Soft bag filling manual lamp inspection compliance timing monitoring system

Abstract

The invention discloses a soft bag filling manual lamp inspection compliance timing monitoring system, and relates to the technical field of manual lamp inspection posts of large-capacity soft bag pharmaceutical equipment, and the system comprises an integrity detection module which is used for detecting and transmitting accurate time data corresponding to a signal sent by pressing a microswitch and picture information of the same time; the control program conversion system PLC module is used for calculating a difference value between the accurate time data and a preset threshold value and carrying out label grouping on the difference value; the picture calling and analyzing module is used for calling picture information in a time period corresponding to the label group, screening label group data by using an AI identification technology, and then transmitting ultimate screening data and an identification result to the data display system module; and the data display system module processes and traces the ultimate screening data, and displays manual lamp inspection data, including manual lamp inspection qualified rate, reject ratio, defect distribution, compliance in the manual lamp inspection process and occurrence of abnormal data.

Description

Technical Field

[0001] The patented technology of this invention relates to the manual light inspection post of large-capacity soft bag pharmaceutical equipment, and specifically to a monitoring system for compliance timing of manual light inspection of soft bag filling. Background Art

[0002] The final sterilized products after drug production and filling have been widely used in the pharmaceutical field. In the prior art, large-capacity soft bag filling products have always adopted manual light inspection in the field of light inspection after sterilization. Whether the personnel can fully comply with the light inspection time requirements during the operation is a problem that has been bothering the pharmaceutical industry. Among them, problems such as false detection and missed detection will arise, which will cause hidden dangers of market complaints and product quality accidents caused by unqualified product quality. Therefore, a device with a reasonable structural design and timing of light inspection time during manual light inspection is designed and developed. The time taken by each light inspection employee during the soft bag light inspection is monitored, and the screen record of the manual light inspection is used as a basis for supervision. In this regard, we propose a monitoring system for compliance timing of manual light inspection of soft bag filling. Summary of the invention

[0003] In order to solve the above technical problems, a monitoring system for compliance timing of manual light inspection for soft bag filling is provided. This technical solution solves the above problems.

[0004] In order to achieve the above purpose, the technical solution adopted by the present invention is:

[0005] A monitoring system for compliance timing of manual light inspection for soft bag filling, comprising:

[0006] Integrity detection module: The integrity detection module is used to detect the accurate time data corresponding to the signal sent by pressing the micro switch and the screen information at the same time, and transmit the time data and the screen information at the same time to the control program conversion system PLC module;

[0007] Control program conversion system PLC module: The control program conversion system PLC module is electrically connected to the integrity detection module, and the control program conversion system PLC module is used to calculate the difference between the time data and a preset threshold value, group the abnormal time data by labels according to the difference, and transmit the label group data to the screen retrieval analysis module;

[0008] Screen retrieval and analysis module: The screen retrieval and analysis module is electrically connected to the control program conversion system PLC module, and is used to retrieve the screen information within the time period corresponding to the label grouping data, filter the label grouping data through AI recognition technology, and transmit the final filtering data and AI recognition results to the data display system module;

[0009] Data display system module: The data display system module is electrically connected to the control program conversion system PLC module. The data display system module processes and traces the final screening data, and displays the manual light inspection data including the manual light inspection pass rate, unqualified rate, defect distribution, compliance during the manual light inspection process, and the occurrence of abnormal data.

[0010] Preferably, the integrity detection module specifically includes:

[0011] Micro switch unit: The signals of pressing the micro switch twice at the beginning and end of each manual light inspection represent the completion of the timing of the manual soft bag light inspection;

[0012] Image acquisition unit: Pressing the micro switch twice at the beginning and end of each manual light inspection also completes the image acquisition function. The camera at the manual light inspection station records the image information in the same time period;

[0013] Data detection unit: detects the signal sent by the micro switch and the picture information recorded by the picture acquisition unit, and transmits the signal and picture information to the control program conversion system PLC module.

