operations, can effectively capture the spatial structure information in images, and is widely used in image classification and feature extraction tasks.

[0053] SVM (Support Vector Machine): A classic classification algorithm, usually used in conjunction with a feature extractor (such as HOG features). It performs well when processing small-scale data sets and can perform accurate classification tasks.

[0054] YOLO (You Only Look Once): An algorithm for real-time object detection that significantly improves detection speed and efficiency by dividing the image into a grid and predicting both the bounding box and the category in a single forward propagation.

[0055] R-CNN (Region-based Convolutional Neural Network): The R-CNN algorithm first generates region proposals and then uses a convolutional neural network to extract the features of these regions. It is suitable for complex object detection tasks. Although the processing speed is slow, the accuracy is high.

[0056] In this embodiment, for the convenience of description, the following description is made by taking the control device of the image recognition test method as the execution subject.

[0057] Traditional image recognition testing methods mainly rely on collecting multiple sample data in real-world environments for testing, including but not limited to target objects such as people, vehicles, and animals. However, this method not only consumes a lot of manpower and material resources, but also faces the problem of insufficient data diversity, especially for those targets that are difficult to capture or appear less frequently (such as specific species of animals). This makes it difficult for the test data set to fully cover all potential recognition objects, thereby limiting the evaluation of the accuracy of the system's artificial intelligence recognition algorithm.

[0058] The present application provides a solution that can be used in simulation test scenarios to simulate high-cost objects or targets with low frequency of occurrence through simulation models such as simulated vehicles and simulated animals, thereby solving the problem of insufficient test data diversity in existing testing methods and being able to obtain image recognition results of preset artificial intelligence recognition algorithms in multiple classification simulation test scenarios, thereby improving the ability to evaluate the accuracy of artificial intelligence recognition algorithms in image recognition systems.

[0059] It should be noted that the execution subject of this embodiment may be a computing service device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., or an electronic device capable of realizing the above functions, a control device, etc. The following takes the control device as an example to illustrate this embodiment and the following embodiments.

[0060] Based on this, the present application embodiment provides an image recognition test method, referring to Figure 1 , Figure 1 This is a flowchart of the first embodiment of the image recognition test method of the present application.

[0061] In this embodiment, the image recognition test method includes steps S10 to S50:

[0062] Step S10, obtaining image data in the current simulation test scenario.

[0063] It should be noted that the object under test is specifically a preset artificial intelligence recognition algorithm in an image recognition system, for example, it can be a preset artificial intelligence recognition algorithm in an intelligent security monitoring camera, and the test camera can be an intelligent security monitoring camera with a built-in control device. In the current simulation test scenario, the camera of the intelligent security monitoring camera can collect image data and output it to the control device. Among them, the control device has a built-in preset artificial intelligence recognition algorithm. The preset artificial intelligence recognition algorithm can classify and screen objects in the image according to a pre-trained model. It is understandable that in the research and development and testing stage of the preset artificial intelligence recognition algorithm, due to the lack of a suitable wild forest environment with multiple animals, it is impossible to make a comprehensive and accurate evaluation test of the accuracy of the recognition algorithm. In this embodiment, by simulating a variety of real application scenarios of the camera through simulation models such as simulated vehicles and simulated animals, the ability to evaluate the accuracy of the preset artificial intelligence recognition algorithm in the intelligent security monitoring camera can be improved.

[0064] It should be noted that the simulation test scenario is specifically a simulation test scenario simulated by a realistic simulation model. For example, in a laboratory or a special test site, simulation models such as simulated vehicles and simulated animals can be used to simulate high-cost objects (such as expensive vehicles) or targets with low frequency of appearance (such as specific types of animals). Through simulation of simulation test scenarios of different categories, richer scenarios can be tested, so that the test data set can fully cover most or even all potential recognition objects, thereby comprehensively evaluating the overall image recognition capability of the system.

[0065] In a feasible implementation, step S10 may include steps S11 to S12:

[0066] Step S11, obtaining temperature change information in the current simulation test scenario;

[0067] Step S12: when it is determined, based on the temperature change information, that there is an object to be identified in the current simulation test scene, image data of the object to be identified in the current simulation test scene is obtained.

