Artificial intelligence detection system

Abstract

The invention relates to an artificial intelligence detection system, and the system comprises an end side device which is provided with a detection device; the detection device is used for shooting a target object to obtain a detection image of the target object and sending the detection image to the first communication module; the first communication module is used for sending the detection image to a second communication module in the cloud server; the second processor is used for analyzing and processing the detection image to obtain monitoring state description information of the target object; and the second communication module is used for sending the monitoring state description information of the target object to the end side equipment. According to the technical scheme, the defect that the dynamic information of the target object cannot be recognized in the related technology is overcome. And the storage and analysis advantages of the cloud server can be brought into play, and the defect that the operation of a large-scale detection and recognition model is not facilitated because the storage space in the end-side equipment is relatively small is solved.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, and in particular to an artificial intelligence detection system. Background Art

[0002] Artificial intelligence technology is developing very rapidly. In related technologies, only static objects can be detected and identified, but dynamic objects cannot be detected and identified, and timely and accurate predictions cannot be achieved. Summary of the invention

[0003] In order to overcome the problems existing in the related art, the present disclosure provides an artificial intelligence detection system to solve the above-mentioned problems.

[0004] According to a first aspect of an embodiment of the present disclosure, there is provided an artificial intelligence detection system, including: an end-side device, on which a detection apparatus is provided.

[0005] The terminal side device is provided with a first processor, a first communication module and a detection device respectively connected to the first processor.

[0006] The detection device is used to photograph the target object, obtain a detection image of the target object, and send the detection image to the first processor;

[0007] The detection device is used to photograph the target object, obtain a detection image of the target object, and send the detection image to the first communication module;

[0008] The first communication module is used to send the detection image to the second communication module in the cloud server;

[0009] The cloud server is also provided with a second processor;

[0010] The second processor is used to analyze and process the detection image to obtain monitoring status description information of the target object; the monitoring status description information is dynamic information describing the target object.

[0011] The second communication module is used to send the monitoring status description information of the target object to the terminal side device.

[0012] In some embodiments, a model training module is provided in the second processor for training the motion detection and recognition model to obtain a trained motion detection and recognition model.

[0013] In some embodiments, a parameter identification module is provided in the second processor, and the parameter identification module is used to call a trained action detection and recognition model to identify the detection image and obtain attribute parameters.

[0014] In some embodiments, a second memory is provided in the cloud server, and the second memory is used to store the trained motion detection and recognition model, as well as sample images and the detection images.

[0015] In some embodiments, the detection device includes a camera, or a depth sensor.

[0016] In some embodiments, the monitoring status description information includes: location information and posture information of the target object.

[0017] In some embodiments, the attribute parameters include at least one of the following: spatial position, posture, and volume.

[0018] In some embodiments, a trajectory generation module is provided in the second processor, and the trajectory generation module is used to generate a motion trajectory of the target object according to the position information of the target object within a historical time period.

[0019] In some embodiments, the second communication module is further used to send the motion trajectory of the target object to the terminal side device.

[0020] In some embodiments, a motion control module is further provided in the terminal-side device, and the motion control module is used to control the action of the related object according to the monitoring state description information of the target object.

[0021] The technical solution provided by the embodiments of the present disclosure may have the following beneficial effects:

[0022] The above-mentioned technical solution of the present application analyzes and processes the detection image in the cloud server to obtain monitoring status description information; the monitoring status description information is the dynamic information describing the target object. This solves the shortcoming of the related art that the dynamic information of the target object cannot be identified. It is also conducive to giving full play to the storage and analysis advantages of the cloud server, and solves the shortcoming that the storage space in the terminal device is relatively small, which is not conducive to the operation of large-scale detection and recognition models.

[0023] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0025] Figure 1 The figure is a schematic structural diagram of an artificial intelligence detection system according to an exemplary embodiment.