[0014] Preferably, the control program conversion system PLC module specifically includes:

[0015] Difference calculation unit: compares the signal time period sent by the micro switch with the preset threshold value, and if it exceeds the preset threshold value, it is determined as abnormal data;

[0016] Abnormal data grouping unit: the abnormal data is grouped by labels, the data above the threshold is divided into a light inspection time exceeded group, and the data below the threshold is divided into a light inspection time insufficient group.

[0017] Preferably, the calculation formula of the difference calculation unit specifically includes:

[0018] Calculation formula:

[0019] Δ x =XT

[0020] Δ x It is represented by the difference between the two, the variable X is the signal value of the micro switch pressed (X can be the numerical value corresponding to the specific physical quantity such as time, level, count, etc. obtained each time it is pressed), and the constant T is represented by the preset threshold.

[0021] Preferably, the calculation formula of the abnormal data grouping unit specifically includes:

[0022] Based on the statistical law of normal distribution, the abnormal data is divided, and the difference of micro switch pressing signal N = (μ, σ 2)(where μ is the mean of the differences within the normal range and σ is the standard deviation of the mean)

[0023] The formula for classifying abnormal data is:

[0024] Δ x >μ+3σorΔ x <μ-3σ

[0025] When Δ x When (difference) is greater than μ (the mean of the differences within the normal range) plus 3σ (the standard deviation of the mean), Δ x (difference) is divided into the light inspection time exceeding group in the abnormal data. x When (the difference) is less than μ (the mean of the differences within the normal range) minus 3σ (the standard deviation of the mean), Δ x (Difference) is divided into the insufficient light inspection time group in the abnormal data.

[0026] Preferably, the picture retrieval analysis module specifically includes:

[0027] Picture feature extraction unit: uses a convolutional neural network to extract picture information features within the time period corresponding to the label grouping data. The convolutional neural network achieves the accuracy of extracting the picture information features by learning the shape, color, light and shadow changes of the soft bag;

[0028] Screen object positioning unit: Uses the YOLO series of algorithms to divide the image into grids, predict the category and position of the target in each grid, accurately locate the position of the soft bag, the sealing area, the label area, and the internal liquid part, detect the shape of the soft bag, the sealing flatness, the label integrity, and determine the standardization and working status of the manual light inspection operation process;

[0029] Final screening unit: Based on the recognition results of feature extraction and target positioning, screen abnormal image information, retrieve corresponding label grouping data, and transmit them together to the manual light inspection data display system module.

[0030] Preferably, the convolutional neural network in the picture feature extraction unit specifically includes:

[0031] Convolution layer calculation: used to initially extract image features. Suppose the input image is I, the size is H×W×C (H represents the image height, W represents the image width, and C is the number of channels), the convolution kernel is K, and the size is k h ×k w ×C, the step size is s, the padding is P; the size of the output feature map O is H out ×H put ×N (N is the number of convolution kernels);

[0032] For the element O with coordinates (i, j) in the output feature map O ij (corresponding to the lth convolution kernel), the calculation formula is:

[0033]

[0034] in, is the element of the lth convolution kernel at position (m, n) and channel c, b l is the lth bias term;

[0035] Pooling layer calculation: used to compress data and extract main features, perform average pooling, set the input feature map to be I, the size to be H×W×C (represents the image height, W represents the image width, C is the number of channels), and the pooling window size to be k h ×k w , the step length is s;

[0036] For the element O with coordinates (i, j) in the output feature map ij (corresponding to channel C), the calculation formula is:

[0037]

[0038] Fully connected layer calculation: used for tasks such as integrating features and classification. Let the input feature vector be x, whose size is n, the weight matrix be W, whose size is m×n, and the bias vector be b, whose size is m.