[0068] It should be noted that the smart security monitoring camera can also have a built-in PIR (passive infrared) sensor to detect the movement of heat sources in the test environment. When the object to be identified (such as a person, vehicle or animal) enters its monitoring range and causes a temperature difference change, the PIR sensor triggers the camera to start image acquisition. At this time, the camera will automatically start capturing image data, which includes visible light image information. If the camera is equipped with a thermal imaging sensor, the image data may also include thermal imaging information. Then, by inputting information such as light image information and thermal imaging information into a preset artificial intelligence recognition algorithm, the preset artificial intelligence recognition algorithm can output a recognition result, such as identifying a person, vehicle or animal.

[0069] Step S20, inputting the image data into a preset artificial intelligence recognition algorithm to obtain a recognition result of the preset artificial intelligence recognition algorithm for the current simulation test scene.

[0070] In this embodiment, the control device pre-stores a preset artificial intelligence recognition algorithm for image recognition, and this embodiment is intended to comprehensively evaluate the accuracy of the preset artificial intelligence recognition algorithm. Among them, the preset artificial intelligence recognition algorithm may include various deep learning models, such as CNN (convolutional neural network), SVM (support vector machine), YOLO (You Only LookOnce) and R-CNN (Region-based Convolutional Neural Network). These algorithms can optimize the recognition model parameters by learning a large amount of annotated training data, thereby improving the recognition ability of image data. In this way, when the image data in the current simulation test scenario is input, the recognition model in the preset artificial intelligence recognition algorithm can predict the object category information in the image based on the previous learning results, and generate corresponding recognition results.

[0071] It should be noted that when the camera includes a display module, the control device can control the corresponding recognition result to be marked on the display module accordingly, so as to facilitate the tester to obtain the recognition result of the current simulation test scene. For example, when the current simulation test scene is a simulated animal passing by, the recognition result is marked on the display module, such as marking the animal classification, which can inform the tester that the preset artificial intelligence recognition algorithm has recognized the animal image in the current simulation test scene.

[0072] Step S30, determining the classification of the current simulation test scene according to the recognition result, and classifying the image data into a storage partition of the corresponding class.

[0073] It should be noted that the preset artificial intelligence recognition algorithm under different application scenarios can design simulation test scenes of different classifications. For example, when the preset artificial intelligence recognition algorithm is used to classify people, vehicles and animals, the classification of the current simulation test scene can include people test scenes, vehicle test scenes, animal test scenes and coexistence test scenes. Among them, the people test scene is used to test people images, the vehicle test scene is used to test vehicle images, the animal test scene is used to test animal images, and the coexistence test scene is used to test at least two types of coexistence images among people, vehicles and animals. The storage partition includes a people image storage partition, a vehicle image storage partition, an animal image storage partition and a coexistence image storage partition. Then, when the preset artificial intelligence recognition algorithm can correctly identify, the person image in the person test scene will be classified into the person image storage partition, the vehicle image in the vehicle test scene will be classified into the vehicle image storage partition, the animal image in the animal test scene will be classified into the animal image storage partition, and at least two types of coexistence images of the coexistence test scene will be saved to the coexistence image storage partition.

[0074] Step S40, executing steps S10 to S30 respectively for various types of simulation test scenarios.

[0075] It should be noted that in order to ensure the test accuracy of the preset artificial intelligence recognition algorithm, a comprehensive test of multiple classifications of simulation test scenarios can be carried out. In this way, the performance of the algorithm in various complex environments can be fully verified. Specifically, in the character test scene, ten different character test scenes can be arranged, each of which simulates different environmental conditions, such as different light (day, night), different backgrounds (city streets, country roads), different character actions (walking, running, riding), etc. In the vehicle test scene, ten different vehicle test scenes can be arranged, each of which contains different types of vehicles (such as cars, trucks, motorcycles, etc.), different driving states (such as uniform speed, acceleration, deceleration), and different environmental conditions (such as busy traffic hours, empty roads). In the animal test scene, ten different animal test scenes can be arranged, each of which simulates different types of animals (such as dogs, cats, birds, small mammals), different activity modes (such as running, flying, stillness), and different environments (such as parks, residential areas, and the wild). In the coexistence test scenario, ten different coexistence test scenarios can be arranged, each of which contains people, vehicles and animals, simulating various complex scenarios that may be encountered in real life. In this way, a large amount of image data can be obtained in different simulation test scenarios, and the recognition and classification results can be obtained through the preset artificial intelligence recognition algorithm.

