Feed production equipment defect identification method and system based on video monitoring

Abstract

The invention discloses a feed production equipment defect identification method and system based on video monitoring, and relates to the technical field of equipment defect identification, when the system operates, video frames and equipment operation state information are collected in real time through a video monitoring system, and the integrity and real-time performance of data are ensured. And the defect detection module adopts a convolutional neural network to analyze the de-noised video frames, so that the accuracy and efficiency of defect identification are greatly improved. And the data analysis module calculates and generates an initial threshold THZ through statistical analysis of defect occurrence frequency and types. Then, a threshold value adjusting module dynamically adjusts a threshold value according to real-time detection data, and the self-adaptability of the detection model under different conditions is ensured; according to the process, the adaptive capacity of the system to different production environments is remarkably improved, the missing report rate and the false report rate are reduced, and the system can clearly know the occurrence trend and frequency of each type of defects in a specific time period through accurate defect statistics and classification.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment defect recognition, and in particular to a method and system for identifying defects in feed production equipment based on video monitoring. Background Art

[0002] In modern manufacturing, intelligent manufacturing and industrial automation are key trends driving the development of the industry. Intelligent manufacturing involves optimizing the production process through advanced technology and data analysis to improve production efficiency and product quality. In the field of intelligent manufacturing, the management and maintenance of feed production equipment is particularly important. As the core equipment in the industrial production line, the stability and operating efficiency of feed production equipment directly affect the overall quality and economic benefits of production. The feed production equipment defect recognition system based on video surveillance came into being in this context. It aims to detect and identify potential defects of equipment through real-time video stream analysis to ensure efficient operation of equipment and high quality of products.

[0003] Most current systems rely on fixed thresholds to determine equipment defects, but this method has certain limitations. Since the equipment operating status may change due to environmental conditions, production load and other factors, the fixed threshold may not be able to adapt to the dynamic changes in the equipment status. This limitation leads to poor detection results under different operating conditions, and may result in missed or false alarms, affecting the overall reliability and accuracy of the system.

[0004] Problems caused by fixed thresholds often lead to a series of adverse effects. When the equipment is running under different production loads or environmental conditions, the fixed detection threshold may not accurately reflect the actual status of the equipment. This will lead to inaccurate defect detection, such as missed reports that lead to potential problems not being discovered in time, or false reports that lead to excessive shutdown of the production line, thus affecting production efficiency and product quality. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention provides a method and system for identifying defects in feed production equipment based on video monitoring, which solves the problems mentioned in the background technology.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: a feed production equipment defect identification system based on video monitoring, including a data acquisition module, a defect detection module, a data analysis module, a threshold adjustment module, a feedback and adjustment module and a report generation module;

[0007] The data acquisition module collects data from the video monitoring system, including video frames F and equipment operation status information DS;

[0008] The defect detection module preprocesses the video frame F collected by the data acquisition module, including denoising the video frame F, using a convolutional neural network to perform defect detection on the denoised video frame F, and obtaining defect information D, specifically performing defect detection on the feed production equipment through a defect detection algorithm;

[0009] The data analysis module analyzes the frequency and type of defects based on the information from the defect detection module and calculates statistical indicators, including the mean μ Defect and standard deviation σ defect , get the initial threshold TH Z ;

[0010] The threshold adjustment module dynamically adjusts the initial threshold TH of defect detection according to the data analysis module Z , including based on defect frequency FQ i (T) and defect type i, calculate statistical indicators, and adjust the initial detection threshold TH according to the calculation results Z , get the new threshold TH new ;

[0011] The feedback and adjustment module is based on the new threshold TH new Changes in the comprehensive threshold TH to Applied to real-time defect detection system and monitor the effect of adjustment, including setting the comprehensive threshold TH to Input into the defect detection model to monitor the comprehensive threshold TH to Check the detection results and record the improvement issues;

[0012] The report generation module generates real-time reports and analysis, showing the comprehensive threshold TH to Adjustment effects, system operation status and defect statistics, summary data information, threshold adjustment records and detection effects.

[0013] The dynamic threshold adjustment mechanism can adapt to changes in equipment status, improve the accuracy of defect detection, reduce false alarms and missed alarms, and the system can adjust in real time to adapt to changes in equipment status such as aging and wear.

[0014] Preferably, the data acquisition module includes a video stream capturing unit and a data recording unit;

[0015] The video stream capture unit captures the video stream Vi deoS in real time through the camera t , obtain continuous video frames F, and store the captured video frames F into the memory;

[0016] The data recording unit records the defect information D detected in the video frame F and records the equipment operation status information DS; the equipment operation status information DS includes temperature parameters, pressure parameters and vibration parameters, which are obtained through various sensors, specifically including temperature sensors, pressure sensors and vibration sensors.

[0017] By accurately capturing and recording each video frame and defect detection results, high data quality is ensured. Real-time capture and recording can promptly reflect the status and defect information of the equipment, facilitating rapid response.

[0018] Preferably, the defect detection module includes a video preprocessing unit and a defect detection unit;

[0019] The video preprocessing unit performs denoising on the video frame F through a denoising algorithm to improve the contrast and brightness of the video, highlight defects, and obtain a denoised video frame DF;

[0020] The defect detection unit performs defect detection on the denoised video frame DF, inputs the denoised video frame DF into a convolutional neural network, performs defect detection, generates a defect detection result, and marks a defect area;

[0021] The preprocessing step can effectively improve image quality and reduce noise, thereby improving the accuracy of convolutional neural network detection. Through high-quality preprocessing, false positives and false negatives caused by poor image quality are reduced.