[0039] The calculation formula for the output vector y is:

[0040] y=W x +b

[0041] Preferably, the data display system module specifically includes:

[0042] Data processing unit: used for data sorting and analysis, classifying and archiving label grouping data and abnormal screen information according to product batch, production date and other dimensions, analyzing and calculating the pass rate, unqualified rate and defect distribution of manual light inspection;

[0043] Data tracing unit: Operation and maintenance personnel can query historical data to understand the compliance and abnormal data occurrence during the manual light inspection process over a period of time, providing a basis for quality tracing and improvement;

[0044] Data display unit: displays manual light inspection data in the form of tables, charts, etc., including manual light inspection pass rate, unqualified rate, defect distribution, compliance during manual light inspection, and the occurrence of abnormal data. It supports data export function for operation and maintenance personnel to share data and generate reports.

[0045] Compared with the prior art, the beneficial effects of the present invention are: solving the problem that it is difficult for operators in the pharmaceutical industry to fully comply with the time requirements for the manual light inspection of large-capacity soft bag filled products after sterilization. In view of the fact that false detection and missed detection are prone to occur during the manual light inspection, the occurrence of such situations can be minimized by monitoring the time for the light inspection and retaining the screen records of the light inspection, which helps to reduce market complaints caused by unqualified product quality due to false detection and missed detection, and reduces the risk of patients encountering quality accidents, thereby improving the quality and standardization of the manual light inspection link for soft bag filled products. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 A framework diagram of a monitoring system for compliance timing of manual light inspection for soft bag filling;

[0047] Figure 2 This is the internal system framework diagram of the integrity detection module;

[0048] Figure 3 The internal system framework diagram of the PLC module of the control program conversion system;

[0049] Figure 4 Retrieve the internal system framework diagram of the analysis module for the screen;

[0050] Figure 5 The internal system framework diagram of the system module is displayed for data. DETAILED DESCRIPTION

[0051] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art may think of other obvious variations.

[0052] Reference Figure 1 As shown, a monitoring system for compliance timing of manual light inspection for soft bag filling includes:

[0053] Integrity detection module: The integrity detection module is used to detect the accurate time data corresponding to the signal sent by pressing the micro switch and the screen information at the same time, and transmit the time data and the screen information at the same time to the control program conversion system PLC module;

[0054] Control program conversion system PLC module: The control program conversion system PLC module is electrically connected to the integrity detection module, and the control program conversion system PLC module is used to calculate the difference between the time data and a preset threshold value, group the abnormal time data by labels according to the difference, and transmit the label group data to the screen retrieval analysis module;

[0055] Screen retrieval and analysis module: The screen retrieval and analysis module is electrically connected to the control program conversion system PLC module, and is used to retrieve the screen information within the time period corresponding to the label grouping data, filter the label grouping data through AI recognition technology, and transmit the final filtering data and AI recognition results to the data display system module;

[0056] Data display system module: The data display system module is electrically connected to the control program conversion system PLC module. The data display system module processes and traces the final screening data, and displays the manual light inspection data including the manual light inspection pass rate, unqualified rate, defect distribution, compliance during the manual light inspection process, and the occurrence of abnormal data.

[0057] Reference Figure 2 As shown, the integrity detection module specifically includes:

[0058] Micro switch unit: The signals of pressing the micro switch twice at the beginning and end of each manual light inspection represent the completion of the timing of the manual soft bag light inspection;

[0059] Image acquisition unit: Pressing the micro switch twice at the beginning and end of each manual light inspection also completes the image acquisition function. The camera at the manual light inspection station records the image information in the same time period;

[0060] Data detection unit: detects the signal sent by the micro switch and the picture information recorded by the picture acquisition unit, and transmits the signal and picture information to the control program conversion system PLC module.

[0061] This solution records the exact time of manual light inspection by detecting the signal from the micro switch, and collects image information within the corresponding time through the camera installed at each manual light inspection station, providing a clear and definite basis for the subsequent analysis of abnormal situations in manual light inspection.