[0076] Step S50, based on the preset partitions of multiple classified simulation test scenarios, and according to whether the storage partitions classified by the image data under different simulation test scenarios match the preset partitions of the corresponding simulation test scenarios, obtain the image recognition results of the preset artificial intelligence recognition algorithm; wherein the preset partitions are used to indicate the storage partitions of the preset classification of the image data under different simulation test scenarios.

[0077] It should be noted that the preset partitions of multiple classifications of simulation test scenes can be preset before the simulation test is performed. For example, a total of four groups of simulation test scenes are preset. The first group of simulation test scenes are human simulation test scenes, which will be scheduled for testing in the first period; the second group of simulation test scenes are vehicle simulation test scenes, which will be scheduled for testing in the second period; the third group of simulation test scenes are animal simulation test scenes, which will be scheduled for testing in the third period; the fourth group of simulation test scenes are coexistence simulation test scenes, which will be scheduled for testing in the fourth period. The preset partitions can indicate that the image data is classified into the human storage partition in the first period, the vehicle storage partition in the second period, the animal storage partition in the third period, and the coexistence storage partition in the fourth period. Then, the tests are performed in the first period to the fourth period respectively. If all the tests pass, that is, the image data is classified into the human storage partition in the first period, the vehicle storage partition in the second period, the animal storage partition in the third period, and the coexistence storage partition in the fourth period, it is concluded that the image recognition accuracy of the current preset artificial intelligence recognition algorithm is high. If the image data is misclassified or not recognized in a certain simulation test scenario, it will be concluded that the algorithm has misidentified or missed recognition and has low accuracy. It is necessary to further optimize and improve the algorithm based on the image recognition results of the preset artificial intelligence recognition algorithm.

[0078] In this way, this embodiment can simulate high-cost objects or targets with low frequency of occurrence through simulation models such as simulated vehicles and simulated animals for testing, which solves the problem of insufficient diversity of test data in existing testing methods, and can obtain the image recognition results of the preset artificial intelligence recognition algorithm in multiple classifications of simulation test scenarios, thereby improving the accuracy of the evaluation of the image recognition ability of the preset artificial intelligence recognition algorithm. Moreover, in this embodiment, according to whether the storage partitions classified by the image data in different simulation test scenarios match the preset partitions of the corresponding simulation test scenarios, the image recognition results of the preset artificial intelligence recognition algorithm for image recognition in different simulation test scenarios can be automatically obtained, and then the accuracy of the artificial intelligence recognition algorithm can be obtained. Compared with manual matching, this embodiment saves labor costs.

[0079] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the above-mentioned embodiment one can refer to the above introduction, and will not be repeated later. On this basis, the classification of simulation test scenes includes human test scenes, vehicle test scenes, animal test scenes and coexistence test scenes. Among them, the human test scene is used to test human images, the vehicle test scene is used to test vehicle images, the animal test scene is used to test animal images, and the coexistence test scene is used to test at least two types of coexistence images of people, vehicles and animals. The storage partitions include human image storage partitions, vehicle image storage partitions, animal image storage partitions and coexistence image storage partitions.

[0080] Reference Figure 2 , step S30 includes step S31:

[0081] Step S31, according to the recognition result, if it is determined that the current simulation test scene is classified as a human test scene, the image data is classified into a human image storage partition;

[0082] If it is determined that the current simulation test scene is classified as a vehicle test scene, the image data is classified into a vehicle image storage partition;

[0083] If it is determined that the current simulation test scene is classified as an animal test scene, the image data is classified into an animal image storage partition;

[0084] If it is determined that the current simulation test scene is classified as a coexistence test scene, the image data is classified into a coexistence image storage partition.

