Thermo-sensitive excitation image analysis and identification system

Abstract

The invention discloses a thermosensitive excitation image analysis and recognition system, which belongs to the technical field of image analysis, and comprises the following steps of: monitoring the yield of a production line in real time, dynamically triggering sampling detection, generating a temperature control instruction when a sampling instruction is triggered, and controlling a thermosensitive head to implement stepped preheating and grouped thermal excitation; synchronously collecting sample surface temperature for monitoring and early warning; a multi-stage image acquisition strategy is made according to excitation parameters, response time and non-uniformity are analyzed after a thermosensitive image sequence is obtained, quality judgment and classified output are achieved, and the reject ratio is counted for production line early warning; through real-time monitoring, precise thermal excitation, multi-dimensional image analysis and closed-loop feedback, the detection precision and the production stability are improved, the rate of unqualified products is effectively reduced, and intelligent quality control is achieved.

Description

Technical Field

[0001] This invention relates to the field of image analysis technology, and more specifically to a thermally excited image analysis and recognition system. Background Technology

[0002] Dry thermal film, as an important imaging medium, plays a crucial role in fields such as medical and industrial inspection. In the medical industry, it is a core consumable for medical imaging diagnosis. Images generated by X-ray examinations, CT scans, and MRI all require dry thermal film for clear visualization, providing doctors with accurate diagnostic information. Its imaging quality directly affects diagnostic accuracy and testing reliability.

[0003] The imaging quality of dry thermal films depends on their thermal response characteristics, namely, the dynamic color development ability of the coating material under different temperature and time parameters. Traditional detection methods suffer from the following core technical problems: existing thermal performance tests cannot simulate the real response characteristics under dynamic changes in printhead temperature, printing speed, etc., during actual printing, leading to significant deviations between test results and clinical application scenarios; single-spectrum machine vision-based defect detection technology struggles to capture the microscopic response differences of the coating material during thermal excitation, lacking effective means to identify latent defects such as uneven thermal layer thickness and abnormal crystal phase distribution; the detection process is independent of the production line temperature control system, making it impossible to obtain the correlation between thermal excitation parameters and image response data in real time, and difficult to establish a mapping model between thermal response characteristics and production process parameters. Consequently, traditional detection methods cannot accurately assess the dynamic thermal performance of the film, nor can they guide the precise control of the production process, thus restricting the quality of dry thermal films. Therefore, to overcome these limitations, this invention proposes a thermal excitation image analysis and recognition system. Summary of the Invention

[0004] To address the shortcomings of existing technologies, the present invention aims to provide a thermal excitation image analysis and recognition system. By triggering controllable temperature thermal excitation detection through dynamic sampling rules, and combining stepped preheating, grouped thermal excitation, and real-time surface temperature monitoring technologies, a multi-level image acquisition scheme and thermal image sequence analysis method are employed to achieve real-time assessment of film quality and early warning for the production line.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A thermally excited image analysis and recognition system, applied to dry thermal film quality inspection equipment, for detecting the quality of dry thermal film, includes:

[0007] The monitoring production line determines whether a sampling command is triggered. If triggered, the target film sample is sampled and a temperature control command is generated. The thermal head is controlled to preheat the detection plate in stages, and a controllable temperature thermal excitation is applied to the target film sample group based on the matched thermal excitation curve.

[0008] By monitoring and identifying feature points of the thermal excitation curve parameters under controllable temperature excitation through slope mutation, a multi-level image acquisition scheme is constructed to acquire the first image, obtain the sample thermal image sequence, and screen the sample thermal image sub-sequence in the heating stage. By configuring the gray change rate threshold, cross-correlation analysis is performed on the image gray change rate of the sample thermal image sub-sequence to obtain the response time of the target film.

[0009] Extract the grayscale matrix of each group of sample thermal images in the sample thermal image sequence to segment the thermal head excitation area in the sample thermal image and obtain the non-uniformity of the sample thermal image; and obtain the edge sharpness of the sample thermal image by calculating the gradient magnitude matrix of the sample thermal image to determine whether the target film sample has qualified thermal response.

[0010] If qualified, the target film sample is flipped over, and a second image acquisition is performed to obtain the image of the back of the sample and select its comparison image to generate a grayscale difference image. By verifying the consistency of the grayscale difference image on both sides, it is determined whether the quality of the target film sample is qualified. The quality judgment result of the target film sample is classified and output, and the frequency of the production line quality failure is counted to provide early warning for the production line.

[0011] Specifically, the steps for applying controllable temperature thermal excitation to the target film sample grouping based on the matched thermal excitation curve include:

[0012] The temperature control zone is divided, and a stepped heating mode is used to preheat the test plate in each temperature control zone; the surface temperature distribution of the test plate in each temperature control zone is obtained, and the temperature standard deviation of the global temperature control zone is calculated by constructing a temperature distribution matrix;

[0013] Configure a temperature deviation threshold, determine whether the temperature standard deviation of the global temperature control area is greater than the temperature deviation threshold. If it is greater than the temperature deviation threshold, locate the deviation from the temperature control area according to the stepped temperature, and adjust the power output of the thermal head that is deviating from the temperature control area to dynamically compensate for the deviation from the temperature control area, so as to control the temperature standard deviation of the global temperature control area to be less than or equal to the temperature deviation threshold.

[0014] The matching thermal excitation curve parameters are called. When the target film sample arrives at the center of the detection plate, the surface temperature of the target film sample is collected. Based on the rate of change of the surface temperature of the target film sample, it is determined whether to start thermal excitation.

[0015] Specifically, the steps for applying controllable temperature thermal excitation to the target film sample grouping based on the matched thermal excitation curve also include:

[0016] The number of test groups is set to determine the number of groups for the thermal heads in each temperature control area, and the thermal heads in the temperature control area are grouped according to the grouping rules.

[0017] The grouping rules include: the number of groups in each temperature control area is equal to the number of test groups; each group contains at least one thermal head; the total number of thermal heads in all groups is less than the total number of thermal heads in that temperature control area; and the number of thermal heads in each group is determined randomly.

[0018] Based on the grouping results of the thermal heads, the temperature of the thermal heads is controlled according to the thermal excitation curve parameters to apply thermal excitation to the target film sample, and the surface temperature of the target film sample is monitored in real time during the thermal excitation process.

[0019] Configure a temperature deviation threshold to determine whether the deviation between the surface temperature of the target film sample and the temperature of the corresponding thermal head is greater than the temperature deviation threshold. If it is greater than the temperature deviation threshold, adjust the power output of the corresponding thermal head to perform dynamic temperature compensation. If temperature compensation fails, trigger a thermal excitation abnormality warning.

[0020] Specifically, the steps for constructing a multi-level image acquisition scheme include:

[0021] Obtain the parameters of the thermal excitation curve and perform smooth fitting on the data points of the thermal excitation curve to generate a continuous temperature-time curve;

[0022] The second derivative of the temperature-time curve is calculated, and the characteristic points of the temperature-time curve are identified by slope change detection.