[0022] Preferably, the data analysis module includes a defect statistics unit and a statistical analysis unit;

[0023] The defect statistics unit calculates the frequency of occurrence of each type of defect within a certain time period, and records and classifies the detected defect types, counts the number of occurrences of each defect type within a given time period, classifies different types of defects, and counts the number of defect types, marking the total number of defects N;

[0024] The statistical analysis unit calculates the defect frequency FQ i The mean μ of (T) Defect and standard deviation σ defect Statistical indicators, through the mean μ Defect and standard deviation σ defect Perform fitting and obtain the initial threshold TH Z ;

[0025] The mean μ Defect By the defect frequency FQ i (T) aggregate acquisition;

[0026] The standard deviation σ defect By the defect frequency FQ i (T) and mean μDefect Calculation acquisition of

[0027] The initial threshold TH Z By the mean μ Defect and standard deviation σ defect Get.

[0028] Preferably, the threshold adjustment module includes a threshold updating unit and a threshold adjustment unit;

[0029] The threshold updating unit receives the initial threshold TH of the statistical analysis unit. Z , and according to the initial threshold TH Z Calculate the new threshold TH new , and set the new threshold TH new Applied to the defect detection module;

[0030] Reflects the adjustment effect of image quality on the detection threshold. The value of this function is usually between 0 and 1. The better the image quality, the closer γ(DF(t)) is to 1.

[0031] Preferably, the threshold adjustment unit performs adjustment according to the device operation status information DS and the denoised video frame DF; and compares the device operation status information change ΔDS, the denoised video frame change ΔDF and the trigger judgment value ΔTo:

[0032] When the trigger judgment value ΔTo is greater than or equal to the sum of the absolute value of the device operation status information change ΔDS and the denoised video frame change ΔDF, the threshold adjustment mechanism is triggered;

[0033] When the trigger judgment value ΔTo is less than the sum of the absolute value of the device operation status information change ΔDS and the denoised video frame change ΔDF, the threshold adjustment mechanism is not triggered;

[0034] The device operation status information change amount ΔDS is obtained by the difference between the denoised video frame data at time point t and the denoised video frame data at time point t-1;

[0035] The denoised video frame variation ΔDF is obtained by the difference between the denoised video frame data at time point t and the denoised video frame data at time point t-1;

[0036] When the threshold adjustment mechanism is triggered, the device state weight factor β and the video frame quality adjustment factor γ (DF(t)) are recalculated according to the device operation state information DS, and the new threshold TH is recalculated. new ;

[0037] The equipment state weight factor β is obtained according to the deviation between the current value and the historical value of the equipment operation state information DS;

[0038] The device state weight factor β is obtained by comparing the device operation state information DS and the historical average value DS of the device operation state information. avg Calculation acquisition of

[0039] Substitute the calculated device state weight factor β(t) into the new threshold TH new In the formula, the new state threshold TH is obtained newB ;

[0040] When the equipment operation status information DS deviates from the normal working range, the equipment state weight factor β(t) changes, thus affecting the new state threshold TH newB changes occur;

[0041] The denoised video frame DF is a video frame extracted from the video surveillance system and subjected to denoising processing. The system evaluates the clarity, brightness and contrast of the video frame to determine the quality level of the current frame DF;

[0042] According to the quality of the denoised video frame, the video frame quality adjustment factor γ(DF(t)) is calculated;

[0043] Substitute the calculated video frame quality adjustment factor γ(DF(t)) into the new threshold TH new In the formula, get the new denoising threshold TH newF ;

[0044] When the video frame quality fluctuates, the video frame quality adjustment factor γ(DF(t)) and the new denoising threshold TH newF Changes linearly to compensate for fluctuations in video frame quality;

[0045] According to the new state threshold TH newB and the new denoising threshold TH newF Composed of threshold set W;

[0046] Preferably, the feedback and adjustment module includes a threshold application unit and an effect monitoring and recording unit;

[0047] The threshold application unit receives the threshold set W from the threshold adjustment module and obtains the comprehensive threshold TH to For defect detection;

[0048] Through the comprehensive threshold TH to , update the output value of the convolutional neural network model so that the convolutional neural network model is based on the comprehensive threshold TH to Conduct defect detection;

[0049] During the defect detection process, the system is within a fixed period of time t to Use the current comprehensive threshold TH to Conduct continuous monitoring;

[0050] The time range within a fixed period t to If the monitoring and detection effect is evaluated as not triggering the threshold adjustment mechanism, the system will restore the initial threshold TH Z , and maintain the initial threshold TH Z For subsequent detection, continue to use the initial threshold TH Z Until the trigger threshold adjustment mechanism is obtained.

[0051] This mechanism ensures that the system can quickly return to the default state under stable operating conditions, reducing unnecessary adjustment operations, thereby improving the operating efficiency and stability of the system.

[0052] Preferably, the effect monitoring and recording unit records the comprehensive threshold TH to The detection accuracy ACnew, false alarm rate FNRnew and false alarm rate FPRnew under this condition;

[0053] The detection accuracy ACnew is obtained by the ratio of the sum of the number of true positive examples and the number of true negative examples to the number of false positive examples and the number of false negative examples;

[0054] The false negative rate FNRnew is obtained by the ratio of the number of false negative examples to the number of true positive examples;

[0055] The false alarm rate FPRnew is obtained by the ratio of the number of false positive examples to the number of true negative examples;

[0056] The detection results are recorded in the log and a report is generated.