[0026] Figure 2 is a flow chart of a method for artificial intelligence detection according to an exemplary embodiment;

[0027] Figure 3 is a flow chart of a motion trajectory processing method according to an exemplary embodiment;

[0028] Figure 4 is a flow chart of a method for training and using an action detection and recognition module according to an exemplary embodiment;

[0029] Figure 5 The figure is a signal flow chart according to an exemplary embodiment.

[0030] Description of Reference Numerals

[0031] 1-end-side device; 11-electronic device; 2-cloud server; 21-first processor; 22-first memory; 211-model training module; 212-parameter identification module; 213-first communication module; 214-motion control module; 215-second communication module; 216-trajectory generation module; 31-second processor; 32-second memory. DETAILED DESCRIPTION

[0032] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0033] See attached Figure 1 , the present application proposes an artificial intelligence detection system, comprising: a terminal side device 1, on which a detection device is provided.

[0034] The terminal device may be a mobile phone, a computer or a mixed reality (MR) device.

[0035] The terminal device is provided with a first processor 21, a first communication module 213 and a detection device 3 respectively connected to the first processor 21.

[0036] The detection device 3 is used to photograph the target object, obtain a detection image of the target object, and send the detection image to the first processor 21.

[0037] In this embodiment, a first processor 21 and a first memory 22 may be provided in the terminal device, and the detection device 3 and the first communication module 213 are respectively connected to the first processor 21. The first processor 21 controls the detection device 3 to perform actions, and the first communication module 213 to perform communication actions.

[0038] Wherein, the detection device 3 includes a camera or a depth sensor.

[0039] The camera or the depth sensor can photograph the target object under the control of the first processor 21 to obtain a detection image.

[0040] The first communication module 213 is used to send the detection image to the second communication module 215 in the cloud server 2.

[0041] In this embodiment, the first communication module 213, under the control of the first processor 21, sends the detection image to the second communication module 215 in the cloud server.

[0042] The cloud server 2 is also provided with a second processor 31 .

[0043] The second processor 31 is used to analyze and process the detection image to obtain monitoring status description information of the target object.

[0044] Wherein, the monitoring status description information includes: the position information and posture information of the target object.

[0045] The second communication module 215 is used to send the monitoring status description information of the target object to the terminal side device.

[0046] In the above technical solution of the present application, the detection device 3 takes a picture of the target object to obtain a detection image of the target object; the first communication module 213 sends the detection image to the cloud server 2. The detection image is analyzed and processed in the cloud server 2 to obtain the monitoring status description information of the target object. It is beneficial to give full play to the storage and analysis advantages of the cloud server 2. The cloud server 2 has a large-capacity storage space and is provided with a second memory 32, which can store a relatively large-scale detection and recognition model, solving the shortcoming that the storage space in the terminal device is relatively small and is not conducive to the operation of large-scale detection and recognition models.

[0047] In some embodiments, the second processor 21 is provided with a model training module 211 for training the motion detection and recognition model to obtain a trained motion detection and recognition model.

[0048] In this embodiment, when training the motion detection and recognition model, a large number of test sample images can be collected, the test sample images can be divided into standard samples and test samples, the test samples can be input into the motion detection and recognition model, the output results can be compared with the standard samples to determine whether the output results are accurate, and the parameters in the motion detection and recognition model can be adjusted so that the output results are close to the standard samples. In this way, after multiple adjustment tests, for example, thousands of tests, a trained motion detection and recognition model can be obtained.

[0049] In some embodiments, the second processor is provided with a graphics card 33, and the graphics card 33 is provided with a parameter identification module 212, and the parameter identification module 212 is used to call a trained action detection and recognition model to identify the detection image and obtain attribute parameters.

[0050] The attribute parameters include at least one of the following: spatial position, posture and volume.

[0051] In this embodiment, a trained motion detection and recognition model is called to recognize the detection image, so that the spatial position can be determined.

[0052] In this embodiment, a trained motion detection and recognition model is called to recognize the detection image to determine the posture.