[0085] It should be noted that the coexistence test scene is used to test two types of coexistence images of people and vehicles, two types of coexistence images of people and animals, two types of coexistence images of vehicles and animals, or three types of coexistence images of people, vehicles and animals. That is, when the simulation test scene obtains image data of at least two of people, vehicles and animals, the image data under the simulation test scene is stored in the coexistence image storage partition.

[0086] It should be noted that in actual image recognition application scenarios, people, vehicles and pets are the most common recognition objects. Therefore, in this embodiment, the classification of simulation test scenes can be divided into four categories: people test scenes, vehicle test scenes, animal test scenes and coexistence test scenes for testing, which can cover a wide range of application scenarios on the basis of simplifying the complexity of the test. In this embodiment, according to the recognition results, it can be determined that the current simulation test scene is classified as a people test scene, a vehicle test scene, an animal test scene or a coexistence test scene, and the image data is classified into the storage partition of the corresponding class.

[0087] Based on the second embodiment of the present application, in the third embodiment of the present application, the same or similar contents as those in the above-mentioned second embodiment can refer to the above introduction, and will not be repeated in the following. Figure 2 , step S40 includes step S41:

[0088] Step S41 , executing steps S10 to S30 for the human test scene, the vehicle test scene, the animal test scene, and the coexistence test scene, respectively.

[0089] In this embodiment, seven groups of test scenarios can be set. In the first group of test scenarios, simulated characters are included, and the actions of the simulated characters are remotely controlled to simulate real character test scenarios. In the second group of test scenarios, simulated animals are included, and the actions of the simulated animals are remotely controlled to simulate real animal test scenarios. In the third group of test scenarios, simulated vehicles are included, and the actions of the simulated vehicles are remotely controlled to simulate real vehicle test scenarios. In the fourth group of test scenarios, simulated characters and simulated animals are included, and the actions of the simulated characters and simulated animals are remotely controlled to simulate real coexistence test scenarios. In the fifth group of test scenarios, simulated characters and simulated vehicles are included, and the actions of the simulated characters and simulated vehicles are remotely controlled to simulate real coexistence test scenarios. In the sixth group of test scenarios, simulated vehicles and simulated animals are included, and the actions of the simulated vehicles and simulated animals are remotely controlled to simulate real coexistence test scenarios. In the seventh group of test scenarios, simulated characters, simulated vehicles and simulated animals are included, and the actions of the simulated characters, simulated vehicles and simulated animals are remotely controlled to simulate real coexistence test scenarios. By executing steps S10 to S30 for the seven groups of test scenes respectively, classification results of the image data in the human test scene, the vehicle test scene, the animal test scene and the coexistence test scene can be obtained respectively.

[0090] Based on the third embodiment of the present application, in the fourth embodiment of the present application, the same or similar contents as those in the third embodiment can be referred to the above description, and will not be described in detail later. Figure 2 , step S50 includes step S51:

[0091] Step S51, based on the preset partitions of the simulation test scenes of multiple classifications, the storage partitions classified according to the image data in the human test scene match the preset classified human image storage partitions, and obtain the human image recognition result of the preset artificial intelligence recognition algorithm;

[0092] According to whether the storage partition of the image data in the vehicle test scenario matches the storage partition of the preset classified vehicle image, the vehicle image recognition result of the preset artificial intelligence recognition algorithm is obtained;

[0093] According to whether the storage partition of the image data in the animal test scene matches the preset classified animal image storage partition, the animal image recognition result of the preset artificial intelligence recognition algorithm is obtained;

[0094] According to whether the storage partition of the image data in the coexistence test scenario matches the storage partition of the coexistence image of the preset classification, the coexistence image recognition result of the preset artificial intelligence recognition algorithm is obtained.