[0023] The acquisition stages are divided according to the feature points of the thermal excitation curve to set the image acquisition frequency of each acquisition stage, and the number of image frames acquired is calculated according to the duration of each acquisition stage.

[0024] Plan the sequence of image acquisition and integrate the acquisition stages, acquisition frequency, number of image frames acquired, and acquisition sequence to construct a multi-level image acquisition scheme.

[0025] Specifically, the steps for obtaining the sample thermal image sequence include:

[0026] Based on a multi-level image acquisition scheme, the first image of the target film sample is acquired to obtain the thermal image sequence of each sample. The thermal response analysis of the acquired thermal image sequence is then performed to determine whether the thermal response of the target film sample is qualified.

[0027] If the thermal response is unqualified, the target film sample is deemed to be of unqualified quality; if the thermal response is qualified, the target film sample is flipped over, a second image is acquired, and the image of the back of the sample is obtained for quality defect analysis to determine whether the target film sample is of qualified quality.

[0028] Specifically, the steps for determining whether a target film sample meets the thermal response requirements include:

[0029] After preprocessing the acquired sample thermal image sequence frame by frame, the sample thermal image subsequence in the heating stage is selected according to the acquisition stage marker. Cross-correlation analysis is performed to calculate the time difference from the moment thermal excitation is applied to the moment the image grayscale change reaches a stable state, which is taken as the response time of the target film under the action of the thermal head.

[0030] Configure a response time threshold. If the response time of the target film sample is greater than the response time threshold when a set of thermal heads applies controllable temperature thermal excitation, the thermal response of the target film sample is deemed unqualified.

[0031] Extract the gray value matrix of each preprocessed sample thermal image, and segment the thermal head excitation region of the sample thermal image according to the region growing algorithm. Calculate the gray value variation coefficient of the thermal head excitation region as the non-uniformity of the sample thermal image.

[0032] Configure a uniformity threshold. If the non-uniformity of the thermal image of a sample exceeds the uniformity threshold, the thermal response of the target film sample is deemed unqualified.

[0033] Specifically, the steps for determining whether a target film sample has a qualified thermal response also include:

[0034] The gradient magnitude matrix of the preprocessed thermal image of each sample is calculated, and the edge spread function of the gradient magnitude matrix is ​​estimated by Gaussian blur, which is used to quantify the edge blur of the sample thermal image as the edge sharpness of the sample thermal image.

[0035] Configure a sharpness threshold. If the edge sharpness of a sample thermal image is greater than the sharpness threshold, the target film sample is deemed to have an unqualified thermal response.

[0036] If the thermal response of the target film sample is deemed unqualified, the thermal head stops applying controllable temperature thermal excitation and records the image features and timestamp of the unqualified target film sample.

[0037] If the thermal response of the target film sample is not deemed unqualified at the end of the thermal response analysis of the sample thermal image sequence, then the thermal response of the target film sample is deemed qualified.

[0038] Specifically, the steps for obtaining the back image of the sample for quality defect analysis include:

[0039] The uniformity of the back image of the sample is checked through image preprocessing. The non-uniformity of the back image is calculated. If the non-uniformity of the back image is greater than the uniformity threshold, the thermal response of the target film sample is deemed unqualified. Otherwise, double-sided consistency verification is performed.

[0040] The last thermal image in the sample thermal image sequence is flipped so that it is oriented in the same direction as the back image of the sample, and used as a reference image for the back image of the sample.

[0041] Feature points are extracted and matched from the back image of the sample and its control image. The images are then aligned, and the grayscale difference between corresponding pixels in the aligned back image of the sample and the control image is obtained to generate a grayscale difference image.

[0042] Configure an abnormal deviation threshold, filter pixels whose pixel values ​​in the grayscale difference image are greater than the abnormal deviation threshold, and use them as difference points. Calculate the proportion of difference points to the total number of pixels in the grayscale difference image by counting the number of difference points.

[0043] Configure a consistency threshold to determine whether the proportion of difference points to the total number of pixels in the grayscale difference image is greater than the consistency threshold. If it is greater, the double-sided consistency verification is deemed to have failed, and the thermal response of the target film sample is deemed to be unqualified; otherwise, the double-sided consistency verification is deemed to have succeeded, and the thermal response of the target film sample is deemed to be qualified.

[0044] Specifically, the steps for determining whether a sampling instruction has been triggered include:

[0045] Set up a sampling counter to record the production output data of the production line in each round of sampling inspection. Initialize the sampling counter and obtain the real-time production output of film based on the collected production data to update the sampling counter.

[0046] Configure the sampling interval threshold, compare the real-time data of the sampling counter with the sampling interval threshold, determine whether the real-time data of the sampling counter is equal to the sampling interval threshold, if they are equal, determine to trigger a sampling command and clear the sampling counter to zero.

[0047] When no sampling instruction is triggered, detect fluctuations in production data, including:

[0048] Real-time analysis of production line pressure data; based on historical production line pressure data, setting a pressure fluctuation range; determining whether the production line pressure data exceeds the pressure fluctuation range; if it does, marking it as an abnormal pressure fluctuation and triggering a sampling command to clear the sampling counter.

[0049] Simultaneously, the production line speed data is monitored in real time, a change rate threshold is configured, and it is determined whether the real-time change rate of the production line speed is greater than the change rate threshold. If it is greater, it is marked as an abnormal speed change, and a sampling instruction is triggered to clear the sampling counter.

[0050] Specifically, the steps for determining whether a sampling instruction has been triggered also include:

[0051] When a sampling command is triggered, the sampling interval threshold update process is initiated, i.e.:

[0052] Obtain the quality inspection results of the sampled target film. If the quality of the target film sample is determined to be qualified, maintain the sampling inspection threshold.

[0053] If the quality of the target film sample is deemed unqualified, the sampling interval threshold will be adjusted to the minimum sampling interval unit.

[0054] Configure an adjustment threshold to measure the duration of monitoring. Within the adjustment threshold, calculate the pass rate of the sampled target film. Set a pass rate threshold and determine whether the pass rate within the adjustment threshold is greater than the pass rate threshold. If it is greater, restore the sampling detection threshold; otherwise, issue a production quality warning and stop the production line.

[0055] The beneficial effects of this invention are:

[0056] 1. By dynamically capturing the parameters of the controllable temperature-induced thermal excitation curve, the thermal response of the film under thermal excitation is accurately grasped. Simultaneously, real-time surface temperature monitoring technology provides a comprehensive understanding of the details of surface temperature changes. Based on this, multi-level image sequence analysis is combined to deeply analyze the dynamic color development process of the thermal layer from both temporal and spatial dimensions. This analytical method overcomes the limitations of traditional single-spectral detection, effectively identifying latent defects such as uneven coating thickness. Traditional single-spectral detection often only provides information in a single dimension, making it difficult to detect defects inside the film or at the microscopic level. This solution, through multi-dimensional analysis, significantly improves the accuracy and comprehensiveness of defect detection.