[0053] In this embodiment, a trained action detection and recognition model is first required. The action detection and recognition model may be built based on a deep learning architecture (such as a convolutional neural network CNN, a recurrent neural network RNN, or a long short-term memory network LSTM, etc.). The training process usually involves a large amount of labeled data, that is, images or videos containing various actions or postures, and corresponding labels. These data are used to train the model so that it can learn the characteristics and patterns of different actions. After the model training is completed, it can be used to identify new detection images. This usually involves inputting the image into the model, and then the model outputs one or more recognition results, indicating the actions or postures that may exist in the image.

[0054] Based on the output of the model, the specific pose exhibited in the image can be determined. This may involve further processing or interpretation of the output, such as matching the output against a library of predefined poses to find the best match.

[0055] In this embodiment, the volume of the target object can be determined by calling a trained motion detection and recognition model to recognize the detection image.

[0056] In this embodiment, first, a large amount of image data needs to be collected, which should contain target objects of different angles, different lighting conditions and different sizes. The collected images are annotated to clarify the position of the target object in each image and possible volume information. Use the annotated data to train the action detection and recognition model. This action detection and recognition model can be based on deep learning, such as a convolutional neural network (CNN) or a regional convolutional neural network (R-CNN). The trained action detection and recognition model is deployed to a cloud server. The image to be detected is input into the deployed action detection and recognition model. The trained action detection and recognition model will identify the target object in the image and output its location information (such as a bounding box).

[0057] If the action detection recognition model only provides a two-dimensional bounding box, then additional information is needed to calculate the volume. This can be done by obtaining the depth information of the target object through a depth sensor (such as a lidar or a depth camera). Using the depth information and the two-dimensional bounding box, a three-dimensional reconstruction can be performed to obtain a three-dimensional model of the target object. Based on the three-dimensional model, the volume of the target object can be calculated. This usually involves geometric calculations, such as calculating the volume of a polyhedron. The calculated volume information is output to the user or used for subsequent processing. It should be noted that each step in the above process may be affected by a variety of factors, such as image quality, lighting conditions, shape and size of the target object, etc. Therefore, in practical applications, the model may need to be tuned to improve the accuracy of recognition and the accuracy of volume calculation. In addition, if the shape of the target object is irregular or the volume calculation requires high accuracy, more complex algorithms and techniques may need to be used, such as three-dimensional scanning, point cloud processing, etc. In general, by calling a trained action detection recognition model and combining depth information and geometric calculations, the volume of the target object in the detection image can be determined. However, this process requires comprehensive consideration of multiple factors and may require the use of multiple technologies and methods.

[0058] In some embodiments, the cloud server 2 is provided with a second memory 32, and the second memory 32 is used to store the trained motion detection and recognition model, as well as the sample images and the detection images.

[0059] In this embodiment, the trained action detection and recognition model is one of the core contents stored in the second memory 32. The action detection and recognition model is obtained through a large amount of data training and can recognize actions or postures in images or videos. The trained action detection and recognition model file is usually large and contains information such as parameters and structure of the model, so it needs to be stored in a reliable storage device. When the cloud server needs to perform an action detection and recognition task, it will load the trained action detection and recognition model from the second memory 32 and perform reasoning operations on it. Sample images are image data used to train the action detection and recognition model. These images usually contain examples of various actions or postures with corresponding labels or annotations. In the model training stage, the cloud server 2 will read these sample images from the second memory 32 and use them to train the action detection and recognition model. In addition, the sample images can also be used for verification and testing of the action detection and recognition model to ensure the accuracy and reliability of the action detection and recognition model. The detection images are image data to be processed, which are input into the action detection and recognition model for recognition. These images may come from different sources, such as cameras, image databases, etc. During the processing, the cloud server reads the detection image from the second storage and processes it using the trained motion detection recognition model to identify the action or posture in the image. In order to ensure the security and reliability of the data, the second storage 32 usually implements a data backup strategy. This means that important data (such as motion detection recognition model files and image data) will be regularly copied to other storage devices or cloud storage to prevent data loss or damage. In order to protect the privacy and security of the data, the cloud server 2 will strictly control the data access to the second storage 32. Only authorized users or applications can access and modify these data. In order to improve storage efficiency and performance, the second storage 32 may adopt various storage optimization technologies, such as data compression, deduplication, hierarchical storage, etc. These technologies can reduce the occupancy of storage space and improve the reading and writing speed of data. The second storage plays a vital role in the cloud server. It stores important data resources such as trained motion detection recognition models, sample images, and detection images, providing strong support for the various functions and services of the cloud server.