[0057] Preferably, the report generation module generates a report including text, tables and charts based on the summarized data and analysis results, specifically including:

[0058] System operation status: Displays the historical change trend of the equipment operation status information DS;

[0059] Defect statistics: Displays the distribution of defect types, total number and statistical charts of each defect type;

[0060] Threshold adjustment record: record each threshold adjustment and its impact on detection results;

[0061] Detection effect: shows the comparison of system detection accuracy ACnew, false alarm rate FNRnew, and false alarm rate FPRnew before and after threshold adjustment.

[0062] A method for identifying defects in feed production equipment based on video monitoring comprises the following steps:

[0063] Step 1: The data acquisition module collects data from the video surveillance system, including video frames F and equipment operation status information DS;

[0064] Step 2: The defect detection module preprocesses the video frame F collected by the data acquisition module, including denoising the video frame F, using a convolutional neural network to perform defect detection on the denoised video frame F, and obtaining defect information D, and specifically performing defect detection on the feed production equipment through a defect detection algorithm;

[0065] Step 3: The data analysis module analyzes the frequency and type of defects based on the information from the defect detection module and calculates statistical indicators, including the mean μ Defect and standard deviation σ defect , get the initial threshold TH Z ;

[0066] Step 4: The threshold adjustment module dynamically adjusts the initial threshold TH of defect detection according to the data analysis module Z , including based on defect frequency FQ i (T) and defect type i, calculate statistical indicators, and adjust the initial detection threshold TH according to the calculation results Z , get the new threshold TH new ;

[0067] Step 5: Feedback and adjustment module according to the new threshold TH new Changes in the comprehensive threshold TH to Applied to real-time defect detection system and monitor the effect of adjustment, including setting the comprehensive threshold TH to Input into the defect detection model to monitor the comprehensive threshold TH to Check the detection results and record the improvement issues;

[0068] Step 6: The report generation module generates real-time reports and analysis, showing the comprehensive threshold TH to Adjustment effects, system operation status and defect statistics, summary data information, threshold adjustment records and detection effects.

[0069] The present invention provides a method and system for identifying defects in feed production equipment based on video monitoring, which has the following beneficial effects:

[0070] (1) When the system is running, the video monitoring system collects video frames and equipment operation status information in real time to ensure the integrity and real-time nature of the data. The defect detection module uses a convolutional neural network to analyze the denoised video frames, which greatly improves the accuracy and efficiency of defect identification. The data analysis module calculates and generates the initial detection threshold through statistical analysis of the frequency and type of defects. Subsequently, the threshold adjustment module dynamically adjusts the threshold according to the real-time detection data to ensure the adaptability of the detection model under different conditions. This process significantly improves the system's adaptability to different production environments and reduces the rate of missed and false alarms.

[0071] (2) By introducing the threshold adjustment module, the system can dynamically adjust the threshold of defect detection in real time according to the equipment operation status information and video frame analysis results. This dynamic adjustment mechanism enables the system to automatically adapt to changes in equipment status and reduce false alarms and missed alarms. The feedback and adjustment module forms a closed-loop feedback system by monitoring the application effect of the new threshold and combining the equipment operation status and detection results. This mechanism ensures that the actual effect of the threshold adjustment can be monitored and optimized in real time, ensuring the continuous and efficient operation of the system and being able to quickly respond to abnormal situations in the production process.

[0072] (3) A defect statistics unit is introduced to calculate the frequency of occurrence of each type of defect within a specific time period, and to record and classify the detected defect types in detail, classify different types of defects, and count the number of occurrences of each type of defect. Through accurate defect statistics and classification, the system can clearly understand the occurrence trend and frequency of each type of defect within a specific time period. This precise data classification and statistics provides a reliable data basis for subsequent threshold adjustment and defect warning, which helps to accurately identify high-risk defects and reduce the occurrence of equipment failures.

[0073] (4) The new threshold is calculated through the threshold update unit, and the device state weight factor and the denoised video frame quality adjustment factor are introduced. The device operation state information DS and the quality of the denoised video frame DF are comprehensively considered to dynamically adjust the threshold of defect detection. The threshold calculation mechanism improves the system's sensitivity to different operation states and video quality changes, ensuring that the system can adaptively adjust the threshold of defect detection in real time, thereby improving the accuracy and reliability of detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Figure 1 This is a schematic diagram of a flow chart of a feed production equipment defect identification system based on video monitoring according to the present invention;

[0075] Figure 2 The present invention is a schematic diagram of the steps of a method for identifying defects in feed production equipment based on video monitoring. DETAILED DESCRIPTION

[0076] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0077] Example 1

[0078] The present invention provides a feed production equipment defect recognition system based on video monitoring, please refer to Figure 1 , including data acquisition module, defect detection module, data analysis module, threshold adjustment module, feedback and adjustment module and report generation module;

[0079] The data acquisition module collects data from the video monitoring system, including video frames F and equipment operation status information DS;

[0080] The defect detection module preprocesses the video frame F collected by the data acquisition module, including denoising the video frame F, using a convolutional neural network to perform defect detection on the denoised video frame F, and obtaining defect information D, specifically performing defect detection on the feed production equipment through a defect detection algorithm;

[0081] The data analysis module analyzes the frequency and type of defects based on the information from the defect detection module and calculates statistical indicators, including the mean μ Defect and standard deviation σ defect , get the initial threshold TH Z ;

[0082] The threshold adjustment module dynamically adjusts the initial threshold TH of defect detection according to the data analysis module Z , including based on defect frequency FQ i (T) and defect type i, calculate statistical indicators, and adjust the initial detection threshold TH according to the calculation results Z , get the new threshold TH new ;

[0083] The feedback and adjustment module is based on the new threshold TH new Changes in the comprehensive threshold TH to Applied to real-time defect detection system and monitor the effect of adjustment, including setting the comprehensive threshold TH to Input into the defect detection model to monitor the comprehensive threshold TH to Check the detection results and record the improvement issues;

[0084] The report generation module generates real-time reports and analysis, showing the comprehensive threshold TH to Adjustment effects, system operation status and defect statistics, summary data information, threshold adjustment records and detection effects.

