An Edge Computing-Based Autonomous Object Recognition Method and Device

Abstract

The present invention discloses an autonomous object recognition method and device based on edge computing, which relates to the technical field of object recognition. The performance of the edge computing device is detected. When it is detected that the edge computing device has a fault or performance degradation, the status of the edge computing device is first recorded, and then the preprocessing of the image data is forwarded to the control center based on the wireless network for processing. After combining the status of the edge computing device with the object recognition effect, a maintenance score is generated for each recognition device, and all recognition devices are sorted according to the maintenance score to generate a device list, and the maintenance order of all recognition devices is selected based on the device list. When analyzing the fault or performance degradation of the edge computing device, the recognition method sends the data to the control center for processing, ensuring the data processing efficiency and processing effect, and regularly generating a management list by combining the status of the edge computing device and the data recognition result, so that the recognition devices with poor overall status can be managed preferentially, improving the management efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of object recognition, and particularly to an autonomous object recognition method and device based on edge computing. Background Art

[0002] Edge computing is a technology that transfers computing and data processing capabilities from the cloud center to the network edge. It enables data to be processed closer to the data source, aiming to reduce data transmission latency, improve processing speed, save bandwidth, and enhance data privacy. Autonomous object recognition devices based on edge computing utilize the advantages of this distributed computing to perform real-time object recognition.

[0003] The prior art has the following deficiencies:

[0004] 1. Recognition devices adopting the edge computing method usually rely solely on edge computing devices for data processing. However, in actual applications, when the performance of the edge computing device deteriorates due to an anomaly (at this time, the edge computing device is not in a faulty state and can still be used), the recognition device cannot detect such an anomaly, and continuing to use the edge computing device cannot guarantee data processing efficiency and processing effect;

[0005] 2. The methods for maintaining and managing recognition devices are generally divided into two types. The first type is that when a recognition device fails, after the recognition device sends a warning signal to the control center, the control center sends a maintenance signal to the maintenance personnel. The second type is that the control center pre-sets a maintenance cycle, and when the maintenance date arrives, randomly maintains and manages all recognition devices in the control center. Neither of the above two methods can preferentially manage recognition devices with overall anomalies, reducing the management efficiency of recognition devices.

[0006] Based on this, the present invention proposes an autonomous object recognition method and device based on edge computing, which first analyzes the performance of the edge computing device during operation. When analyzing that the edge computing device fails or its performance deteriorates, the data is sent to the control center for processing to ensure data processing efficiency and processing effect, and regularly generates a management list in combination with the status of the edge computing device and the data recognition result, so as to be able to preferentially manage recognition devices with poor overall status and improve management efficiency. Summary of the Invention

[0007] The purpose of the present invention is to provide an autonomous object recognition method and device based on edge computing to solve the deficiencies in the background art.

[0008] To achieve the above purpose, the present invention provides the following technical solution: An autonomous object recognition method based on edge computing, the recognition method comprising the following steps:

[0009] When an object appears in the monitoring area of the recognition device, the recognition device automatically collects image data and performs a performance detection on the edge computing device;

[0010] When it is detected that the edge computing device has a fault or its performance has declined, record the status of the edge computing device, and then forward the preprocessed image data to the control center via a wireless network for preprocessing by the control center. After the control center preprocesses the image data, it is sent to the recognition device for object recognition;

[0011] After the object recognition is completed this time, the control center evaluates the object recognition effect of the recognition device, combines the status of the edge computing device with the object recognition effect to generate a maintenance score for each recognition device, sorts all the recognition devices according to the maintenance score to generate a device list, and selects the maintenance order of all the recognition devices based on the device list.

[0012] In a preferred embodiment, the performance detection of the edge computing device includes the following steps:

[0013] Obtain the power loss coefficient and the Fourier transform coefficient of heat conduction of the edge computing device;

[0014] Comprehensively calculate the power loss coefficient and the Fourier transform coefficient of heat conduction to obtain a device status index;

[0015] Compare the obtained device status index with a preset first anomaly threshold and a second anomaly threshold. The first anomaly threshold is used to determine whether the performance of the edge computing device has declined, and the second anomaly threshold is used to determine whether the edge computing device has a fault;

[0016] If the device status index is less than or equal to the first anomaly threshold, it is determined that the performance of the edge computing device has not declined. If the device status index is greater than the first anomaly threshold and less than or equal to the second anomaly threshold, it is determined that the performance of the edge computing device has declined. If the device status index is greater than the second anomaly threshold, it is determined that the edge computing device has a fault.