[0095] In this way, this embodiment can obtain the results of person image recognition, vehicle image recognition, animal image recognition and coexistence image recognition respectively, and can test the accuracy of the preset artificial intelligence recognition algorithm in recognizing images of people, vehicles and animals respectively, and test the accuracy of the preset artificial intelligence recognition algorithm in recognizing more than two images of people, vehicles and animals, and can comprehensively analyze and test the image recognition ability of the preset artificial intelligence recognition algorithm. In this way, if the preset artificial intelligence recognition algorithm has weak image recognition ability for a specific scene, the recognition model built into the preset artificial intelligence recognition algorithm can be further trained to enhance the image recognition ability of the recognition model in the corresponding scene until the image recognition accuracy requirements can be met.

[0096] The present application provides a control device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the image recognition test method in the above-mentioned embodiment one.

[0097] Refer to the following Figure 3 , Figure 3 A schematic diagram of a control device according to an embodiment of the present application. The control device according to the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, laptop computers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as desktop computers, etc. Figure 3 The control device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0098] like Figure 3As shown, the control device may include a processing device 1001 (e.g., a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM: Read Only Memory) 1002 or a program loaded from a storage device 1003 to a random access memory (RAM: Random Access Memory) 1004. In RAM1004, various programs and data required for the operation of the control device are also stored. The processing device 1001, ROM1002, and RAM1004 are connected to each other through a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Generally, the following systems can be connected to the I / O interface 1006: an input device 1007 including, for example, a touch screen, a touch pad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output device 1008 including, for example, a liquid crystal display (LCD: Liquid Crystal Display), a speaker, a vibrator, etc.; a storage device 1003 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 1009. The communication device 1009 can allow the control device to communicate with other devices wirelessly or by wire to exchange data. Although the control device with various systems is shown in the figure, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems can be implemented or have alternatively.

[0099] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0100] The control device provided by the present application adopts the image recognition test method in the above embodiment, which can solve the limitations of the existing method in evaluating the accuracy of the artificial intelligence recognition algorithm in the image recognition system. Compared with the prior art, the beneficial effects of the control device provided by the present application are the same as the beneficial effects of the image recognition test method provided by the above embodiment, and the other technical features in the control device are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.

[0101] This application provides a test camera, referring to Figure 4The test camera includes the control device as described above, and also includes a camera module, which is used to collect image data of simulation test scenes of different categories and input the image data to the control device.

[0102] The test camera provided by the present application, including the control device as described above, can solve the limitations of the existing methods in evaluating the accuracy of artificial intelligence recognition algorithms in image recognition systems. Compared with the prior art, the beneficial effects of the test camera provided by the present application are the same as those of the control device as described above, and will not be repeated here.

[0103] In one embodiment of the test camera provided in the present application, refer to Figure 4 , the camera module includes:

[0104] Passive infrared sensor, the passive infrared sensor is used to obtain temperature change information in the simulation test scene and input the temperature change information into the control device;

[0105] A camera assembly, the camera assembly is used to collect image data of the simulation test scene and input the image data into the control device;

[0106] The control device is also used to control the camera module to work when it is determined that there is an object to be identified in the simulation test scene according to the temperature change information; wherein the object to be identified includes at least one of a human object, a vehicle object and an animal object.

[0107] In this embodiment, when an object to be identified (such as a person, simulated vehicle or simulated animal) enters the monitoring range of the test camera and causes a temperature difference change, the control device can identify it through the passive infrared sensor, and then control the camera assembly to start image acquisition. The test camera of this embodiment is closer to a camera that can be used in the wild, such as an intelligent security monitoring camera, and can effectively test the accuracy of its built-in preset artificial intelligence recognition algorithm.

[0108] The present application provides a testing system, referring to Figure 4 The test system includes the test camera as described above, and also includes a simulation test component, which is used to construct multiple classifications of simulation test scenarios.

[0109] The test system provided by the present application, including the test camera described above, can solve the limitations of existing methods in evaluating the accuracy of artificial intelligence recognition algorithms in image recognition systems. Compared with the prior art, the beneficial effects of the test system provided by the present application are the same as those of the test camera described above, and will not be repeated here.

[0110] In one embodiment of the test system provided by the present application, the simulation test component includes at least one of a simulation animal component, a simulation human component, and a simulation vehicle component. Thus, the simulation test component of this embodiment can be used for simulating human test scenarios, vehicle test scenarios, animal test scenarios, and coexistence test scenarios, respectively.