[0085] In this embodiment, by introducing the threshold adjustment module, the system can dynamically adjust the threshold of defect detection in real time according to the equipment operation status information D and the analysis results of the video frame F. This dynamic adjustment mechanism enables the system to automatically adapt to changes in the equipment status and reduce false alarms and missed alarms. The feedback and adjustment module forms a closed-loop feedback system by monitoring the application effect of the new threshold, combining the equipment operation status and the detection results. This mechanism ensures that the actual effect after the threshold adjustment can be monitored and optimized in real time, ensuring the continuous and efficient operation of the system and being able to quickly respond to abnormal situations in the production process.

[0086] The convolutional neural network is used to denoise and detect defects in the video frame F, which enhances the intelligence and automation level of detection. Through intelligent preprocessing and detection, the system can accurately identify tiny defects in equipment operation, effectively improving the detection accuracy and the response speed of the system. The report generation module can generate a comprehensive report in real time, showing the system operation status, defect statistics and the effect of threshold adjustment. The real-time report provides timely and accurate data support for production managers, helping them to better monitor the production process, make decisions and optimize the production process.

[0087] Example 2

[0088] This embodiment is explained in Example 1, please refer to Figure 1 ,Specifically: the data acquisition module includes a video stream capturing unit and a data recording unit;

[0089] The video stream capture unit captures the video stream Vi deoS in real time through the camera t , obtain continuous video frames F, and store the captured video frames F into the memory;

[0090] F=Capture(VideoS t );

[0091] In the formula, Capture represents the video capture function, VideoS t Represents the video stream at time t;

[0092] The data recording unit records the defect information D detected in the video frame F and records the equipment operation status information DS; the equipment operation status information DS includes temperature parameters, pressure parameters and vibration parameters, which are obtained through various sensors, specifically including temperature sensors, pressure sensors and vibration sensors.

[0093] The defect detection module includes a video preprocessing unit and a defect detection unit;

[0094] The video preprocessing unit performs denoising on the video frame F through a denoising algorithm to improve the contrast and brightness of the video, highlight defects, and obtain a denoised video frame DF;

[0095] The defect detection unit performs defect detection on the denoised video frame DF, inputs the denoised video frame DF into a convolutional neural network, performs defect detection, generates a defect detection result, and marks a defect area;

[0096] D = CNN (DF);

[0097] Wherein, CNN represents the CNN model, and D represents the defect information, including the defect location and type.

[0098] In this embodiment, a video stream capture unit and a data recording unit are introduced. The video stream capture unit captures and stores continuous video frames F in real time through a camera, and the data recording unit records in detail the defect information D and equipment operation status information DS in the video frames. This mechanism ensures the system's real-time monitoring of production equipment, provides high-quality and continuous image data, and synchronously records key equipment status data, providing sufficient basic data support for subsequent defect detection and analysis.

[0099] In the defect detection module, the video preprocessing unit processes the video frame F through an advanced denoising algorithm, which improves the contrast and brightness of the video and highlights the possible defect areas. This preprocessing method improves the quality of the video data, making the defects more obvious in the image, and enhances the effectiveness and accuracy of subsequent defect detection, thereby reducing the risk of false detection and missed detection. The defect detection unit uses a convolutional neural network to perform deep learning analysis on the denoised video frame DF, automatically identifying and marking the defect area, including the specific location and type of the defect.

[0100] Example 3

[0101] This embodiment is explained in Example 2. Please refer to Figure 1 ,Specifically: the data analysis module includes a defect statistics unit and a statistical analysis unit;

[0102] The defect statistics unit calculates the frequency of occurrence of each type of defect within a certain time period, and records and classifies the detected defect types, counts the number of occurrences of each defect type within a given time period, classifies different types of defects, and counts the number of defect types, marking the total number of defects N;

[0103]

[0104] In the formula, FQ i (T) represents the occurrence frequency of the i-th type of defect in the time period T, Ni (T) represents the total number of defects of type i in time period T; T represents the length of time period;

[0105]

[0106] Where Count(T) represents the total number of defects in time period T;

[0107] The statistical analysis unit calculates the defect frequency FQ i The mean μ of (T) Defect and standard deviation σ defect Statistical indicators, through the mean μ Defect and standard deviation σ defect Perform fitting and obtain the initial threshold TH Z ;

[0108] The mean μ Defect Obtained by the following formula:

[0109]

[0110] In the formula, μ Defect represents the mean of defect frequency, and N represents the total number of defect types;

[0111] The standard deviation σ defect Obtained by the following formula:

[0112]

[0113] In the formula, σ defect represents the standard deviation of the defect frequency;

[0114] The initial threshold TH Z Obtained by the following formula:

[0115] TH Z =μ Defect +α*σ defect ;

[0116] Wherein, α represents the adjustment coefficient, which is specifically used to control the sensitivity of the threshold.