[0057] 2. Through a multi-level image acquisition scheme, multiple frames of sample thermal images are acquired at specific frequencies during different stages of thermal excitation, such as heating and isothermal states, constructing a dynamic thermal response dataset. After preprocessing the acquired sample thermal images, the thermal response characteristics are analyzed from multiple dimensions. When analyzing response time, cross-correlation analysis is used to calculate the time difference from the application of thermal excitation to the image grayscale reaching a stable state, accurately evaluating thermal performance. Regarding non-uniformity detection, a region growing algorithm is used to segment the thermal head excitation area, and the grayscale variation coefficient is calculated, effectively identifying non-uniformity issues in the thermal head excitation area. When quantifying edge sharpness, the gradient magnitude matrix is ​​calculated, and the edge diffusion function is estimated using Gaussian blur, accurately determining the degree of image edge blur. These image processing techniques work together to overcome the limitations of traditional single-spectral detection, effectively identifying latent defects such as uneven coating thickness and abnormal crystal phase distribution, greatly improving the accuracy and comprehensiveness of thermal defect detection. Attached Figure Description

[0058] Figure 1 This is a schematic diagram of the structure of a thermally stimulated image analysis and recognition system according to the present invention;

[0059] Figure 2 This is a flowchart illustrating the specific steps involved in determining whether a sampling instruction has been triggered according to the present invention.

[0060] Figure 3 A flowchart illustrating the specific steps involved in continuously detecting fluctuations in production data according to this invention;

[0061] Figure 4 This is a flowchart illustrating the specific steps of applying controllable temperature and heat stimulation to the target film sample grouping according to the present invention;

[0062] Figure 5 A flowchart illustrating the specific steps involved in developing a multi-level image acquisition scheme for this invention;

[0063] Figure 6 This is a flowchart illustrating the specific steps involved in determining whether a target film sample meets the thermal response requirements according to the present invention. Detailed Implementation

[0064] Please see Figure 1 This embodiment introduces a thermal excitation image analysis and recognition system, including: a sampling module, a thermal excitation module, a visual detection module, and a control module;

[0065] In this embodiment, the sampling module is configured as follows:

[0066] The system monitors the operation status of the dry thermal film production line in real time using sensors and collects production data. Based on sampling trigger conditions or real-time quality inspection results, it dynamically sets sampling interval thresholds to determine whether a sampling command is triggered and to sample and inspect the film on the production line.

[0067] Sensors include: photoelectric sensors, pressure sensors, and speed sensors;

[0068] Production data includes: production output data, production time data, production quality data, production line speed data, and production line pressure data;

[0069] In this embodiment, photoelectric sensors on the dry thermal film production line monitor the position information of the dry thermal film and collect production data. Sampling trigger conditions are set based on time intervals, production quantities, and abnormal conditions. For example, a sampling operation is triggered every 10 minutes of production according to a preset time interval. When the set time is reached, a sampling command is sent to the robotic arm. The photoelectric sensors count the number of films passing through in real time. When the cumulative production quantity reaches a preset value, such as sampling every 500 films produced, the counting signal is judged. If the condition is met, a sampling operation is triggered. When abnormal pressure fluctuations or sudden changes in conveyor belt speed are detected, a sampling operation is triggered to check for films that may have quality problems.

[0070] Please see Figure 2 Preferably, the specific steps for determining whether a sampling instruction has been triggered include:

[0071] A sampling counter is set up to record the production output data of the production line in each round of sampling inspection. The sampling counter is initialized to zero. Based on the collected production data, the real-time production output of dry thermal film on the production line is obtained, and the sampling counter is updated. This provides an initial state for subsequent accurate production data statistics, ensuring that the production statistics of each sampling cycle start from 0, which facilitates the clear definition of the production range of each round of sampling and provides basic data support for determining whether the sampling interval threshold has been reached.

[0072] A sampling interval threshold is configured. Within each round of sampling inspection, the real-time data of the sampling counter is compared with the sampling interval threshold. If the real-time data of the sampling counter equals the threshold, a sampling command is triggered, and the sampling counter is reset to zero, preparing for a new round of sampling inspection. By setting a fixed sampling interval threshold, sampling is ensured to be performed at predetermined production intervals under normal production conditions, guaranteeing the regularity and periodicity of sampling. Regular sampling helps to promptly identify potential quality problems in the production process, preventing the batch production of defective products. Simultaneously, resetting the sampling counter provides a clean counting environment for the next round of statistics, ensuring the independence and continuity of the sampling cycle.

[0073] And within each round of sampling inspection, i.e., when no sampling instruction is triggered, please refer to [link / reference]. Figure 3 Continuously monitor fluctuations in production data, including:

[0074] The system analyzes production line pressure data collected by pressure sensors in real time. Based on historical production line pressure data, it obtains the average and standard deviation of production pressure to set the pressure fluctuation range. If the real-time production line pressure data collected by the pressure sensors exceeds the pressure fluctuation range, it is marked as an abnormal pressure fluctuation, triggering a sampling command and resetting the sampling counter. By using historical data and statistical principles to set the pressure fluctuation range, it can effectively identify abnormal pressure situations. Once an abnormal pressure fluctuation occurs, a sampling command is triggered, allowing for timely inspection of films that may be affected by pressure issues. This prevents defective products caused by unstable pressure from entering subsequent processes. At the same time, resetting the sampling counter allows production statistics to restart, facilitating the tracking of production under abnormal pressure.

[0075] The system monitors production line speed data collected by speed sensors in real time, configures a rate of change threshold, and calculates the real-time rate of change of production line speed. If the rate of change exceeds the threshold, it is marked as an abnormal speed change, triggering a sampling command and resetting the sampling counter to zero. By monitoring the rate of change of speed, abnormal fluctuations in production line speed can be accurately detected. Speed ​​anomalies may affect the stability and quality of film production. Triggering a sampling command in this situation allows for timely inspection of the relevant film quality. Resetting the sampling counter helps to accurately count the output during speed anomalies, providing data for subsequent analysis of the impact of speed anomalies on output and quality.

[0076] Once a sampling command is triggered, the sampling interval threshold update process is initiated, i.e.:

[0077] The quality inspection results of the target film samples are obtained. If the target film sample is deemed to be of acceptable quality, the sampling inspection threshold for the current round of sampling is maintained. When the quality is acceptable, the sampling interval threshold remains stable, thus maintaining the stability and consistency of sampling during the production process. This indicates that the current sampling strategy and production process are relatively stable, requiring no adjustment to the sampling frequency, which helps maintain a balance between production rhythm and quality monitoring.