[0017] In a preferred embodiment, the power loss coefficient and the Fourier transform coefficient of heat conduction are comprehensively calculated to obtain a device status index, and the expression is:

[0018] , where is the device status index, is the power loss coefficient, is the Fourier transform coefficient of heat conduction, , are the proportionality coefficients of the power loss coefficient and the Fourier transform coefficient of heat conduction respectively, and , are both greater than 0.

[0019] In a preferred embodiment, the control center evaluates the object recognition effect of the recognition device, including the following steps:

[0020] The control center obtains the recognition error frequency and the recognition duration deviation during the process of the recognition device recognizing an object, normalizes the recognition error frequency and the recognition duration deviation, maps the value ranges of the recognition error frequency and the recognition duration deviation to between [0, 1], obtains the normalized value of the recognition error frequency and the normalized value of the recognition duration deviation, and sums the normalized value of the recognition error frequency and the normalized value of the recognition duration deviation to obtain the recognition weakening index;

[0021] Compare the obtained recognition weakening index with the effect threshold. The effect threshold is used to evaluate the quality of the object recognition effect of the recognition device. If the recognition weakening index is less than or equal to the effect threshold, it is evaluated that the object recognition effect of the recognition device is excellent. If the recognition weakening index is greater than the effect threshold, it is evaluated that the object recognition effect of the recognition device is poor.

[0022] In a preferred embodiment, sorting all recognition devices according to the maintenance score to generate a device list includes the following steps:

[0023] Combining the edge computing device status and the object recognition effect to generate a maintenance score for each recognition device. The calculation expression is:

[0024] , where in the formula, is the management score, is the device status index, is the recognition weakening index, are the weights of the device status index and the recognition weakening index respectively, and ;

[0025] Sort all recognition devices in descending order according to the management score to generate a device list, select the maintenance order of all recognition devices based on the device list, and send the maintenance order information to the relevant maintenance personnel;

[0026] When there is a recognition device in the device list with a management score greater than the score threshold, the control center generates an emergency maintenance flag for the recognition device and sends it to the relevant maintenance personnel.

[0027] In a preferred embodiment, the acquisition logic of the recognition error frequency is: obtaining the number of recognition errors of the recognition device within the object recognition time period;

[0028] The acquisition logic of the recognition duration deviation is: subtracting the standard recognition duration from the actual recognition duration to obtain the recognition duration deviation.

[0029] In a preferred embodiment, the calculation expression of the power loss coefficient is:

[0030] , where is the power loss coefficient, is the nominal voltage of the device, is the instantaneous voltage, is the monitoring time range;

[0031] The calculation expression of the thermal conduction Fourier transformation coefficient is:

[0032] , where is the temperature value at position and time . In this application, the temperature value at position and time is defined as the thermal conduction Fourier transformation coefficient, is the temperature value at position and time , is the time step, representing the amount of time advancement in each iteration, is the space step, representing the interval length of space discretization, is the thermal diffusion coefficient, defined as , where is the thermal conductivity, is the density, is the specific heat capacity, and respectively represent the temperature values at positions and .

[0033] An autonomous object recognition device based on edge computing, including a performance detection module, a data forwarding module, and an object recognition module;

[0034] Performance detection module: When an object appears in the monitoring area of the recognition device, the recognition device automatically collects image data and performs performance detection on the edge computing device;

[0035] Data forwarding module: When it is detected that the edge computing device has a fault or performance degradation, first record the status of the edge computing device, and then forward the preprocessed image data to the management and control center based on a wireless network;

[0036] Object recognition module: Used to perform object recognition on the preprocessed image data, and send the object recognition result to the management and control center.

[0037] In the above technical solution, the technical effects and advantages provided by the present invention:

[0038] The present invention automatically collects image data and performs performance detection on edge computing devices. When a fault or performance degradation is detected in an edge computing device, the status of the edge computing device is first recorded, and then the preprocessed image data is forwarded to a management and control center via a wireless network. After the object recognition is completed this time, the management and control center evaluates the object recognition effect of the recognition device, generates a maintenance score for each recognition device by combining the status of the edge computing device with the object recognition effect, and generates a device list by sorting all the recognition devices according to the maintenance score. The maintenance order of all the recognition devices is selected based on the device list. During the operation of this recognition method, the performance of the edge computing device is first analyzed. When a fault or performance degradation of the edge computing device is analyzed, the data is sent to the management and control center for processing to ensure the data processing efficiency and effect. A management list is regularly generated by combining the status of the edge computing device and the data recognition result, so that the recognition devices with poor overall status can be preferentially managed, improving the management efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings.