[0117] In this embodiment, a defect statistics unit is introduced to calculate the frequency of occurrence of each type of defect in a specific time period, and the detected defect types are recorded and classified in detail, different types of defects are classified, and the number of occurrences of each type of defect is counted. Through accurate defect statistics and classification, the system can clearly understand the occurrence trend and frequency of each type of defect in a specific time period. This accurate data classification and statistics provide a reliable data basis for subsequent threshold adjustment and defect warning, which helps to accurately identify high-risk defects and reduce the occurrence of equipment failures.

[0118] The statistical analysis unit calculates the mean μ of the defect frequency Defect and standard deviation σ defect Calculate and use these statistical indicators to fit and obtain the initial threshold TH Z , by the mean value μ of the defect frequency Defect and standard deviation σ defect By calculating, the system can quantify the occurrence of defects, help identify abnormal defect frequency fluctuations, and then pass the initial threshold TH Z The sensitivity of defect detection can be reasonably set through calculation to avoid false positives or negative negatives.

[0119] Example 4

[0120] This embodiment is explained in Example 3, please refer to Figure 1 , specifically: the threshold adjustment module includes a threshold updating unit and a threshold adjustment unit;

[0121] The threshold updating unit receives the initial threshold TH of the statistical analysis unit. Z , and according to the initial threshold TH Z Calculate the new threshold TH new , and set the new threshold TH new Applied to the defect detection module;

[0122] The new threshold TH new Obtained by the following formula:

[0123]

[0124] In the formula, β represents the device state weight factor, which is used to adjust the device operation state DS to the initial threshold TH Z , γ(DF(t)) represents the quality adjustment factor of the video frame after denoising.

[0125] The threshold adjustment unit makes adjustments according to the device operation status information DS and the denoised video frame DF; and compares the device operation status information change ΔDS, the denoised video frame change ΔDF and the trigger judgment value ΔTo:

[0126] When ΔTo≥|ΔDS|+|ΔDF|, the threshold adjustment mechanism is triggered;

[0127] When ΔTo<ΔDS|+|ΔDF|, the threshold adjustment mechanism is not triggered;

[0128] The equipment operation status information change ΔDS is obtained by the following formula:

[0129] ΔDS=DS t -DS t-1 ;

[0130] Where, DS t Represents the denoised video frame data at the current time point t, DS t-1 Represents the denoised video frame data at time point t-1;

[0131] The denoised video frame variation ΔDF is obtained by the following formula:

[0132] ΔDF=DF t -DF t-1 ;

[0133] Where DF t Represents the denoised video frame data at the current time point t, DF t-1 Represents the denoised video frame data at time point t-1;

[0134] When the threshold adjustment mechanism is triggered, the device state weight factor β and the video frame quality adjustment factor γ (DF(t)) are recalculated according to the device operation state information DS, and the new threshold TH is recalculated. new ;

[0135] The equipment state weight factor β is obtained according to the deviation between the current value and the historical value of the equipment operation state information DS;

[0136] The device state weight factor β is obtained by the following formula:

[0137]

[0138] Where, DS avg represents the historical average value of the equipment operation status information, and k represents the adjustment factor;

[0139] Substitute the calculated device state weight factor β(t) into the new threshold TH new In the formula, the new state threshold TH is obtained newB ;

[0140] The new state threshold TH newB Obtained by the following formula:

[0141]

[0142] When the equipment operation status information DS deviates from the normal working range, the equipment state weight factor β(t) changes, thus affecting the new state threshold TH newB changes occur;

[0143] The denoised video frame DF is a video frame extracted from the video surveillance system and subjected to denoising. The system evaluates the clarity, brightness and contrast of the video frame to determine the quality level of the current frame DF.

[0144] According to the quality of the denoised video frame, the video frame quality adjustment factor γ(DF(t)) is calculated;

[0145] The video frame quality adjustment factor γ(DF(t)) is obtained by the following formula:

[0146] γ(DF(t)=1-c*(1-Q(DF(t)));

[0147] Where Q(DF(t)) represents the quality score of the denoised video frame, ranging from 0 to 1, and c represents the quality impact factor;

[0148] Substitute the calculated video frame quality adjustment factor γ(DF(t)) into the new threshold TH new In the formula, get the new denoising threshold TH newF ;

[0149] The new denoising threshold TH newF Obtained by the following formula:

[0150]

[0151] When the video frame quality fluctuates, the video frame quality adjustment factor γ(DF(t)) and the new denoising threshold TH newF Changes linearly to compensate for fluctuations in video frame quality;

[0152] According to the new state threshold TH newB and the new denoising threshold TH newF Composed of threshold set W;

[0153] The threshold set W is obtained by the following formula:

[0154] W={TH newB1 ,TH newB2 ,...,TH newBn ,TH newF1 ,TH newF2 ,...,TH newFn}.

[0155] In this embodiment, a new threshold is calculated by a threshold updating unit, and the device state weight factor β and the denoised video frame quality adjustment factor γ(DF(t)) are introduced to comprehensively consider the device operation state information DS and the quality of the denoised video frame DF, and dynamically adjust the threshold of defect detection. The comprehensive threshold TH toThe calculation mechanism improves the system's sensitivity to different operating states and video quality changes, ensuring that the system can adaptively adjust the threshold of defect detection in real time, thereby improving the accuracy and reliability of detection. Through the threshold adjustment unit, the threshold adjustment mechanism is determined based on the change in the device operating state information D and the denoised video frame DF to ensure that the system can respond to changes in the device state or video quality in a timely manner. This dynamic response mechanism enables the system to quickly adjust the detection threshold when the device operating state or video quality changes significantly, avoiding false alarms or missed alarms caused by fluctuations in the environment or device state, and enhancing the robustness and stability of the system.