[0078] If the target film sample is determined to be substandard, the sampling interval threshold is adjusted to the minimum sampling interval unit. An adjustment threshold is configured to measure the duration of monitoring. Within this threshold, the pass rate of the sampled target film samples is statistically analyzed, and a pass rate threshold is set. If the pass rate exceeds the threshold, the sampling inspection threshold is restored; otherwise, a production quality warning is issued, and the production line is stopped. When quality is substandard, the sampling interval is reduced to the minimum sampling interval unit, thus maximizing the detection of problematic products and preventing further production of substandard goods. By statistically analyzing the pass rate within the adjustment threshold and comparing it with the threshold, a decision is made on whether to restore the original sampling interval threshold or stop the production line. This allows for timely response to quality issues and the resumption of normal sampling when production line quality improves, balancing quality control and production efficiency. Stopping the production line if the pass rate is below the target prevents the production of more substandard products, reduces losses, and allows for timely identification of the root cause of production problems.

[0079] When a sampling command is triggered, the sampling module drives the gripping robotic arm to identify, locate and pick up the target film sample. Through the coordinated movement of the pushing component, the target film sample is transported to the detection plate.

[0080] In this embodiment, after receiving the sampling instruction, the industrial camera of the gripping robotic arm is activated to acquire images of the film on the conveyor belt. Combined with the position data provided by the photoelectric sensor, the position of the target film is identified, the coordinates of the dry thermal film sample are determined, and after gripping the film, the robotic arm places it on the conveyor belt. The conveyor belt transports the sample to the detection plate at a preset speed. During the transport process, the speed sensor monitors the speed of the conveyor belt in real time to ensure the stability and accuracy of the transport.

[0081] Preferably, the thermal excitation module is integrated below the detection plate and includes a thermal head array and a temperature feedback unit. The detection plate has a built-in visible light excitation module, including a uniformly distributed LED light source array and a light homogenizing layer. The LED light source array emits visible light, and the light homogenizing layer is used to homogenize the light emitted by the LED light source array to illuminate the film on the detection plate, so that the thermal changes of the film can form a clear image through visible light reflection or transmission. During the detection process, when the thermal head applies controllable temperature thermal excitation to the film, the light-emitting elements of the detection plate start to work. As the thermal layer of the film undergoes color changes under thermal excitation, its absorption and reflection characteristics of visible light will also change accordingly.

[0082] When a sampling command is triggered, the thermal head array is activated. The thermal head array is used to preheat the detection plate in stages. Based on the grouping results of the thermal head array, a controllable temperature thermal excitation is applied to the target film sample group. The temperature feedback unit collects the surface temperature of the detection plate and the target film sample in real time for temperature monitoring and early warning. This ensures the stability and accuracy of the thermal excitation and provides a reliable test image basis for subsequent accurate image detection.

[0083] Please see Figure 4 Preferably, the specific steps for applying controllable temperature stimulation to the target film sample grouping include:

[0084] The array of thermal heads beneath the detection plate is divided into multiple independent temperature-controlled zones. A stepped heating mode is used for each zone to preheat the detection plate. The temperature-controlled zones are defined based on the size of the detection plate, the distribution of the thermal heads, and their thermal conductivity characteristics; for example, a large-area detection plate might be divided into nine temperature-controlled zones in a nine-square grid. Stepped heating parameters are set for each zone, such as an initial temperature of T0, a step increase of ΔT, and a heating time interval of t. According to these parameters, each temperature-controlled zone sequentially heats from T0 to T0+ΔT, holds for t, then heats to T0+2ΔT, and so on, completing the stepped heating preheating process.

[0085] After each temperature step is completed, the surface temperature distribution of the detection plate in each temperature-controlled area is acquired using an infrared thermal imager to calculate the temperature standard deviation of the entire temperature-controlled area. The infrared thermal imager accurately collects temperature data from multiple sampling points within each temperature-controlled area to construct a temperature distribution matrix. Using statistical methods, the temperature standard deviation of the entire temperature-controlled area is calculated based on the temperature data from each temperature-controlled area, thereby quantifying the uniformity of the surface temperature of the detection plate.

[0086] A temperature deviation threshold is configured. If the standard deviation of the temperature in the global temperature control area exceeds the threshold, the area deviating from the temperature control is located based on the stepped temperature. The power output of the thermistor in this deviating area is adjusted to dynamically compensate for the surface temperature of the detection plate in that area until the standard deviation of the temperature in the global temperature control area is less than or equal to the temperature deviation threshold. By comparing the temperature data of each temperature control area with the average temperature, areas with large temperature deviations are identified. For these deviating areas, the power output of the thermistor is adjusted proportionally according to the degree of temperature deviation; for example, power is reduced in areas with excessively high temperatures and increased in areas with excessively low temperatures. After each adjustment, temperature data is collected again using an infrared thermal imager, and the temperature standard deviation is recalculated. This adjustment continues until the standard deviation of the temperature in the global temperature control area is less than or equal to the temperature deviation threshold.

[0087] Based on the batch information of the dry thermal film, the matching thermal excitation curve parameters, including heating rate, isothermal temperature, heating period duration and isothermal period duration, are retrieved from the production information database.

[0088] The production information database stores the standard parameters for thermal excitation corresponding to different batches of film, ensuring that the thermal excitation process is compatible with the characteristics of the film and improving the accuracy and reliability of the test images.

[0089] When the target film sample arrives at the center of the detection plate, the surface temperature of the target film sample is collected by the temperature feedback unit. When the rate of change of the surface temperature of the target film sample is zero, thermal excitation is started. The zero rate of change of the surface temperature of the target film sample indicates that the sample has reached thermal equilibrium with the detection plate. At this time, thermal excitation is started to ensure that the thermal excitation process is not affected by the initial temperature difference of the sample.

[0090] The number of test groups is set to determine the number of groups for the thermal heads within each temperature control zone. The thermal heads within each temperature control zone are randomly grouped, adhering to the following rules: the number of groups within each temperature control zone is equal to the number of test groups; each group contains at least one thermal head; the total number of thermal heads in all groups is less than the total number of thermal heads in that temperature control zone; and the specific number of thermal heads in each group is randomly determined. Random grouping simulates the working conditions of different thermal head combinations, avoiding the regular errors that may result from fixed grouping. This makes the thermal excitation test more comprehensive and objective, better reflecting the thermal excitation effect of different thermal head combinations on the target film sample. Ensuring that all thermal heads within each temperature control zone participate in the grouped test, and that each group contains at least one thermal head, allows for a comprehensive examination of the working performance of each thermal head and different combinations of numbers of thermal heads, providing diverse data support for subsequent thermal excitation control.

[0091] Based on the grouping results of the thermal heads, each group of thermal heads is controlled according to the parameters of the thermal excitation curve. Specifically, the temperature of the thermal head is raised to the target temperature at a set heating rate, and after reaching the target temperature, the temperature is maintained at a constant temperature for a preset holding time. Thermal excitation control is performed sequentially on each group of thermal heads, facilitating comparative analysis of the thermal excitation effects of different groups. This helps to identify differences between groups and thus optimize the grouping method of the thermal heads or the setting of thermal excitation parameters.