[0040] Figure 1 It is a flowchart of the method of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0041] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0042] Embodiment 1: Please refer to Figure 1 As shown, a method for autonomous object recognition based on edge computing in this embodiment includes the following steps:

[0043] When an object appears in the monitoring area of the recognition device, the recognition device automatically collects image data and performs a performance detection on the edge computing device. When it is detected that the edge computing device has a fault or performance degradation, the status of the edge computing device is first recorded, and then the preprocessed image data is forwarded to the management and control center via a wireless network for preprocessing. After the management and control center preprocesses the image data, it is sent to the recognition device for object recognition. After this object recognition is completed, the management and control center evaluates the object recognition effect of the recognition device, combines the status of the edge computing device with the object recognition effect to generate a maintenance score for each recognition device, sorts all the recognition devices according to the maintenance score to generate a device list, selects the maintenance order of all the recognition devices based on the device list, and sends the maintenance order information to the relevant maintenance personnel.

[0044] In this application, image data is automatically collected, and a performance detection is performed on the edge computing device. When it is detected that the edge computing device has a fault or performance degradation, the status of the edge computing device is first recorded, and then the preprocessed image data is forwarded to the management and control center via a wireless network for preprocessing. After this object recognition is completed, the management and control center evaluates the object recognition effect of the recognition device, combines the status of the edge computing device with the object recognition effect to generate a maintenance score for each recognition device, sorts all the recognition devices according to the maintenance score to generate a device list, and selects the maintenance order of all the recognition devices based on the device list. This recognition method first performs a performance analysis on the edge computing device during operation. When analyzing the fault or performance degradation of the edge computing device, the data is sent to the management and control center for processing, ensuring the data processing efficiency and processing effect, and regularly generating a management list by combining the status of the edge computing device and the data recognition result, so as to be able to preferentially manage the recognition devices with poor overall status and improve the management efficiency.

[0045] Embodiment 2: When an object appears in the monitoring area of the recognition device, the recognition device automatically collects image data, including the following steps:

[0046] 1) Object detection and triggering

[0047] Area monitoring: The recognition device continuously monitors the specified area and captures environmental information in real time through sensors (such as cameras, infrared sensors, radars, etc.).

[0048] Motion detection: Use background subtraction, optical flow method or other motion detection algorithms to identify the movement of objects in the monitoring area. Once an object is detected entering or leaving the monitoring area, the image data collection is triggered.

[0049] 2) Image capture

[0050] Automatic focus and exposure adjustment: Before capturing the image, the recognition device automatically adjusts the focus and exposure parameters of the camera to adapt to the current lighting conditions, ensuring that a clear and properly lit image is captured.

[0051] High-frequency sampling: Adjust the frame rate of image capture according to the speed and importance of the moving object. In the case of high-speed movement, the device may capture image data at a higher frame rate.

[0052] 3) Multi-angle and multi-frame acquisition

[0053] Multi-angle capture: If the device is equipped with multiple cameras or has a moving function, it can capture objects from different angles to obtain comprehensive information.

[0054] Multi-frame superposition: Continuously capture multiple frames of images in a short time, and enhance the image clarity and detail performance through image superposition or stitching technology.

[0055] 4) Image data compression and storage

[0056] Data compression: Compress the captured image data to reduce the storage space occupancy and transmission bandwidth requirements. Common compression methods include JPEG, PNG, or video coding (such as H.264, H.265), etc.

[0057] Local storage: Store the processed image data in a local storage device. Especially for application scenarios with high real-time requirements, it can be preferentially saved locally for subsequent analysis.

[0058] 5) Data tagging and classification

[0059] Automatic annotation: Automatically annotate the captured images according to preset rules or using a simple model, such as recording the time, location, environmental conditions, and detected object categories.

[0060] Classification storage: Classify and store the image data according to tags or categories for subsequent rapid retrieval and analysis.