[0156] By calculating the device state weight factor β and comprehensively considering the deviation between the current device operating state and the historical average, the threshold adjustment can fully reflect the long-term trend and short-term changes of the device state. This method can not only accurately capture abnormal changes in device state, but also use historical data to smooth sudden fluctuations, thereby more stably adjusting the detection threshold, reducing the system's oversensitivity to short-term fluctuations, and improving the overall system operation stability. The system uses the new state threshold TH newB and the new denoising threshold TH newF Generate a comprehensive threshold set W and apply it to the defect detection model to form a more accurate detection strategy. The introduction of the comprehensive threshold set W enables the system to form a more accurate detection standard based on multi-dimensional parameter information and adapt to complex production environments. This method improves the detection accuracy of the system and ensures that efficient defect detection capabilities can be maintained under various operating conditions and video quality.

[0157] Example 5

[0158] This embodiment is explained in Example 4. Please refer to Figure 1 ,Specifically: the feedback and adjustment module includes a threshold application unit and an effect monitoring and recording unit;

[0159] The threshold application unit receives the threshold set W from the threshold adjustment module and obtains the comprehensive threshold TH to For defect detection, the integrated threshold TH to Obtained through the following formula:

[0160]

[0161] Through the comprehensive threshold TH to , update the output value of the convolutional neural network model so that the convolutional neural network model is based on the comprehensive threshold TH to Conduct defect detection;

[0162] During the defect detection process, the system is within a fixed period of time t toUse the current comprehensive threshold TH to Conduct continuous monitoring;

[0163] The time range within a fixed period t to If the monitoring and detection effect is evaluated as not triggering the threshold adjustment mechanism, the system will restore the initial threshold TH Z , and maintain the initial threshold TH Z For subsequent detection, continue to use the initial threshold TH Z Until the trigger threshold adjustment mechanism is obtained.

[0164] The effect monitoring and recording unit records the comprehensive threshold value TH to The detection accuracy ACnew, false alarm rate FNRnew and false alarm rate FPRnew under this condition;

[0165]

[0166] Where TP new Indicates the new threshold TH new The number of true positives under new Indicates the new threshold TH new The number of true negative examples, FP new Indicates the new threshold TH new The number of false positives under new Indicates the new threshold TH new The number of false negatives under

[0167] The detection results are recorded in the log and a report is generated.

[0168] The report generation module generates a report including text, tables and charts based on the summarized data and analysis results, specifically including:

[0169] System operation status: Displays the historical change trend of the equipment operation status information DS;

[0170] Defect statistics: Displays the distribution of defect types, total number and statistical charts of each defect type;

[0171] Threshold adjustment record: record each threshold adjustment and its impact on detection results;

[0172] Detection effect: shows the comparison of system detection accuracy ACnew, false alarm rate FNRnew, and false alarm rate FPRnew before and after threshold adjustment.

[0173] In this embodiment, through the threshold application unit, the system receives and uses the comprehensive threshold TH calculated by the threshold set W. to , in order to update the output value of the convolutional neural network model, so that the model can detect defects according to the comprehensive threshold THto This mechanism ensures the dynamics and flexibility of the detection process. to The dynamic adjustment enables the system to adaptively adjust the detection criteria when facing different device states and video quality, reducing missed alerts and false alerts, thereby improving the detection accuracy of the system.

[0174] During the detection process, the system is within a fixed period of time range t to Use the current comprehensive threshold TH to If the detection effect is stable during the period and the threshold adjustment mechanism is not triggered, the system will restore and maintain the initial threshold TH Z , until the next event that triggers the adjustment mechanism occurs. This mechanism ensures the stability of the system, avoids frequent threshold adjustments when the detection environment is relatively stable, reduces the consumption of system resources, and can make timely adjustments when the detection environment changes significantly, ensuring the accuracy of detection.

[0175] Example 6

[0176] A method for identifying defects in feed production equipment based on video surveillance, please refer to Figure 2 , specifically:

[0177] The following steps are involved:

[0178] Step 1: The data acquisition module collects data from the video surveillance system, including video frames F and equipment operation status information DS;

[0179] Step 2: The defect detection module preprocesses the video frame F collected by the data acquisition module, including denoising the video frame F, using a convolutional neural network to perform defect detection on the denoised video frame F, and obtaining defect information D, and specifically performing defect detection on the feed production equipment through a defect detection algorithm;

[0180] Step 3: The data analysis module analyzes the frequency and type of defects based on the information from the defect detection module and calculates statistical indicators, including the mean μ Defect and standard deviation σ defect , get the initial threshold TH Z ;

[0181] Step 4: The threshold adjustment module dynamically adjusts the initial threshold TH of defect detection according to the data analysis module Z , including based on defect frequency FQ i (T) and defect type i, calculate statistical indicators, and adjust the initial detection threshold TH according to the calculation results Z , get the new threshold TH new ;

[0182] Step 5: Feedback and adjustment module according to the new threshold TH new Changes in the comprehensive threshold TH to Applied to real-time defect detection system and monitor the effect of adjustment, including setting the comprehensive threshold TH to Input into the defect detection model to monitor the comprehensive threshold TH to Check the detection results and record the improvement issues;

[0183] Step 6: The report generation module generates real-time reports and analysis, showing the comprehensive threshold TH to Adjustment effects, system operation status and defect statistics, summary data information, threshold adjustment records and detection effects.