[0062] Reference Figure 3 As shown, the control program conversion system PLC module specifically includes:

[0063] Difference calculation unit: compares the signal time period sent by the micro switch with the preset threshold value, and if it exceeds the preset threshold value, it is determined as abnormal data;

[0064] Abnormal data grouping unit: group abnormal data by labels, and divide data above the threshold into a light inspection time exceeded group, and data below the threshold into a light inspection time insufficient group;

[0065] The calculation formula of the difference calculation unit specifically includes:

[0066] Calculation formula:

[0067] Δx =XT

[0068] Δ x It is expressed as the difference between the two, the variable X is the signal value of the micro switch pressed (X can be the numerical value corresponding to the specific physical quantity such as time, level, count, etc. obtained each time it is pressed), and the constant T is expressed as the preset threshold value;

[0069] The calculation formula of the abnormal data grouping unit specifically includes:

[0070] Based on the statistical law of normal distribution, the abnormal data is divided, and the difference of micro switch pressing signal N = (μ, σ 2 )(where μ is the mean of the differences within the normal range and σ is the standard deviation of the mean)

[0071] The formula for classifying abnormal data is:

[0072] Δ x >μ+3σorΔ x <μ-3σ

[0073] When Δ x When (difference) is greater than μ (the mean of the differences within the normal range) plus 3σ (the standard deviation of the mean), Δ x (difference) is divided into the light inspection time exceeding group in the abnormal data. x When (the difference) is less than μ (the mean of the differences within the normal range) minus 3σ (the standard deviation of the mean), Δ x (Difference) is divided into the insufficient light inspection time group in the abnormal data.

[0074] This solution performs difference calculations and abnormal data classification on the collected time data and image information, screens and divides the huge amount of data, and gradually narrows the scope of data that needs to be studied. This greatly shortens the time cost for operation and maintenance personnel to discover, analyze and solve problems, and also serves as a warning for the supervision of compliance of manual lighting inspections.

[0075] Reference Figure 4 As shown, the picture retrieval analysis module specifically includes:

[0076] Picture feature extraction unit: uses a convolutional neural network to extract picture information features within the time period corresponding to the label grouping data. The convolutional neural network achieves the accuracy of extracting the picture information features by learning the shape, color, light and shadow changes of the soft bag;

[0077] Screen object positioning unit: Uses the YOLO series of algorithms to divide the image into grids, predict the category and position of the target in each grid, accurately locate the position of the soft bag, the sealing area, the label area, and the internal liquid part, detect the shape of the soft bag, the sealing flatness, the label integrity, and determine the standardization and working status of the manual light inspection operation process;

[0078] Final screening unit: Based on the recognition results of feature extraction and target positioning, screen abnormal image information, retrieve corresponding label grouping data, and transmit them to the manual light inspection data display system module;

[0079] The convolutional neural network in the picture feature extraction unit specifically includes:

[0080] Convolution layer calculation: used to initially extract image features. Suppose the input image is I, the size is H×W×C (H represents the image height, W represents the image width, and C is the number of channels), the convolution kernel is K, and the size is k h ×k w ×C, the step size is s, the padding is P; the size of the output feature map O is H out ×H put ×N (N is the number of convolution kernels);

[0081] For the element O with coordinates (i, j) in the output feature map O ij (corresponding to the lth convolution kernel), the calculation formula is:

[0082]

[0083] in, is the element of the lth convolution kernel at position (m, n) and channel c, b l is the lth bias term;

[0084] Pooling layer calculation: used to compress data and extract main features, perform average pooling, set the input feature map to be I, the size to be H×W×C (represents the image height, W represents the image width, C is the number of channels), and the pooling window size to be k h ×k w , the step length is s;

[0085] For the element O with coordinates (i, j) in the output feature map ij (corresponding to channel C), the calculation formula is:

[0086]

[0087] Fully connected layer calculation: used for tasks such as integrating features and classification. Let the input feature vector be x, whose size is n, the weight matrix be W, whose size is m×n, and the bias vector be b, whose size is m.