[0061] Perform performance detection on the edge computing device, including the following steps:

[0062] Obtain the power loss coefficient and the Fourier transform coefficient of heat conduction of the edge computing device;

[0063] Comprehensively calculate the power loss coefficient and the Fourier transform coefficient of heat conduction to obtain the device status index, and the expression is: , where is the device status index, is the power loss coefficient, is the Fourier transform coefficient of heat conduction, , are the proportionality coefficients of the power loss coefficient and the Fourier transform coefficient of heat conduction respectively, and , are both greater than 0;

[0064] The larger the device status index, the worse the performance of the edge computing device. Compare the obtained device status index with a preset first anomaly threshold and a second anomaly threshold. The first anomaly threshold is used to determine whether the performance of the edge computing device has declined, and the second anomaly threshold is used to determine whether the edge computing device has failed;

[0065] If the device status index is less than or equal to the first anomaly threshold, it is determined that the performance of the edge computing device has not declined and can continue to be used. If the device status index is greater than the first anomaly threshold and less than or equal to the second anomaly threshold, it is determined that the performance of the edge computing device has declined and it is not supported for use. If the device status index is greater than the second anomaly threshold, it is determined that the edge computing device has failed and immediate maintenance is required.

[0066] The calculation expression of the power loss coefficient is: , where is the power loss coefficient, is the nominal voltage of the device, is the instantaneous voltage, is the monitoring time range.

[0067] The larger the power loss coefficient, the higher the power consumed by the device at a certain moment compared to the normal level, which usually indicates that the device may have anomalies and may lead to performance degradation or failure. The reasons include the following aspects:

[0068] 1) Overload operation and component loss

[0069] When the instantaneous power loss of the device is large, it often means that the device is in an overload operation state. Overload operation causes the internal components of the device to bear higher current and voltage, increasing the heat generation and accelerating the aging of the components. Especially the power supply module, processor, and memory, these core components are sensitive to heat, and long-term high-power operation may cause:

[0070] Overheating: When the device generates too much heat, the cooling system may not be able to effectively discharge the heat, resulting in an increase in the internal temperature, which in turn affects the normal operation of electronic components.

[0071] Insufficient power supply: If the power consumed by the device exceeds the design capacity of the power supply module, the power supply may not be able to provide sufficient and stable power, thus causing problems such as device restart and data loss.

[0072] 2) Current spikes and circuit stress

[0073] The increase in instantaneous power loss is usually accompanied by current spikes. Current spikes can cause significant stress on the circuit and may lead to the following problems:

[0074] Circuit damage: Instantaneous excessive current may damage the device's circuit boards, wires, or other sensitive components, causing partial function failure or complete damage to the device.

[0075] Transient voltage changes: High-power load switching can cause short-term fluctuations in the power supply voltage. This transient voltage change may affect the stability of other circuits or equipment.

[0076] 3) Unstable operating state

[0077] Large instantaneous power fluctuations are often a sign that the device is performing heavy computing, data processing, or other high-load operations. In this case, the device may:

[0078] Excessive resource usage: Critical resources such as CPU and memory are heavily occupied, causing slow system response, and even freezing and restarting.

[0079] Increased electromagnetic interference: Increased power may also increase electromagnetic radiation, affecting other electronic devices inside or outside the device, causing signal interference and data errors.

[0080] 4) System load and failure risk

[0081] When a device is in a high power consumption state for a long time, it means that it is performing heavy computing tasks or processing large amounts of data. This load increases the risk of system errors or crashes, especially when the system's cooling or power supply cannot keep up with the load. Possible risks include:

[0082] Software failure: Excessive load may trigger untested edge conditions, leading to software errors, process deadlocks, and other problems.

[0083] Hardware failure: Persistent high-power states can accelerate hardware aging and failure rates.

[0084] 5) Device anomalies or signs of attack

[0085] Abnormal increases in instantaneous power may also be an early sign of a device being attacked by a cyberattack or hardware failure. For example:

[0086] Malware activity: Malware may consume a large amount of resources for mining, DDoS attacks, etc., which can cause a significant increase in instantaneous power.

[0087] Premonitions of hardware failure: Some hardware failures may manifest as abnormal power consumption before they occur, such as unstable power supply caused by a power module failure.

[0088] The calculation expression of the Fourier variation coefficient of heat conduction is:

[0089] , where is the temperature value at the position and time In this application, the temperature value at the position and time is defined as the Fourier transform coefficient of heat conduction, is the temperature value at the position and time is the time step, representing the amount of time advancement in each iteration, is the space step, representing the interval length of space discretization, is the thermal diffusivity, defined as , where is the thermal conductivity, is the density, is the specific heat capacity, and respectively represent the temperature values at the positions and .