[0184] In this embodiment, through steps one to six, this method significantly improves the accuracy and response speed of detection by integrating convolutional neural networks and intelligent threshold adjustment mechanisms. Traditional methods often rely on fixed thresholds, cannot flexibly respond to changes in equipment status and environment, and are prone to high false alarms and missed reports. This method solves these problems through dynamic adjustment and intelligent feedback, ensuring that high-precision detection effects are maintained under different working conditions. This embodiment realizes all-round monitoring and intelligent identification of defects in feed production equipment, and improves the stability, reliability and detection accuracy of the system. Through dynamic threshold adjustment and intelligent feedback mechanisms, the system can adaptively optimize the detection process, greatly reducing the shortcomings of traditional methods, and providing a more accurate and efficient defect identification solution.

[0185] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A feed production equipment defect identification system based on video monitoring, characterized in that: It includes data acquisition module, defect detection module, data analysis module, threshold adjustment module, feedback and adjustment module and report generation module; The data acquisition module collects data from the video monitoring system, including video frames F and equipment operation status information DS; The defect detection module preprocesses the video frame F collected by the data acquisition module, including denoising the video frame F, using a convolutional neural network to perform defect detection on the denoised video frame F, and obtaining defect information D, specifically performing defect detection on the feed production equipment through a defect detection algorithm; The data analysis module analyzes the frequency and type of defects based on the information from the defect detection module and calculates statistical indicators, including the mean μ Defect and standard deviation σ defect , get the initial threshold TH Z ; The threshold adjustment module dynamically adjusts the initial threshold TH of defect detection according to the data analysis module Z , including based on defect frequency FQ i (T) and defect type i, calculate statistical indicators, and adjust the initial detection threshold TH according to the calculation results Z , get the new threshold TH new ; The feedback and adjustment module is based on the new threshold TH new Changes in the comprehensive threshold TH to Applied to real-time defect detection system and monitor the effect of adjustment, including setting the comprehensive threshold TH to Input into the defect detection model to monitor the comprehensive threshold TH to Check the detection results and record the improvement issues; The report generation module generates real-time reports and analysis, showing the comprehensive threshold TH to Adjustment effects, system operation status and defect statistics, summary data information, threshold adjustment records and detection effects.

2. A feed production equipment defect identification system based on video monitoring according to claim 1, characterized in that: The data acquisition module includes a video stream capturing unit and a data recording unit; The video stream capture unit captures the video stream Vi deoS in real time through the camera t , obtain continuous video frames F, and store the captured video frames F into the memory; The data recording unit records the defect information D detected in the video frame F and records the equipment operation status information DS; the equipment operation status information DS includes temperature parameters, pressure parameters and vibration parameters, which are obtained through various sensors, specifically including temperature sensors, pressure sensors and vibration sensors.

3. The feed production equipment defect identification system based on video monitoring according to claim 1 is characterized in that: The defect detection module includes a video preprocessing unit and a defect detection unit; The video preprocessing unit performs denoising on the video frame F through a denoising algorithm to improve the contrast and brightness of the video, highlight defects, and obtain a denoised video frame DF; The defect detection unit performs defect detection on the denoised video frame DF, inputs the denoised video frame DF into a convolutional neural network, performs defect detection, generates a defect detection result, and marks a defect area.

4. A feed production equipment defect identification system based on video monitoring according to claim 1, characterized in that: The data analysis module includes a defect statistics unit and a statistics analysis unit; The defect statistics unit calculates the frequency of occurrence of each type of defect within a certain time period, and records and classifies the detected defect types, counts the number of occurrences of each defect type within a given time period, classifies different types of defects, and counts the number of defect types, marking the total number of defects N; The statistical analysis unit calculates the defect frequency FQ i The mean μ of (T) Defect and standard deviation σ defect Statistical indicators, through the mean μ Defect and standard deviation σ defect Perform fitting and obtain the initial threshold TH Z ; The mean μ Defect By the defect frequency FQ i (T) aggregate acquisition; The standard deviation σ defect By the defect frequency FQ i (T) and mean μ Defect Calculation acquisition of The initial threshold TH Z By the mean μ Defect and standard deviation σ defect Get.

5. A feed production equipment defect identification system based on video monitoring according to claim 4, characterized in that: The threshold adjustment module includes a threshold updating unit and a threshold adjustment unit; The threshold updating unit receives the initial threshold TH of the statistical analysis unit. Z , and according to the initial threshold TH Z Calculate the new threshold TH new , and set the new threshold TH new Applied to the defect detection module; The new threshold TH new By setting the initial threshold TH Z , device state weight factor and denoised video frame quality adjustment factor are calculated and obtained.