[0088] The calculation formula for the output vector y is:

[0089] y=W x +b

[0090] This solution extracts features and locates targets from the collected image information, determines and analyzes the causes of non-compliant behaviors during manual light inspections, provides a clear and concise reference for the supervision of manual light inspection employees, and provides a scientific and effective teaching case for the training and learning of manual light inspection employees. It is an effective tool to improve the efficiency and quality of manual light inspections.

[0091] Reference Figure 5 As shown, the data display system module specifically includes:

[0092] Data processing unit: used for data sorting and analysis, classifying and archiving label grouping data and abnormal screen information according to product batch, production date and other dimensions, analyzing and calculating the pass rate, unqualified rate and defect distribution of manual light inspection;

[0093] Data tracing unit: Operation and maintenance personnel can query historical data to understand the compliance and abnormal data occurrence during the manual light inspection process over a period of time, providing a basis for quality tracing and improvement;

[0094] Data display unit: displays manual light inspection data in the form of tables and charts, including manual light inspection pass rate, unqualified rate, defect distribution, compliance during manual light inspection, and the occurrence of abnormal data. It supports data export function for operation and maintenance personnel to share data and generate reports.

[0095] This solution receives a large amount of label grouping data and abnormal screen information from the front-end light inspection link every day through the data processing unit, and automatically diverts the data to the corresponding storage area based on the key information of the product batch number and the production date accurate to the hour, minute and second; when the drug quality control department receives market feedback that a batch of drugs may have light inspection risks, the operation and maintenance personnel quickly start the data tracing unit, and by entering the search conditions of the product batch and the approximate production date range, the system retrieves the detailed records of manual light inspections during that period from the massive historical data within seconds. The operation and maintenance personnel can view the operational compliance of each light inspection employee at that time, such as whether the light inspection was completed on time, whether there were any signs of missed inspections or false inspections, the frequency and specific links of abnormal data, etc., and then accurately locate the root cause of the problem, providing a detailed basis for improving the process.

[0096] The use process of the present invention is:

[0097] Step 1: Micro switch unit: The signals of pressing the micro switch twice at the beginning and end of each manual light inspection represent that the timing of the manual soft bag light inspection is completed;

[0098] Step 2: Image acquisition unit: Pressing the micro switch twice at the beginning and end of each manual light inspection also completes the image acquisition function. The camera at the manual light inspection station records the image information in the same time period;

[0099] Step 3: Data detection unit: detects the signal sent by the micro switch and the picture information recorded by the picture acquisition unit, and transmits the signal and picture information to the control program conversion system PLC module.

[0100] Step 4: Difference calculation unit: compare the signal time period sent by the micro switch with the preset threshold value, and if it exceeds the preset threshold value, it is determined as abnormal data;

[0101] Step 5: Abnormal data grouping unit: group abnormal data by labels, divide data above the threshold into a light inspection time exceeded group, and divide data below the threshold into a light inspection time insufficient group.

[0102] Step 6: Picture feature extraction unit: Use convolutional neural network to extract picture information features in the time period corresponding to the label grouping data. The convolutional neural network achieves the accuracy of extracting the picture information features by learning the shape, color, light and shadow changes of the soft bag;

[0103] Step 7: Screen object positioning unit: Use the YOLO series algorithm to divide the image into grids, predict the category and position of the target in each grid, accurately locate the position of the soft bag, the sealing area, the label area and the internal liquid part, detect the shape of the soft bag, the sealing flatness, the label integrity, and judge the standardization and working status of the manual light inspection operation process;

[0104] Step 8: Final screening unit: based on the recognition results of feature extraction and target positioning, screen abnormal image information, retrieve corresponding label group data, and transmit them to the manual light inspection data display system module. The characteristic is that the data display system module specifically includes:

[0105] Step 9: Data processing unit: used for data sorting and analysis, classifying and archiving label grouping data and abnormal screen information according to product batch, production date and other dimensions, analyzing and calculating the pass rate, unqualified rate and defect distribution of manual light inspection;

[0106] Step 10: Data tracing unit: Operation and maintenance personnel can query historical data to understand the compliance and abnormal data occurrence during the manual light inspection process over the past period of time, providing a basis for quality tracing and improvement;

[0107] Step 11: Data display unit: displays manual light inspection data in the form of tables, charts, etc., including manual light inspection pass rate, unqualified rate, defect distribution, compliance during manual light inspection, and occurrence of abnormal data. It supports data export function for operation and maintenance personnel to share data and generate reports.