[0092] During thermal excitation, the surface temperature of the target film sample is monitored in real time using a temperature feedback unit. A temperature deviation threshold is pre-configured. When the deviation between the surface temperature of the target film sample and the temperature of the corresponding thermistor exceeds the temperature deviation threshold, the power output of the corresponding thermistor is adjusted for dynamic temperature compensation. If temperature compensation fails, a thermal excitation anomaly warning is triggered. For example, if the threshold is still exceeded after three consecutive corrections, it is determined that the temperature compensation has failed, an anomaly warning is triggered, and the thermal excitation is terminated. This allows for the timely detection of potential problems during thermal excitation, such as thermistor malfunction or abnormal heat conduction, thus preventing inaccurate test results or damage to film quality due to abnormal conditions.

[0093] The visual inspection module is configured as follows: during the real-time controllable thermal excitation of the target film sample by each group of thermal heads, the first image is acquired, a multi-level image acquisition scheme is executed, and the thermal image sequence of each group of samples is obtained for thermal response analysis. If the thermal response of the target film sample is qualified, the flipping component is triggered to flip the target film sample, and the second image is acquired and the quality is evaluated on the back of the target film sample to determine whether the quality of the target film sample is qualified.

[0094] In this embodiment, when each set of thermal heads applies thermal excitation to the target film sample, a multi-level image acquisition scheme is used to acquire multiple frames of sample thermal images in real time during the heating and isothermal periods. This breaks through the limitations of traditional single-point-of-time detection and constructs a dynamic and comprehensive thermal response dataset. Thermal response analysis is performed on the sample thermal images, including response time, response grayscale range, response uniformity, and edge sharpness. This enables high-precision evaluation of the thermal performance of the target film sample, accurately identifying the film's response characteristics under thermal excitation, and improving the accuracy of judging whether the thermal response is qualified. This effectively reduces product quality problems caused by poor thermal performance. When the thermal response is determined to be qualified, the flip-up film is flipped, ensuring that the back of the film can be accurately acquired for the second image acquisition, providing a reliable image data foundation for back-side quality assessment. The assessment of heat penetration uniformity and double-sided consistency of the back image delves into the internal heat transfer of the film during thermal excitation and the quality matching degree between the front and back sides. It detects potential quality problems caused by uneven heat penetration or double-sided quality differences, enabling comprehensive and in-depth quality control of the target film sample from the front thermal performance to the back quality. It can timely and accurately determine whether the film quality is qualified, significantly improving the overall quality of dry thermal film and the quality stability in the production process, providing a solid guarantee for product quality, and reducing production costs and market risks caused by quality problems.

[0095] Preferably, the visual inspection module includes: a first inspection head, a second inspection head, and an image processing unit;

[0096] During the process of applying controlled-temperature thermal excitation to the target film sample by each set of thermal heads, a multi-level image acquisition scheme is formulated based on the thermal excitation curve parameters. Since the thermal response of the film changes dynamically during thermal excitation, the multi-level image acquisition scheme ensures targeted image acquisition during key stages such as heating and isothermal control. This allows the acquired image data to comprehensively and accurately reflect the film's thermal response during thermal excitation, providing a rich and high-quality data foundation for subsequent thermal response analysis.

[0097] Based on a multi-level image acquisition scheme, the first image acquisition is performed. A first detection head performs multi-level image acquisition on the target film sample, obtaining a thermal image sequence for each sample, and marking the acquisition stage for each thermal image. The first detection head can capture images in a timely manner at different stages of thermal excitation according to the multi-level image acquisition scheme, ensuring that the acquired images contain information about the dynamic changes of the film throughout the entire thermal excitation process. These sample thermal images can clearly show the changes in the film's color, grayscale, and other characteristics under heat, providing intuitive and effective data support for subsequent accurate analysis of the film's thermal response.

[0098] After acquiring the thermal image of the sample, the image processing unit performs thermal response analysis on the acquired thermal image to determine whether the target film sample has a qualified thermal response. The thermal response analysis of the acquired thermal image by the image processing unit can promptly identify problems with the thermal performance of the film, providing a reliable basis for subsequent quality judgment.

[0099] If the thermal response of the thermal image of a target film sample is unqualified, the target film sample is deemed unqualified; this allows for the rapid screening out of films with substandard thermal performance, saving time and resources.

[0100] If the thermal images of the target film sample all have qualified thermal responses, the clamping component is flipped using the flipping component to flip the target film sample. After the target film sample is flipped, the second detection head is used to acquire an image of the back side of the flipped target film sample, thus obtaining the image of the back side of the sample. This improves the quality inspection dimension, provides the image data foundation of the back side for a comprehensive evaluation of film quality, avoids missing back side quality issues due to only inspecting the front side, and enhances the completeness of quality inspection.

[0101] After acquiring the image of the back of the sample, the image processing unit performs quality defect analysis on the acquired image of the back of the sample in order to further determine whether the quality of the target film sample is up to standard.

[0102] Please see Figure 5 Preferably, the specific steps for developing a multi-level image acquisition scheme include:

[0103] The system acquires the thermal excitation curve parameters for the current dry thermal film production batch, including the heating rate, isothermal temperature, heating period duration, and isothermal period duration. It then performs smooth fitting on the thermal excitation curve data points to generate a continuous temperature-time curve. This provides a data foundation that closely reflects the actual production situation for subsequent operations. Furthermore, the temperature-time curve effectively removes data noise and fluctuations, improving the accuracy of understanding and analyzing the thermal excitation process.

[0104] By calculating the second derivative of the temperature-time curve and detecting abrupt slope changes, characteristic points of the temperature-time curve are identified, including the heating start point, heating end point, isothermal start point, and isothermal end point. Through this method, key characteristic points can be accurately identified. The heating start point marks the beginning of thermal excitation on the film; the heating end point and isothermal start point define the moment when the thermal excitation enters the stable temperature stage; and the isothermal end point indicates the end of a specific stable phase of the thermal excitation. The accurate identification of these characteristic points clearly delineates the different stages of the thermal excitation process.

[0105] The acquisition stages were divided based on the characteristic points of the thermal excitation curve, including a heating stage and an isothermal stage. An image acquisition frequency was set for each stage. During the heating stage, a higher acquisition frequency captured the rapidly changing thermal response of the film; during the isothermal stage, which is relatively stable, a lower acquisition frequency ensured the monitoring of any subtle changes while avoiding resource waste. The number of image frames required for each stage was calculated based on its duration. This ensured that the acquired data volume adequately reflected the thermal response at each stage without over-acquiring and causing data redundancy, providing a strong guarantee for the comprehensive and efficient acquisition of film thermal response information.

[0106] Based on the thermal excitation curve and the division of acquisition stages, the sequence of image acquisition is planned to ensure the continuity and systematic nature of the acquisition process, avoid omissions or disorder in acquisition, and integrate the acquisition stages, acquisition frequency, number of acquisition frames and acquisition sequence to construct a multi-level image acquisition scheme, so as to clarify the specific operation process and requirements of image acquisition throughout the thermal excitation process.