[0090] The larger the Fourier transform coefficient of heat conduction, the more likely it indicates that the edge computing device may have anomalies leading to performance degradation or failures, as follows:

[0091] Overheating problem: The increase in temperature may cause the device to overheat, which in turn affects the performance of core components such as the processor and memory. Electronic devices usually operate within a specific temperature range. Exceeding this range will lead to slower computing speed, more errors, and even permanent damage to the hardware.

[0092] Increased power consumption: A direct cause of the temperature rise in edge computing devices is the increase in power consumption. When the device processes complex tasks or due to system anomalies (such as short circuits, overloads, etc.), the power consumption will rise sharply, generating more heat. This will not only cause overheating but also accelerate the wear of the battery or power supply module, reducing the device life.

[0093] Increased burden on the cooling system: When the device temperature is too high, the cooling system (such as a fan or heatsink) will operate at a higher level to maintain the normal temperature. However, if the temperature continues to rise, the cooling system may exceed its design capacity and be unable to dissipate heat effectively, resulting in a further increase in the internal temperature of the device.

[0094] Possible signs of hardware failure: A sudden increase in temperature may be a sign that some hardware components are about to fail. For example, the power supply module, circuit board, or some sensors may generate abnormal heat due to aging, wear, or faults.

[0095] When it is detected that the edge computing device has a failure or performance degradation, first record the status of the edge computing device, and then forward the preprocessed image data to the management and control center based on the wireless network, including the following steps:​

[0096] According to the type of wireless network (such as Wi-Fi, 4G / 5G, LoRa, etc.), select an appropriate transmission protocol (such as TCP / IP, UDP, etc.) for data transmission, and send the preprocessed image data to the control center through the wireless network;

[0097] After receiving the data packet transmitted by the wireless network, the control center preprocesses the image data, including the following:

[0098] Noise reduction and filtering: Perform noise reduction on the image, and use filtering technology to remove unnecessary information in the image to improve the image quality.

[0099] Image cropping and scaling: According to the requirements of the recognition device, crop or scale the image to ensure that the image meets the input requirements of the recognition device.

[0100] Color space conversion: Convert the image from one color space to another (such as from RGB to grayscale image or HSV) so that the recognition algorithm can process the data more effectively.

[0101] Feature enhancement: Enhance the feature information in the image through techniques such as edge detection and histogram equalization to improve the recognition accuracy of the recognition device.

[0102] Compression: Compress the image data to reduce the bandwidth requirements for data transmission.

[0103] After the control center preprocesses the image data, it sends it to the recognition device for object recognition, including the following steps:

[0104] Image loading: The recognition device loads the received image data into memory for further processing.

[0105] Image normalization: Normalize the image, for example, adjust brightness, contrast, etc., to make it meet the input standards of the recognition model.

[0106] Data augmentation: Perform data augmentation before recognition, such as flipping, rotating, scaling, etc., to increase the robustness of recognition.

[0107] Model loading: The recognition device loads a pre-trained object recognition model (such as a convolutional neural network model) to prepare for processing the input image data.

[0108] Object recognition: Input the image data into the recognition model, and the model extracts features and classifies the image to recognize the objects in the image.

[0109] Result generation: The recognition model outputs the recognition results, including information such as the category, location, and confidence of the object.

[0110] After the object recognition is completed, the control center evaluates the object recognition effect of the recognition device, including the following steps:

[0111] The control center obtains the recognition error reporting frequency and the recognition duration deviation during the process of the recognition device recognizing the object, normalizes the recognition error reporting frequency and the recognition duration deviation, maps the value ranges of the recognition error reporting frequency and the recognition duration deviation to between [0, 1], obtains the normalized value of the recognition error reporting frequency and the normalized value of the recognition duration deviation, sums the normalized value of the recognition error reporting frequency and the normalized value of the recognition duration deviation to obtain the recognition weakening index. The larger the recognition weakening index, the worse the recognition effect of the recognition device;

[0112] Compare the obtained recognition weakening index with the effect threshold. The effect threshold is used to evaluate the quality of the object recognition effect of the recognition device. If the recognition weakening index is less than or equal to the effect threshold, it is evaluated that the object recognition effect of the recognition device is excellent. If the recognition weakening index is greater than the effect threshold, it is evaluated that the object recognition effect of the recognition device is poor.