[0111] In another embodiment of the test system provided by the present application, referring to Figure 5 , the simulation test components include:

[0112] The simulation model body is used to simulate the morphology of the object to be identified;

[0113] A temperature control module is arranged on the surface of the simulation model body and is used to simulate the temperature of the object to be identified;

[0114] A driving module is arranged at a movable part of the simulation model body and is used to simulate the action of the object to be identified;

[0115] The remote control module includes a control unit and multiple controlled units; the multiple controlled units of the remote control module are electrically connected to the temperature control module and the drive module respectively; the control unit of the remote control module is communicatively connected to the multiple controlled units of the remote control module; the control unit of the remote control module is used to output control instructions when triggered by the user; the control instructions are used to instruct the operation of the temperature control module and / or the operation of the drive module.

[0116] In this embodiment, the temperature control module may include a heating wire and a thermostat for controlling the heating temperature of the heating wire, the drive module may include a motor and a motor drive circuit, and the control unit of the remote control module may include a remote control handle, and a Bluetooth antenna is provided on the remote control handle. The assembly steps of the simulation test component may specifically be:

[0117] Step S1001, setting a heating wire to be wound around the simulation model body.

[0118] It should be noted that if the real test component is a simulated animal component, the shape of the simulated model body can be an animal model such as a cat, dog, or chicken. If the real test component is a simulated vehicle component, the shape of the simulated model body can be a vehicle model such as a car, truck, or motorcycle. If the real test component is a simulated human component, the shape of the simulated model body can be a human model such as an adult, a child, or an elderly person.

[0119] Step S1002, a temperature controller is set on the simulation model body to raise the temperature of the heating wire to the surface temperature of the object to be identified. The heating wire can be spread all over the simulation model body.

[0120] It should be noted that, if the real test component is a simulated animal component, after step S1002, the surface of the animal model may be covered with fur of the corresponding animal color and fixed to simulate the animal's form more realistically.

[0121] Step S1003, setting a motor driving circuit on the simulation model body, and setting a motor on the movable part of the simulation model body, so that the motor can drive the movable part of the simulation model body to move; wherein the motor driving circuit is used to drive the motor to work.

[0122] Step S1004, setting a Bluetooth controlled part on the simulation model body so that the operation of the temperature controller and the motor drive circuit can be controlled by the remote control handle; wherein, the Bluetooth controlled part may include a control circuit and a Bluetooth antenna, and the control circuit may convert the Bluetooth signal (i.e., control instruction) received by the Bluetooth antenna into a control signal output to the temperature controller and the motor drive circuit.

[0123] It should be noted that the thermostat can also feedback temperature information, and the motor drive circuit can also feedback drive information. The feedback temperature information and / or drive information can be sent to the remote control handle through the Bluetooth controlled part, and the temperature information and / or drive information is displayed on the remote control handle. An action control button can be provided on the handle. When the action control button is triggered by the tester, an action control instruction is output to control the direction and speed of motor rotation to control the direction and speed of movement of the movable parts of the simulation model body. A temperature control button can also be provided on the handle. When the temperature control button is triggered by the tester, a temperature control instruction is output to control the heating wire to heat to a corresponding temperature. Among them, the temperature control button can realize stepless adjustment of temperature control to more accurately simulate the temperature of the identified object.

[0124] It should be noted that the simulation of the simulation test scene can include three modules: temperature simulation, motion control, and information feedback. When simulating a human test scene, for example, a scene of a pedestrian crossing the street can be simulated. The tester can adjust the heating temperature of the heating wire through the temperature control button on the handle to simulate the body temperature of the human body. For example, the heating wire can be set to heat to about 36 degrees Celsius to 37 degrees Celsius to simulate the body temperature of a normal person. The tester can control the motor through the motion control button on the handle to control the action of the simulation model. Among them, by adjusting the speed and direction of the motor, the simulated "pedestrian" can walk in the simulation environment to simulate the process of crossing the road. The thermostat can feedback temperature information in real time, and the motor drive circuit can feedback the status of the motor. This information is transmitted to the remote control handle via Bluetooth and displayed on the handle, allowing the tester to monitor the human test scene at any time.