[0107] Please see Figure 6 Preferably, the specific steps for determining whether a target film sample has a qualified thermal response include:

[0108] For each group of thermal heads, the sample thermal image sequence is collected during the controlled temperature thermal excitation process. Each sample thermal image is preprocessed frame by frame, including: using non-local mean filtering for noise reduction to preserve edge details while reducing noise interference, and using an adaptive histogram equalization algorithm to enhance contrast to highlight the grayscale changes in the thermal response area.

[0109] For the sample thermal image sequence after image preprocessing, the sample thermal image subsequence in the heating stage is selected according to the acquisition stage marker. Cross-correlation analysis is performed to calculate the time difference from the application of thermal excitation to the steady state of image grayscale change, which is used as the response time of the target film under the action of the set of thermal heads. This realizes the accurate timing of the target film from the application of thermal excitation to the steady change of grayscale, and provides key time parameters for evaluating thermal performance.

[0110] Preferably, the specific steps of cross-correlation analysis include:

[0111] Based on the acquisition order of the sample thermal image subsequence, the image grayscale change rate of the sample thermal image during the heating stage is calculated sequentially.

[0112] Configure a grayscale change rate threshold. When the grayscale change rate of the sample thermal image during the heating phase is less than the grayscale change rate threshold, obtain the current sample thermal image acquisition timestamp.

[0113] By combining the timestamp of the heating start point, the time difference between the application of controllable temperature thermal excitation and the stabilization of the grayscale change in the sample thermal image is obtained.

[0114] Configure a response time threshold. If the response time of the target film sample is greater than the response time threshold when a set of thermal heads applies controllable temperature thermal excitation, the thermal response of the target film sample is deemed unqualified. Otherwise, no operation is performed. Unqualified target film samples with excessively long response times are screened out to promptly detect possible abnormal thermal excitation response problems in the production process.

[0115] For the preprocessed sample thermal image sequence, the gray value matrix of each sample thermal image is extracted, and the thermal head excitation region of the sample thermal image is segmented according to the region growing algorithm. The gray value variation coefficient of the thermal head excitation region is calculated as the non-uniformity of the sample thermal image.

[0116] Preferably, the specific steps for calculating the grayscale variation coefficient of the thermal head excitation area include:

[0117] Based on the gradient magnitude matrix of the sample thermal image, the coordinates of the pixel with the highest gray level are selected as the seed point for the region growing algorithm.

[0118] Set a standard grayscale difference and set the growth condition to be that the difference between the grayscale value of a pixel and the grayscale value of a seed pixel is less than the standard grayscale difference, and then perform region growth.

[0119] Based on the segmentation results of the region growing algorithm, a binary mask is generated to mark the thermal head excitation region. The mean and standard deviation of gray values ​​within the thermal head excitation region are calculated to obtain the gray value variation coefficient of the thermal head excitation region.

[0120] Configure a uniformity threshold. If the non-uniformity of the thermal image of a sample exceeds the uniformity threshold, the thermal response of the target film sample is deemed unqualified; otherwise, no operation is performed. This prevents imaging quality problems caused by non-uniformity in the thermal head excitation area and improves product quality stability.

[0121] For the preprocessed sample thermal image sequence, the gradient magnitude matrix of each sample thermal image is calculated, and the edge spread function of the gradient magnitude matrix is ​​estimated by Gaussian blur to quantify the edge blur of the sample thermal image, which is used as the edge sharpness of the sample thermal image.

[0122] Preferably, the specific steps for calculating the edge sharpness of the sample thermal image include:

[0123] The gradient operator is used to obtain the directional gradient of the sample thermal image, and the gradient magnitude matrix of the sample thermal image is obtained based on the directional gradient. The gradient magnitude matrix reflects the edge intensity at each pixel in the image. The larger the magnitude, the more likely the point is to be an edge point.

[0124] Gaussian blurring is performed on the gradient magnitude matrix by convolving the Gaussian kernel with the gradient magnitude matrix to obtain a smoothed matrix, namely the edge spread function. The edge spread function describes the degree of blurring of the image edges, smoothing out sharp changes in the edges for easier subsequent analysis.

[0125] Obtain the peak value of the edge spread function and the distance between two points corresponding to half of the peak value, which is used as the edge sharpness of the thermal image of the sample.

[0126] Configure a sharpness threshold. If the edge sharpness of a sample thermal image is greater than the sharpness threshold, the target film sample is deemed to have an unqualified thermal response; otherwise, no operation is performed.

[0127] When a target film sample is determined to have an unacceptable thermal response, the application of controlled temperature thermal excitation by the thermal head is stopped, and the image characteristics and timestamp of the unacceptable target film sample are recorded. This prevents further ineffective operations and saves energy and production time. Simultaneously recording the image characteristics and timestamp of the unacceptable sample facilitates subsequent traceability analysis of the root cause of the problem, contributing to the optimization of production processes and improvement of product quality control.

[0128] If the thermal response of the target film sample is not deemed unqualified at the end of the thermal response analysis of the sample thermal image sequence, then the thermal response of the target film sample is deemed qualified.

[0129] Preferably, the specific steps for quality defect analysis of the acquired sample back images include:

[0130] After image preprocessing of the back image of the sample, a uniformity test is performed. The non-uniformity of the back image is calculated. If the non-uniformity of the back image is greater than the uniformity threshold, the thermal response of the target film sample is deemed unqualified; otherwise, a double-sided consistency verification is performed.

[0131] Obtain the last thermal image of the sample in the sample thermal image sequence and flip it so that it is aligned with the orientation of the sample back image, and use it as a reference image for the sample back image;

[0132] Feature points are extracted and matched on the back-side image of the sample and its control image to align them. The gray-level difference between corresponding pixels in the aligned back-side image and control image is calculated to obtain a gray-level difference image. This ensures the consistency of orientation between the control image and the back-side image, providing an accurate reference standard for subsequent feature point matching and difference analysis. This allows for a more reliable comparison of the differences between the two images during double-sided consistency verification, avoiding misjudgments caused by inconsistencies in orientation and improving the accuracy and reliability of the detection results. By calculating the gray-level difference between corresponding pixels in the aligned back-side image and control image, the gray-level difference image clearly shows the gray-level difference distribution between the two images. This step allows the differences between the images to be presented in an intuitive way, providing strong support for subsequent analysis and judgment of difference points.

[0133] By configuring a deviation anomaly threshold, pixels with grayscale difference images whose pixel values ​​exceed the threshold are selected and marked as difference points. The number of difference points is counted, and the proportion of difference points to the total number of pixels in the grayscale difference image is calculated. By counting the number of difference points and calculating their proportion to the total number of pixels in the grayscale difference image, the degree of difference between images is quantified. This operation makes the evaluation of two-sided consistency more objective and accurate, and can determine whether there are quality problems in the sample based on the proportion of difference points.