[0060] In some embodiments, the second processor 31 is provided with a trajectory generation module 216, and the trajectory generation module 216 is used to generate a motion trajectory of the target object according to the position information of the target object within a historical time period.

[0061] In this embodiment, the trajectory generation module 216 first collects the location information of the target object in the historical time period from the data source (such as sensors, cameras, GPS systems, etc.). This information usually includes a timestamp and corresponding location coordinates (such as latitude and longitude, X / Y / Z coordinates, etc.). The preprocessing steps may include data cleaning (removing erroneous or abnormal data), data smoothing (reducing noise), and data alignment (ensuring the consistency of timestamps). Using the collected location information, the trajectory generation module 216 connects these points in chronological order to form a continuous path, that is, the motion trajectory of the target object. If the location information is dense and accurate enough, the generated trajectory can be very close to the actual motion path of the target object. The trajectory generation module 216 can apply some interpolation algorithms to fill in the missing data points, or apply filtering algorithms to smooth the trajectory to make it more continuous and accurate. The generated trajectory can be represented in a variety of forms, such as line graphs, polygons, trajectory point lists, etc. The trajectory data is usually stored in a database for subsequent analysis, query and visualization. The generated trajectory can be used in a variety of application scenarios, such as behavior analysis, path planning, anomaly detection, etc. The trajectory generation module 216 may also be integrated with other modules (such as a behavior recognition module, a prediction module, etc.) to achieve more advanced functions.

[0062] In some embodiments, the second communication module 215 is further configured to send the motion trajectory of the target object to the terminal device.

[0063] In some embodiments, a motion control module 214 is further provided in the terminal device 1, and the motion control module 214 is used to control the action of the related object according to the monitoring state description information of the target object.

[0064] In this embodiment, the terminal device 1 is configured with a motion control module 214. The function of this module is to control and drive related objects to perform actions according to the monitoring state description information of the target object.

[0065] The following is a detailed description of this embodiment:

[0066] The motion control module 214 is a key component in the terminal device 1, which is responsible for formulating and executing the control strategy according to the monitoring state description information of the target object obtained from the monitoring system. The monitoring state description information may include physical parameters such as the position, speed, acceleration, temperature, humidity, pressure, etc. of the target object, as well as possible other state information, such as switch state, working mode, etc.

[0067] The end-side device 1 collects monitoring data of the target object through its built-in sensors or external sensor networks. The collected data is sent to the device's processor for preprocessing and analysis to extract key information related to the state of the target object. The processor converts the processed data into monitoring status description information of the target object, which is usually expressed in a structured data format (such as JSON, XML, etc.). The motion control module 214 formulates corresponding control instructions based on the received monitoring status description information and the preset control strategy or algorithm. The control instructions are sent to related mechanical equipment (such as motors, cylinders, valves, etc.) to drive them to act in a specified manner.

[0068] Since the end-side devices 1 can directly interact with the environment or target objects, they can obtain and respond to changes in the monitoring state in real time. The motion control module 214 can formulate and execute different control strategies according to different application scenarios and requirements. Through automated and intelligent control, work efficiency can be significantly improved and labor costs can be reduced. This end-side device equipped with the motion control module 214 has broad application prospects and important practical value in many fields.