[0113] The acquisition logic of the recognition error reporting frequency is: obtain the number of recognition error reports of the recognition device during the object recognition time period, divide the number of recognition error reports by the recognition duration to obtain the recognition error reporting frequency. The larger the recognition error reporting frequency, the worse the recognition effect of the recognition device.

[0114] The acquisition logic of the recognition duration deviation is: subtract the standard recognition duration from the actual recognition duration to obtain the recognition duration deviation. The larger the recognition duration deviation, the longer the time for the recognition device to recognize the object, and the worse the recognition effect of the recognition device.

[0115] After combining the edge computing device status with the object recognition effect, generate a maintenance score for each recognition device, sort all the recognition devices according to the maintenance score to generate a device list, and select the maintenance order of all the recognition devices based on the device list. The maintenance order information is sent to the relevant maintenance personnel, including the following steps:

[0116] After combining the edge computing device status with the object recognition effect, generate a maintenance score for each recognition device. The calculation expression is: , where is the management score, is the device status index, is the recognition weakening index, are the weights of the device status index and the recognition weakening index respectively, and .

[0117] Sort all the recognition devices in descending order according to the management score to generate a device list. The higher the ranking of a recognition device in the device list, the worse its overall status, and the more it needs to be maintained preferentially. Select the maintenance order of all recognition devices based on the device list and send the maintenance order information to the relevant maintenance personnel;

[0118] When there is a recognition device in the device list with a management score greater than the score threshold, the control center generates an emergency maintenance flag for the recognition device and sends it to the maintenance personnel, indicating that the recognition device needs to be maintained immediately.

[0119] Embodiment 3: The autonomous object recognition device based on edge computing described in this embodiment includes a performance detection module, a data forwarding module, and an object recognition module;

[0120] Performance detection module: When an object appears in the monitoring area of the recognition device, the recognition device automatically collects image data and performs performance detection on the edge computing device. The performance detection result is sent to the data forwarding module;

[0121] Data forwarding module: When it is detected that the edge computing device has a fault or performance degradation, first record the status of the edge computing device, and then forward the preprocessed image data to the control center based on the wireless network. The preprocessed image data is sent to the object recognition module;

[0122] Object recognition module: Used to perform object recognition on the preprocessed image data, and the object recognition result is sent to the control center.

[0123] The above formulas are all dimensionless and take their numerical values for calculation. The formulas are obtained by collecting a large amount of data for software simulation to get a formula that is closest to the actual situation. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0124] It should be understood that the term "and / or" in this article is only a description of the association relationship between associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. Among them, A and B can be singular or plural. In addition, the character " / " in this article generally represents an "or" relationship between the associated objects before and after, but it may also represent an "and / or" relationship. Specifically, it can be understood by referring to the context before and after.

[0125] It should be understood that in various embodiments of the present application, the magnitudes of the serial numbers of the above processes do not mean the order of execution. The execution order of each process should be determined according to its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0126] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed in this text can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application. Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated here.

[0127] As described above, the above is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed in this application can easily think of changes or substitutions, which should all be covered by the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

Claims

1. An autonomous object recognition method based on edge computing, characterized in that: The recognition method includes the following steps: When an object appears in the monitoring area of the recognition device, the recognition device automatically collects image data and performs a performance detection on the edge computing device; When it is detected that the edge computing device has a fault or a performance degradation, record the status of the edge computing device, and then preprocess the image data and forward it to the management and control center via a wireless network. After the management and control center preprocesses the image data, it is sent to the recognition device for object recognition; After the object recognition in this time is completed, the management and control center evaluates the object recognition effect of the recognition device, combines the status of the edge computing device with the object recognition effect to generate a maintenance score for each recognition device, sorts all the recognition devices according to the maintenance score to generate a device list, and selects the maintenance order of all the recognition devices based on the device list.