[0134] A consistency threshold is configured. If the proportion of differing points to the total number of pixels in the grayscale difference image exceeds the consistency threshold, the double-sided consistency verification is deemed a failure, and the thermal response of the target film sample is deemed unqualified. Otherwise, the double-sided consistency verification is deemed successful, and the thermal response of the target film sample is deemed qualified. This judgment process provides a clear conclusion for the quality inspection of dry thermal film, ensuring that the product quality meets the requirements, effectively controlling product quality risks, and improving the reliability and stability of production.

[0135] Preferably, the control module includes: a sorting execution unit and an exception handling unit;

[0136] The sorting unit determines the quality result based on the quality of the target film sample and uses a pusher to sort and output the target film sample to the qualified area or the unqualified area.

[0137] The anomaly handling unit is used to count the frequency of target film samples extracted from the dry thermal film production line that are deemed to be of substandard quality, in order to provide early warning for the production line.

[0138] In this embodiment, the control module plays a core role in scheduling and decision execution in a thermally stimulated image analysis and recognition system. It includes a sorting execution unit and an anomaly handling unit. The sorting execution unit is responsible for classifying and outputting film samples based on quality inspection results, while the anomaly handling unit provides early warnings to the production line by statistically analyzing the frequency of non-conforming samples.

[0139] The control module establishes a communication connection with the vision inspection module, receiving the quality assessment results of the target film sample from the vision inspection module in real time. The assessment results include two states: qualified and unqualified. Based on the received quality assessment results, the control module determines the area where the target film sample should be transported. If the assessment result is qualified, the target is to transport the sample to the qualified area; if the assessment result is unqualified, the sample is transported to the unqualified area. The control module sends control commands to the pushing component, driving it to move in tandem, transporting the target film sample to the corresponding area.

[0140] When the production line starts running, the anomaly handling unit initializes a non-conforming sample counter and a time window. The time window defines the statistical time range, such as hourly or daily. Whenever the visual inspection module determines a target film sample is non-conforming, the anomaly handling unit increments the non-conforming sample counter by 1. At the end of each time window, the anomaly handling unit calculates the frequency of non-conforming samples within that time window. The anomaly handling unit is configured with a warning threshold; when the calculated non-conforming frequency exceeds the warning threshold, a production line warning is triggered. Warning methods include audible and visual alarms and SMS notifications. This allows for timely detection of potential quality problems on the production line, providing early warnings, reducing the batch production of non-conforming products, and lowering production costs. The control module automates the entire quality inspection process, reducing manual intervention and improving the accuracy and reliability of inspection.

[0141] Working principle and its effects:

[0142] This thermal excitation image analysis and recognition system achieves high-precision quality inspection through the collaborative operation of multiple modules. The sampling module monitors the production line in real time using photoelectric, pressure, and speed sensors. Based on preset rules or real-time quality conditions, it determines the sampling timing using a sampling counter and fluctuation monitoring. Once triggered, it drives a robotic arm to precisely grasp the sample and, together with a pusher, delivers it to the inspection plate.

[0143] The thermal excitation module starts upon receiving the sampling command. The thermal head array first preheats the detection plate in steps, dividing the temperature control zone. Temperature uniformity is ensured by using an infrared thermal imager and temperature deviation threshold adjustment. Then, based on the film batch, the appropriate thermal excitation curve parameters are called, and thermal excitation is initiated when the sample surface temperature change rate is zero. Randomly grouped thermal heads are tested, with the temperature feedback unit monitoring in real time. If the deviation exceeds the limit, the power is adjusted to ensure consistent thermal penetration depth.

[0144] During thermal excitation, the visual inspection module formulates a multi-level image acquisition scheme based on the parameters of the thermal excitation curve. The acquisition stages are divided using the second derivative and slope abrupt change, with different frequencies and frame rates set. The first inspection head acquires a sequence of thermal images of the sample. After preprocessing by the image processing unit, the response time is calculated from the sub-sequence of the heating stage, the grayscale variation coefficient is calculated by segmenting the thermal head excitation area, and edge sharpness is quantified to comprehensively judge the thermal response. If qualified, the flipping of the part is triggered. The second inspection head acquires the back image, first verifying uniformity, and then performing double-sided consistency verification, comprehensively controlling the film quality. In the control module, the sorting execution unit classifies samples based on the visual inspection results; the anomaly handling unit counts the frequency of non-conformities, issuing an early warning if the frequency exceeds the standard, ensuring stable production quality.

[0145] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A thermally excited image analysis and recognition system, wherein the thermally excited image analysis and recognition system is applied to a dry thermal film quality inspection equipment to detect the quality of dry thermal film, characterized in that, include: The monitoring production line determines whether a sampling command is triggered. If triggered, the target film sample is sampled and a temperature control command is generated. The thermal head is controlled to preheat the detection plate in stages, and a controllable temperature thermal excitation is applied to the target film sample group based on the matched thermal excitation curve. By monitoring and identifying feature points of the thermal excitation curve parameters under controllable temperature excitation through slope mutation, a multi-level image acquisition scheme is constructed to acquire the first image, obtain the sample thermal image sequence, and screen the sample thermal image sub-sequence in the heating stage. By configuring the gray change rate threshold, cross-correlation analysis is performed on the image gray change rate of the sample thermal image sub-sequence to obtain the response time of the target film. Extract the grayscale matrix of each group of sample thermal images in the sample thermal image sequence to segment the thermal head excitation area in the sample thermal image and obtain the non-uniformity of the sample thermal image; and obtain the edge sharpness of the sample thermal image by calculating the gradient magnitude matrix of the sample thermal image to determine whether the target film sample has qualified thermal response. If qualified, the target film sample is flipped over, and a second image acquisition is performed to obtain the image of the back of the sample and select its comparison image to generate a grayscale difference image. By verifying the consistency of the grayscale difference image on both sides, it is determined whether the quality of the target film sample is qualified. The quality judgment result of the target film sample is classified and output, and the frequency of the production line quality failure is counted to provide early warning for the production line.

2. The thermally stimulated image analysis and recognition system as described in claim 1, characterized in that, The specific steps for applying controllable temperature thermal excitation to the target film sample grouping based on the matched thermal excitation curve include: The temperature control zone is divided, and a stepped heating mode is used to preheat the test plate in each temperature control zone; the surface temperature distribution of the test plate in each temperature control zone is obtained, and the temperature standard deviation of the global temperature control zone is calculated by constructing a temperature distribution matrix; Configure a temperature deviation threshold, determine whether the temperature standard deviation of the global temperature control area is greater than the temperature deviation threshold. If it is greater than the temperature deviation threshold, locate the deviation from the temperature control area according to the stepped temperature, and adjust the power output of the thermal head that is deviating from the temperature control area to dynamically compensate for the deviation from the temperature control area, so as to control the temperature standard deviation of the global temperature control area to be less than or equal to the temperature deviation threshold. The matching thermal excitation curve parameters are called. When the target film sample arrives at the center of the detection plate, the surface temperature of the target film sample is collected. Based on the rate of change of the surface temperature of the target film sample, it is determined whether to start thermal excitation.