[0069] In some embodiments, based on the above embodiments, combined with virtual object interaction, and applied to multiple people, multiple virtual people or moving scenes, the following examples can be used for illustration:

[0070] In a multi-person collaboration scenario, such as a remote collaborative engineering project, multiple team members may be located in different geographical locations. In this case, a shared virtual environment can be set up, which contains various virtual objects and tools to simulate the real engineering project scenario.

[0071] Role setting: Each team member has a corresponding virtual human role in the virtual environment, which can reflect their actions and status in the real world in real time.

[0072] Motion control: Through the motion control module 214, team members can control their virtual human characters to move, rotate and interact in the virtual environment. For example, they can use devices such as keyboards, mice or handles to operate their virtual human characters.

[0073] Virtual object interaction: In the virtual environment, team members can interact with various virtual objects. For example, they can click or drag virtual tools to perform design or modeling work, or interact with virtual engineering project elements to view or modify their properties.

[0074] Real-time collaboration: Through a shared virtual environment and real-time communication functions, team members can see the actions and status of other team members in real time and collaborate. For example, they can discuss design plans together, complete tasks together, or provide real-time technical support.

[0075] Multi-virtual human interaction scenario,In a multi-virtual human interaction scenario, such as a virtual social gathering or game scenario, multiple virtual human characters can interact in a shared virtual environment.

[0076] Virtual human characters, each participant has a unique virtual human character, which has different appearance, movements and interaction capabilities.

[0077] Motion control: Participants can use the motion control module 214 to control their virtual human characters to move, jump, rotate and perform various actions in the virtual environment.

[0078] Virtual objects interact with the environment. Virtual human characters can interact with various objects in the virtual environment. For example, they can pick up or put down items, open or close doors, and talk to virtual NPCs (non-player characters).

[0079] Social interaction: Through virtual social functions and real-time communication functions, participants can interact with other virtual characters. For example, they can greet each other, chat, share experiences or play games.

[0080] Moving virtual object interaction scenarios,In moving virtual object interaction scenarios, such as an augmented reality (AR) navigation or training application,,virtual objects can be updated and interacted in real time as the user moves.

[0081] Location tracking: Through sensors on the device (such as GPS, camera, etc.), the system can track the user's location and orientation in real time.

[0082] Virtual object generation: Based on the user's position and orientation, the system can generate corresponding virtual objects in the user's field of view. For example, in an AR navigation application, the system can display virtual route instructions and landmarks in the user's field of view.

[0083] Motion control: users can control the movement and interaction of virtual objects through the motion control module 214 on the device (such as a touch screen, a handle, etc.). For example, in an AR training application, users can interact with a virtual training assistant through gestures.

[0084] Real-time updates: As users move and interact, virtual objects and scenes can be updated and changed in real time. For example, in an AR navigation application, as users move, virtual route instructions and landmarks will also be updated accordingly.

[0085] These examples show the application scenarios and possibilities of the embodiments combined with virtual object interaction in multiple people, multiple virtual people or in motion. By combining the motion control module 214 and virtual object interaction technology, a richer, more interactive and immersive user experience can be created.

[0086] The present application relates to a technical solution combining MR and AI, the purpose of which is to solve the motion trajectory of objects in real-time dynamic situations, and then respond to each other and interact with other virtual objects. It is applied to real-time dynamic trajectory capture of multiple people, multiple virtual people or moving objects (not limited to people).

[0087] See attached Figure 2 , the present application also discloses an artificial intelligence detection method, comprising the following steps:

[0088] In step S202 , the detection device 3 photographs the target object to obtain a detection image of the target object, and sends the detection image to the first processor 21 .

[0089] In step S204, the first communication module 213 sends the detection image to the second communication module 215 in the cloud server.

[0090] In step S206 , the second processor 31 analyzes and processes the detection image to obtain monitoring state description information of the target object; the monitoring state description information is dynamic information describing the target object.