[0125] When simulating vehicle test scenarios, for example, a car can be simulated driving on the road. The tester can use the temperature control button to adjust the heating wire to the typical temperature of the engine or other vehicle components, such as the engine temperature is approximately between 80 degrees Celsius and 100 degrees Celsius. The tester can use the motion control button to adjust the speed and direction of the motor to simulate vehicle acceleration, deceleration or turning. The information fed back by the motor drive circuit can help the tester understand the operating status of the vehicle, such as the current speed and steering angle, which is also transmitted back to the handle via Bluetooth and displayed.

[0126] When simulating vehicle test scenarios, for example, you can simulate the activities of animals in the wild, such as a deer foraging in the forest. Through the temperature control button, you can set the heating wire to heat to the animal's body temperature range, for example, the body temperature of a deer is about 38 degrees Celsius to 39 degrees Celsius, and use the motion control button to control the motor to simulate the deer's walking, running or stopping. The feedback information from the thermostat and motor drive circuit can help the tester monitor the activity status and temperature changes of the simulated animal.

[0127] When simulating coexistence test scenarios, for example, you can simulate the interaction between pedestrians, vehicles, and animals in a park. At this time, it is necessary to set different temperatures of multiple heating wires at the same time to simulate the temperature characteristics of different objects (such as people, vehicles, and animals). Testers can control different motors separately through multiple action control buttons to simulate the actions of different objects, such as people walking, vehicles driving, and animals moving. The thermostat and motor drive circuit of each object will feedback information, which is transmitted to the handle via Bluetooth to display the status of all objects, allowing testers to monitor the dynamic changes of the entire simulation test scenario.

[0128] In this way, in the simulation test scenario, the simulation test component can become an animal model, a human model or a vehicle model with temperature, recognition and motion, and can simulate various objects to be identified in the real environment. In a variety of test scenarios, the tester can control the action of the simulation test component through the remote control handle, so that the simulation test component enters the sensing range of the passive infrared sensor of the camera module, and the control device determines that there is a simulation test component in the simulation test scene according to the temperature change information in the simulation test scene, and then controls the camera component to collect image data of the simulation test scene. The control device obtains the image data, inputs it into the preset artificial intelligence recognition algorithm, identifies the simulation test component, and classifies it according to the recognition result. The above classification steps are performed for the simulation test scenes of various classifications, and then the storage partitions classified according to the image data in different simulation test scenes are matched with the preset partitions of the corresponding simulation test scenes, and the image recognition results of the preset artificial intelligence recognition algorithm are obtained, and then the image recognition accuracy of the preset artificial intelligence recognition algorithm is tested. In this way, the present embodiment can provide a simulation test component to simulate a variety of test scenarios. In this way, the present application can use simulation test components to simulate higher-cost objects or targets with lower frequency of occurrence for testing, thereby solving the problem of insufficient test data diversity in existing testing methods, and can obtain image recognition results of preset artificial intelligence recognition algorithms in multiple classification simulation test scenarios, thereby improving the ability to evaluate the accuracy of artificial intelligence recognition algorithms in image recognition systems.

[0129] The above descriptions are only some embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structural changes made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect applications in other related technical fields are included in the patent protection scope of the present application.

Claims

1. An image recognition test method, characterized in that: The steps of the method include: S10, obtaining image data in the current simulation test scenario; S20, inputting the image data into a preset artificial intelligence recognition algorithm to obtain a recognition result of the preset artificial intelligence recognition algorithm for the current simulation test scene; S30, determining the classification of the current simulation test scene according to the recognition result, and classifying the image data into a storage partition of the corresponding class; S40, for various types of simulation test scenarios, respectively execute steps S10 to S30; S50. Based on the preset partitions of multiple classified simulation test scenarios, the image recognition results of the preset artificial intelligence recognition algorithm are obtained according to whether the storage partitions classified by the image data under different simulation test scenarios match the preset partitions of the corresponding simulation test scenarios; wherein the preset partitions are used to indicate the storage partitions of the preset classification of the image data under different simulation test scenarios.