[0108] To sum up, the advantages of the present invention are: it solves the problem that it is difficult for operators in the pharmaceutical industry to fully comply with the time requirements for the manual light inspection of large-capacity soft bag filled products after sterilization. In view of the fact that false detection and missed detection are prone to occur during the manual light inspection, the occurrence of such situations can be minimized by monitoring the time for the light inspection and retaining the screen records of the light inspection, which helps to reduce market complaints caused by unqualified product quality due to false detection and missed detection, and reduces the risk of patients encountering quality accidents, thereby improving the quality and standardization of the manual light inspection link of soft bag filled products.

[0109] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions only describe the principles of the present invention. The present invention may be subject to various changes and improvements without departing from the spirit and scope of the present invention. These changes and improvements fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the attached claims and their equivalents.

Claims

1. A monitoring system for compliance timing of manual light inspection of soft bag filling, characterized by: Integrity detection module: The integrity detection module is used to detect the accurate time data corresponding to the signal sent by pressing the micro switch and the screen information at the same time, and transmit the time data and the screen information at the same time to the control program conversion system PLC module; Control program conversion system PLC module: The control program conversion system PLC module is electrically connected to the integrity detection module, and the control program conversion system PLC module is used to calculate the difference between the time data and a preset threshold value, group the abnormal time data by labels according to the difference, and transmit the label group data to the screen retrieval analysis module; Screen retrieval and analysis module: The screen retrieval and analysis module is electrically connected to the control program conversion system PLC module, and is used to retrieve the screen information within the time period corresponding to the label grouping data, filter the label grouping data through AI recognition technology, and transmit the final filtering data and AI recognition results to the data display system module; Data display system module: The data display system module is electrically connected to the control program conversion system PLC module. The data display system module processes and traces the final screening data, and displays the manual light inspection data including the manual light inspection pass rate, unqualified rate, defect distribution, compliance during the manual light inspection process, and the occurrence of abnormal data.

2. A monitoring system for compliance timing of manual light inspection of soft bag filling according to claim 1, characterized in that: The integrity detection module specifically includes: Micro switch unit: The signals of pressing the micro switch twice at the beginning and end of each manual light inspection represent the completion of the timing of the manual soft bag light inspection; Image acquisition unit: Pressing the micro switch twice at the beginning and end of each manual light inspection also completes the image acquisition function. The camera at the manual light inspection station records the image information in the same time period; Data detection unit: detects the signal sent by the micro switch and the picture information recorded by the picture acquisition unit, and transmits the signal and picture information to the control program conversion system PLC module.

3. A monitoring system for compliance timing of manual light inspection of soft bag filling according to claim 2, characterized in that: The control program conversion system PLC module specifically includes: Difference calculation unit: compares the signal time period sent by the micro switch with the preset threshold value, and if it exceeds the preset threshold value, it is determined as abnormal data; Abnormal data grouping unit: the abnormal data is grouped by labels, the data above the threshold is divided into a light inspection time exceeded group, and the data below the threshold is divided into a light inspection time insufficient group.

4. A monitoring system for compliance timing of manual light inspection of soft bag filling according to claim 3, characterized in that: The calculation formula of the difference calculation unit specifically includes: Calculation formula: Δ x =X-T Δ x It is represented by the difference between the two, the variable X is the signal value of the micro switch pressed (X can be the numerical value corresponding to the specific physical quantity such as time, level, count, etc. obtained each time it is pressed), and the constant T is represented by the preset threshold.