[0091] In step S208 , the second communication module 215 sends the monitoring status description information of the target object to the terminal device 1 .

[0092] In step S210 , the motion control module 214 controls the motion of the related object according to the monitoring state description information of the target object.

[0093] See attached Figure 3 In some embodiments, the method may further include the following steps:

[0094] In step S302, the trajectory generation module generates a motion trajectory of the target object according to the position information of the target object within the historical time period.

[0095] In step S304 , the second communication module 215 sends the above-mentioned motion trajectory to the first communication module 213 .

[0096] See attached Figure 4 In some embodiments, the method may further include the following steps:

[0097] In step S402, the model training module 211 trains the action detection and recognition model to obtain a trained action detection and recognition model.

[0098] In step S404, the second memory 32 stores the trained motion detection and recognition model, the sample image, and the detection image.

[0099] In step S406, the parameter recognition module 212 calls the trained motion detection recognition model to recognize the detection image and obtain attribute parameters.

[0100] See attached Figure 5 , the detection device 3 sends the detection image to the first processor 21, the first communication module 213 sends the detection image to the second communication module 215, and the parameter identification module 212 analyzes the detection image to obtain parameters, wherein the parameters include the monitoring state description information of the target object. The trajectory generation module 216 can generate a trajectory according to the position in the parameter and send the trajectory to the second communication module 215; the model training module 211 analyzes and trains the parameter identification module 212, and the second communication module 215 can send the monitoring state description information of the target object and the trajectory to the first communication module 213.

[0101] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the present disclosure. This application is intended to cover any variations, uses or adaptations of the present disclosure, which follow the general principles of the present disclosure and include common knowledge or customary technical means in the art that are not disclosed in the present disclosure. The specification and examples are to be considered as exemplary only, and the true scope and spirit of the present disclosure are indicated by the following claims.

[0102] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. An artificial intelligence detection system, characterized in that: include: A terminal side device, wherein the terminal side device is provided with a detection device; The terminal device is provided with a first processor, and a first communication module and a detection device respectively connected to the first processor; The detection device is used to photograph the target object, obtain a detection image of the target object, and send the detection image to the first processor; The first communication module is used to send the detection image to the second communication module in the cloud server; The cloud server is also provided with a second processor; The second processor is used to analyze and process the detection image to obtain monitoring state description information of the target object; the monitoring state description information is dynamic information describing the target object; The second communication module is used to send the monitoring status description information of the target object to the terminal side device.

2. The artificial intelligence detection system according to claim 1, characterized in that: The second processor is provided with a model training module for training the motion detection and recognition model to obtain a trained motion detection and recognition model.

3. The artificial intelligence detection system according to claim 1, characterized in that: The second processor is provided with a parameter recognition module, and the parameter recognition module is used to call a trained motion detection recognition model to recognize the detection image and obtain attribute parameters.

4. The artificial intelligence detection system according to claim 3, characterized in that: The cloud server is provided with a second memory, which is used to store the trained motion detection and recognition model, sample images, and the detection image.

5. The artificial intelligence detection system according to claim 1, characterized in that: The detection device includes a camera or a depth sensor.

6. The artificial intelligence detection system according to claim 1, characterized in that: The monitoring status description information includes: position information and posture information of the target object.

7. The artificial intelligence detection system according to claim 3, characterized in that: The attribute parameter includes at least one of the following: spatial position, posture and volume.

8. The artificial intelligence detection system according to claim 1, characterized in that: The second processor is provided with a trajectory generation module, and the trajectory generation module is used to generate a motion trajectory of the target object according to the position information of the target object within a historical time period.

9. The artificial intelligence detection system according to claim 8, characterized in that: The second communication module is further configured to send the motion trajectory of the target object to the terminal side device.

10. The artificial intelligence detection system according to claim 1, characterized in that: The terminal side device is also provided with a motion control module, and the motion control module is used to control the action of the related object according to the monitoring state description information of the target object.

Images