2. The method according to claim 1, characterized in that The classification of simulation test scenes includes human test scenes, vehicle test scenes, animal test scenes and coexistence test scenes; wherein the human test scenes are used to test human images, the vehicle test scenes are used to test vehicle images, the animal test scenes are used to test animal images, and the coexistence test scenes are used to test at least two types of coexistence images of humans, vehicles and animals; the storage partitions include human image storage partitions, vehicle image storage partitions, animal image storage partitions and coexistence image storage partitions; The step S30 comprises: According to the recognition result, if it is determined that the current simulation test scene is classified as a human test scene, the image data is classified into a human image storage partition; If it is determined that the current simulation test scene is classified as a vehicle test scene, the image data is classified into a vehicle image storage partition; If it is determined that the current simulation test scene is classified as an animal test scene, the image data is classified into an animal image storage partition; If it is determined that the current simulation test scene is classified as a coexistence test scene, the image data is classified into a coexistence image storage partition.

3. The method according to claim 2, characterized in that The step S40 comprises: For the human test scene, the vehicle test scene, the animal test scene and the coexistence test scene, steps S10 to S30 are respectively performed.

4. The method according to claim 3, characterized in that The image recognition results include person image recognition results, vehicle image recognition results, animal image recognition results and coexistence image recognition results; the step S50 includes: Based on the preset partitions of the simulation test scenes with multiple classifications, the storage partitions classified according to the image data in the human test scene match the preset classified human image storage partitions, and the human image recognition results of the preset artificial intelligence recognition algorithm are obtained; According to whether the storage partition of the image data in the vehicle test scenario matches the storage partition of the preset classified vehicle image, the vehicle image recognition result of the preset artificial intelligence recognition algorithm is obtained; According to whether the storage partition of the image data in the animal test scene matches the preset classified animal image storage partition, the animal image recognition result of the preset artificial intelligence recognition algorithm is obtained; According to whether the storage partition of the image data in the coexistence test scenario matches the storage partition of the coexistence image of the preset classification, the coexistence image recognition result of the preset artificial intelligence recognition algorithm is obtained.

5. A control device, characterized in that: The control device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the image recognition test method according to any one of claims 1 to 4.

6. A test camera, characterized in that: The test camera comprises the control device as claimed in claim 5, and further comprises a camera module, wherein the camera module is used to collect image data of simulation test scenes of different categories and input the image data into the control device.

7. The test camera according to claim 6, characterized in that: The camera module comprises: A passive infrared sensor, the passive infrared sensor is used to obtain temperature change information in a simulation test scenario and input the temperature change information into the control device; A camera assembly, the camera assembly is used to collect image data of the simulation test scene and input the image data to the control device; The control device is also used to control the camera module to work when it is determined that there is an object to be identified in the simulation test scene according to the temperature change information; wherein the object to be identified includes at least one of a person, a vehicle and an animal.

8. A testing system, characterized in that: The test system comprises the test camera as described in any one of claims 6 to 7, and further comprises a simulation test component, wherein the simulation test component is used to construct multiple classifications of simulation test scenes.

9. The test system according to claim 8, characterized in that: The simulation test component includes at least one of a simulation animal component, a simulation human component and a simulation vehicle component.

10. The test system according to claim 9, characterized in that: The simulation test component comprises: The simulation model body is used to simulate the morphology of the object to be identified; A temperature control module, disposed on the surface of the simulation model body, for simulating the temperature of the object to be identified; A driving module, arranged at an actionable part of the simulation model body, for simulating the action of the object to be identified; The remote control module comprises a control unit and a plurality of controlled units; the plurality of controlled units of the remote control module are electrically connected to the temperature control module and the driving module respectively; the control unit of the remote control module is communicatively connected to the plurality of controlled units of the remote control module; the control unit of the remote control module is used to output control instructions when triggered by a user; the control instructions are used to instruct the operation of the temperature control module and / or the operation of the driving module.

Images