5. A monitoring system for compliance timing of manual light inspection of soft bag filling according to claim 3, characterized in that: The calculation formula of the abnormal data grouping unit specifically includes: Based on the statistical law of normal distribution, the abnormal data is divided, and the difference of micro switch pressing signal N = (μ, σ 2 )(where μ is the mean of the differences within the normal range and σ is the standard deviation of the mean) The formula for classifying abnormal data is: D x >μ+3σ or Δ x <μ-3σ When Δ x When (difference) is greater than μ (the mean of the differences within the normal range) plus 3σ (the standard deviation of the mean), Δ x (difference) is divided into the light inspection time exceeding group in the abnormal data. x When (the difference) is less than μ (the mean of the differences within the normal range) minus 3σ (the standard deviation of the mean), Δ x (Difference) is divided into the insufficient light inspection time group in the abnormal data.

6. A monitoring system for compliance timing of manual light inspection of soft bag filling according to claim 5, characterized in that: The picture retrieval analysis module specifically includes: Picture feature extraction unit: uses a convolutional neural network to extract picture information features within the time period corresponding to the label grouping data. The convolutional neural network achieves the accuracy of extracting the picture information features by learning the shape, color, light and shadow changes of the soft bag; Screen object positioning unit: Uses the YOLO series of algorithms to divide the image into grids, predict the category and position of the target in each grid, accurately locate the position of the soft bag, the sealing area, the label area, and the internal liquid part, detect the shape of the soft bag, the sealing flatness, the label integrity, and determine the standardization and working status of the manual light inspection operation process; Final screening unit: Based on the recognition results of feature extraction and target positioning, screen abnormal image information, retrieve corresponding label grouping data, and transmit them together to the manual light inspection data display system module.

7. A monitoring system for compliance timing of manual light inspection of soft bag filling according to claim 6, characterized in that: The convolutional neural network in the picture feature extraction unit specifically includes: Convolution layer calculation: used to initially extract image features. Suppose the input image is I, the size is H×W×C (H represents the image height, W represents the image width, and C is the number of channels), the convolution kernel is k, and the size is k h ×k w ×C, the step size is (s h ,s w ); For the element with coordinates (n, i, j) in the output feature map O (corresponding to the nth convolution kernel), the calculation formula is: m and p are the coordinate indices in the height and width directions of the convolution kernel respectively; and w ×j+p is the coordinate corresponding to the input image, sh and s w They are the steps of the pooling window in the height and width directions respectively; Pooling layer calculation: used to compress data and extract main features, perform average pooling, set the input feature map to be I, the size to be H×W×C (represents the image height, W represents the image width, C is the number of channels), and the pooling window size to be k h ×k w , the step length is (s h ,s w ); For the element with coordinates (c, i, j) in the output feature map (corresponding to channel C), the calculation formula is: i and j are the coordinate indices of the output feature map in the height and width directions respectively; k h and k w are the sizes of the pooling window in the height and width directions respectively; m and p are the coordinate indexes in the height and width directions of the pooling window respectively; and w ×j+p is the coordinate corresponding to the input image; s h and w They are the steps of the pooling window in the height and width directions respectively; Fully connected layer calculation: used for tasks such as integrating features and classification. Let the input feature vector be x, the output vector be y, the weight matrix be W, and the bias vector be b. The calculation formula is: y=W x +b 8. A monitoring system for compliance timing of manual light inspection of soft bag filling according to claim 6, characterized in that: The data display system module specifically includes: Data processing unit: used for data sorting and analysis, classifying and archiving label grouping data and abnormal screen information according to product batch, production date and other dimensions, analyzing and calculating the pass rate, unqualified rate and defect distribution of manual light inspection; Data tracing unit: Operation and maintenance personnel can query historical data to understand the compliance and abnormal data occurrence during the manual light inspection process over a period of time, providing a basis for quality tracing and improvement; Data display unit: displays manual light inspection data in the form of tables and charts, including manual light inspection pass rate, unqualified rate, defect distribution, compliance during manual light inspection, and the occurrence of abnormal data. It supports data export function for operation and maintenance personnel to share data and generate reports.

Images