2. The method for autonomous object recognition based on edge computing according to claim 1, wherein: Performing a performance detection on the edge computing device includes the following steps: Obtain the power loss coefficient and the Fourier transform coefficient of heat conduction of the edge computing device; Comprehensively calculate the power loss coefficient and the Fourier transform coefficient of heat conduction to obtain a device status index; Compare the obtained device status index with a preset first abnormal threshold and a second abnormal threshold. The first abnormal threshold is used to determine whether the performance of the edge computing device has degraded, and the second abnormal threshold is used to determine whether the edge computing device has a fault; If the device status index is less than or equal to the first abnormal threshold, it is determined that the performance of the edge computing device has not degraded. If the device status index is greater than the first abnormal threshold and less than or equal to the second abnormal threshold, it is determined that the performance of the edge computing device has degraded. If the device status index is greater than the second abnormal threshold, it is determined that the edge computing device has a fault.

3. The method for autonomous object recognition based on edge computing according to claim 2, characterized in that: Comprehensively calculate the power loss coefficient and the Fourier transform coefficient of heat conduction to obtain a device status index. The expression is: ， In the formula, is the device status index, is the power loss coefficient, is the Fourier transform coefficient of heat conduction, , are the proportionality coefficients of the power loss coefficient and the Fourier transform coefficient of heat conduction respectively, and , are both greater than 0.

4. The method for autonomous object recognition based on edge computing according to claim 3, wherein: The management and control center evaluates the object recognition effect of the recognition device, including the following steps: The management and control center obtains the recognition error reporting frequency and the recognition duration deviation during the process of the recognition device recognizing an object, normalizes the recognition error reporting frequency and the recognition duration deviation so that the value ranges of the recognition error reporting frequency and the recognition duration deviation are mapped to between [0, 1], obtains the normalized value of the recognition error reporting frequency and the normalized value of the recognition duration deviation, and sums the normalized value of the recognition error reporting frequency and the normalized value of the recognition duration deviation to obtain a recognition weakening index; Compare the obtained recognition weakening index with an effect threshold. The effect threshold is used to evaluate the quality of the object recognition effect of the recognition device. If the recognition weakening index is less than or equal to the effect threshold, it is evaluated that the object recognition effect of the recognition device is excellent. If the recognition weakening index is greater than the effect threshold, it is evaluated that the object recognition effect of the recognition device is poor.

5. The method for autonomous object recognition based on edge computing according to claim 4, characterized in that: Sorting all the recognition devices according to the maintenance score to generate a device list includes the following steps: Combining the status of the edge computing device with the object recognition effect to generate a maintenance score for each recognition device. The calculation expression is: ， Wherein, is the management score, is the equipment status index, is the recognition weakening index, are the weights of the equipment status index and the recognition weakening index respectively, and ; Sort all the recognition devices from largest to smallest according to the management score to generate a device list, select the maintenance order of all the recognition devices based on the device list, and send the maintenance order information to the relevant maintenance personnel; When there is an identification device in the device list whose management score is greater than the score threshold, the control center generates an emergency maintenance identifier for the identification device and sends it to the relevant maintenance personnel.

6. The method for autonomous object recognition based on edge computing according to claim 4, wherein: The acquisition logic of the identification error reporting frequency is as follows: obtain the number of identification error reports of the identification device during the object identification time period; The acquisition logic of the identification duration deviation is as follows: subtract the standard identification duration from the actual identification duration to obtain the identification duration deviation.

7. The method for autonomous object recognition based on edge computing according to claim 3, characterized in that: The calculation expression of the power loss coefficient is: ， In the formula, is the power loss coefficient, is the nominal voltage of the device, is the instantaneous voltage, is the monitoring time range; The calculation expression of the Fourier transform coefficient of heat conduction is: ， Wherein, is the temperature value at position and time ; the temperature value at position and time is defined as the Fourier transform coefficient of heat conduction, is the temperature value at position and time . is the time step, representing the amount of time advancement in each iteration, is the space step, representing the interval length of space discretization, is the thermal diffusivity, defined as , wherein, is the thermal conductivity, is the density, is the specific heat capacity, and represent the temperature values at positions and respectively.

8. An autonomous object recognition device based on edge computing for implementing the recognition method according to any one of claims 1-7, characterized in that: It includes a performance detection module, a data forwarding module, and an object identification module; Performance detection module: When an object appears in the monitoring area of the identification device, the identification device automatically collects image data and performs performance detection on the edge computing device; Data forwarding module: When it is detected that there is a fault or performance degradation in the edge computing device, first record the status of the edge computing device, and then forward the preprocessed image data to the control center based on the wireless network for; Object identification module: used to perform object identification on the preprocessed image data, and send the object identification result to the control center.

Images