6. A feed production equipment defect identification system based on video monitoring according to claim 5, characterized in that: The threshold adjustment unit makes adjustments according to the device operation status information DS and the denoised video frame DF; and compares the device operation status information change ΔDS, the denoised video frame change ΔDF and the trigger judgment value ΔTo: When the trigger judgment value ΔTo is greater than or equal to the sum of the absolute value of the device operation status information change ΔDS and the denoised video frame change ΔDF, the threshold adjustment mechanism is triggered; When the trigger judgment value ΔTo is less than the sum of the absolute value of the device operation status information change ΔDS and the denoised video frame change ΔDF, the threshold adjustment mechanism is not triggered; The device operation status information change amount ΔDS is obtained by the difference between the denoised video frame data at time point t and the denoised video frame data at time point t-1; The denoised video frame variation ΔDF is obtained by the difference between the denoised video frame data at time point t and the denoised video frame data at time point t-1; When the threshold adjustment mechanism is triggered, the device state weight factor β and the video frame quality adjustment factor γ (DF(t)) are recalculated according to the device operation state information DS, and the new threshold TH is recalculated. new ; The equipment state weight factor β is obtained according to the deviation between the current value and the historical value of the equipment operation state information DS; The device state weight factor β is obtained by comparing the device operation state information DS and the historical average value DS of the device operation state information. avg Calculation acquisition of Substitute the calculated device state weight factor β(t) into the new threshold TH new In the formula, the new state threshold TH is obtained newB ; When the equipment operation status information DS deviates from the normal working range, the equipment state weight factor β(t) changes, thus affecting the new state threshold TH newB changes occur; The denoised video frame DF is a video frame extracted from the video surveillance system and subjected to denoising. The system evaluates the clarity, brightness and contrast of the video frame to determine the quality level of the current frame DF. According to the quality of the denoised video frame, the video frame quality adjustment factor γ(DF(t)) is calculated; Substitute the calculated video frame quality adjustment factor γ(DF(t)) into the new threshold TH new In the formula, get the new denoising threshold TH newF ; When the video frame quality fluctuates, the video frame quality adjustment factor γ(DF(t)) and the new denoising threshold TH newF Changes linearly to compensate for fluctuations in video frame quality; According to the new state threshold TH newB and the new denoising threshold TH newF Form a threshold set W.

7. The feed production equipment defect identification system based on video monitoring according to claim 1 is characterized in that: The feedback and adjustment module includes a threshold application unit and an effect monitoring and recording unit; The threshold application unit receives the threshold set W from the threshold adjustment module and obtains the comprehensive threshold TH to For defect detection; Through the comprehensive threshold TH to , update the output value of the convolutional neural network model so that the convolutional neural network model is based on the comprehensive threshold TH to Conduct defect detection; During the defect detection process, the system is within a fixed period of time t to Use the current comprehensive threshold TH to Conduct continuous monitoring; The time range within a fixed period t to If the monitoring and detection effect is evaluated as not triggering the threshold adjustment mechanism, the system will restore the initial threshold TH Z , and maintain the initial threshold TH Z For subsequent detection, continue to use the initial threshold TH Z Until the trigger threshold adjustment mechanism is obtained.

8. A feed production equipment defect identification system based on video monitoring according to claim 7, characterized in that: The effect monitoring and recording unit records the comprehensive threshold value TH to The detection accuracy ACnew, false alarm rate FNRnew and false alarm rate FPRnew under this condition; The detection accuracy ACnew is obtained by the ratio of the sum of the number of true positive examples and the number of true negative examples to the number of false positive examples and the number of false negative examples; The false negative rate FNRnew is obtained by the ratio of the number of false negative examples to the number of true positive examples; The false alarm rate FPRnew is obtained by the ratio of the number of false positive examples to the number of true negative examples; The detection results are recorded in the log and a report is generated.

9. The feed production equipment defect identification system based on video monitoring according to claim 1, characterized in that: The report generation module generates a report including text, tables and charts based on the summarized data and analysis results, specifically including: System operation status: Displays the historical change trend of the equipment operation status information DS; Defect statistics: Displays the distribution of defect types, total number and statistical charts of each defect type; Threshold adjustment record: record each threshold adjustment and its impact on detection results; Detection effect: shows the comparison of system detection accuracy ACnew, false alarm rate FNRnew, and false alarm rate FPRnew before and after threshold adjustment.

10. A method for identifying defects in feed production equipment based on video surveillance, applied to a system for identifying defects in feed production equipment based on video surveillance as claimed in any one of claims 1 to 9, characterized in that: The following steps are involved: Step 1: The data acquisition module collects data from the video surveillance system, including video frames F and equipment operation status information DS; Step 2: The defect detection module preprocesses the video frame F collected by the data acquisition module, including denoising the video frame F, using a convolutional neural network to perform defect detection on the denoised video frame F, and obtaining defect information D, and specifically performing defect detection on the feed production equipment through a defect detection algorithm; Step 3: The data analysis module analyzes the frequency and type of defects based on the information from the defect detection module and calculates statistical indicators, including the mean μ Defect and standard deviation σ defect , get the initial threshold TH Z ; Step 4: The threshold adjustment module dynamically adjusts the initial threshold TH of defect detection according to the data analysis module Z , including based on defect frequency FQ i (T) and defect type i, calculate statistical indicators, and adjust the initial detection threshold TH according to the calculation results Z , get the new threshold TH new ; Step 5: Feedback and adjustment module according to the new threshold TH new Changes in the comprehensive threshold TH to Applied to real-time defect detection system and monitor the effect of adjustment, including setting the comprehensive threshold TH to Input into the defect detection model to monitor the comprehensive threshold TH to Check the detection results and record the improvement issues; Step 6: The report generation module generates real-time reports and analysis, showing the comprehensive threshold TH to Adjustment effects, system operation status and defect statistics, summary data information, threshold adjustment records and detection effects.

Images