3. The thermally stimulated image analysis and recognition system as described in claim 2, characterized in that, The specific steps for applying controllable temperature thermal excitation to the target film sample grouping based on the matched thermal excitation curve also include: The number of test groups is set to determine the number of groups for the thermal heads in each temperature control area, and the thermal heads in the temperature control area are grouped according to the grouping rules. The grouping rules include: the number of groups in each temperature control area is equal to the number of test groups; each group contains at least one thermal head; the total number of thermal heads in all groups is less than the total number of thermal heads in that temperature control area; and the number of thermal heads in each group is determined randomly. Based on the grouping results of the thermal heads, the temperature of the thermal heads is controlled according to the thermal excitation curve parameters to apply thermal excitation to the target film sample, and the surface temperature of the target film sample is monitored in real time during the thermal excitation process. Configure a temperature deviation threshold to determine whether the deviation between the surface temperature of the target film sample and the temperature of the corresponding thermal head is greater than the temperature deviation threshold. If it is greater than the temperature deviation threshold, adjust the power output of the corresponding thermal head to perform dynamic temperature compensation. If temperature compensation fails, trigger a thermal excitation abnormality warning.

4. The thermally stimulated image analysis and recognition system as described in claim 1, characterized in that, The specific steps for constructing the multi-level image acquisition scheme include: Obtain the parameters of the thermal excitation curve and perform smooth fitting on the data points of the thermal excitation curve to generate a continuous temperature-time curve; The second derivative of the temperature-time curve is calculated, and the characteristic points of the temperature-time curve are identified by slope change detection. The acquisition stages are divided according to the feature points of the thermal excitation curve to set the image acquisition frequency of each acquisition stage, and the number of image frames acquired is calculated according to the duration of each acquisition stage. Plan the sequence of image acquisition and integrate the acquisition stages, acquisition frequency, number of image frames acquired, and acquisition sequence to construct a multi-level image acquisition scheme.

5. The thermally stimulated image analysis and recognition system as described in claim 4, characterized in that, The specific steps for obtaining the sample thermal image sequence include: Based on a multi-level image acquisition scheme, the first image of the target film sample is acquired to obtain the thermal image sequence of each sample. The thermal response analysis of the acquired thermal image sequence is then performed to determine whether the thermal response of the target film sample is qualified. If the thermal response is unqualified, the target film sample is deemed to be of unqualified quality; if the thermal response is qualified, the target film sample is flipped over, a second image is acquired, and the image of the back of the sample is obtained for quality defect analysis to determine whether the target film sample is of qualified quality.

6. The thermally stimulated image analysis and recognition system as described in claim 5, characterized in that, The specific steps for determining whether the target film sample has a qualified thermal response include: After preprocessing the acquired sample thermal image sequence frame by frame, the sample thermal image subsequence in the heating stage is selected according to the acquisition stage marker. Cross-correlation analysis is performed to calculate the time difference from the moment thermal excitation is applied to the moment the image grayscale change reaches a stable state, which is taken as the response time of the target film under the action of the thermal head. Configure a response time threshold. If the response time of the target film sample is greater than the response time threshold when a set of thermal heads applies controllable temperature thermal excitation, the thermal response of the target film sample is deemed unqualified. Extract the gray value matrix of each preprocessed sample thermal image, and segment the thermal head excitation region of the sample thermal image according to the region growing algorithm. Calculate the gray value variation coefficient of the thermal head excitation region as the non-uniformity of the sample thermal image. Configure a uniformity threshold. If the non-uniformity of the thermal image of a sample exceeds the uniformity threshold, the thermal response of the target film sample is deemed unqualified.

7. The thermally stimulated image analysis and recognition system as described in claim 6, characterized in that, The specific steps for determining whether the target film sample has a qualified thermal response also include: The gradient magnitude matrix of the preprocessed thermal image of each sample is calculated, and the edge spread function of the gradient magnitude matrix is ​​estimated by Gaussian blur, which is used to quantify the edge blur of the sample thermal image as the edge sharpness of the sample thermal image. Configure a sharpness threshold. If the edge sharpness of a sample thermal image is greater than the sharpness threshold, the target film sample is deemed to have an unqualified thermal response. If the thermal response of the target film sample is deemed unqualified, the thermal head stops applying controllable temperature thermal excitation and records the image features and timestamp of the unqualified target film sample. If the thermal response of the target film sample is not deemed unqualified at the end of the thermal response analysis of the sample thermal image sequence, then the thermal response of the target film sample is deemed qualified.

8. The thermally stimulated image analysis and recognition system as described in claim 7, characterized in that, The specific steps for obtaining the back image of the sample for quality defect analysis include: The uniformity of the back image of the sample is checked through image preprocessing. The non-uniformity of the back image is calculated. If the non-uniformity of the back image is greater than the uniformity threshold, the thermal response of the target film sample is deemed unqualified. Otherwise, double-sided consistency verification is performed. The last thermal image in the sample thermal image sequence is flipped so that it is oriented in the same direction as the back image of the sample, and used as a reference image for the back image of the sample. Feature points are extracted and matched from the back image of the sample and its control image. The images are then aligned, and the grayscale difference between corresponding pixels in the aligned back image of the sample and the control image is obtained to generate a grayscale difference image. Configure an abnormal deviation threshold, filter pixels whose pixel values ​​in the grayscale difference image are greater than the abnormal deviation threshold, and use them as difference points. Calculate the proportion of difference points to the total number of pixels in the grayscale difference image by counting the number of difference points. Configure a consistency threshold to determine whether the proportion of difference points to the total number of pixels in the grayscale difference image is greater than the consistency threshold. If it is greater, the double-sided consistency verification is deemed to have failed, and the thermal response of the target film sample is deemed to be unqualified; otherwise, the double-sided consistency verification is deemed to have succeeded, and the thermal response of the target film sample is deemed to be qualified.

9. The thermally stimulated image analysis and recognition system as described in claim 1, characterized in that, The specific steps for determining whether a sampling instruction has been triggered include: Set up a sampling counter to record the production output data of the production line in each round of sampling inspection. Initialize the sampling counter and obtain the real-time production output of film based on the collected production data to update the sampling counter. Configure the sampling interval threshold, compare the real-time data of the sampling counter with the sampling interval threshold, determine whether the real-time data of the sampling counter is equal to the sampling interval threshold, if they are equal, determine to trigger a sampling command and clear the sampling counter to zero. When no sampling instruction is triggered, detect fluctuations in production data, including: Real-time analysis of production line pressure data; based on historical production line pressure data, setting a pressure fluctuation range; determining whether the production line pressure data exceeds the pressure fluctuation range; if it does, marking it as an abnormal pressure fluctuation and triggering a sampling command to clear the sampling counter. Simultaneously, the production line speed data is monitored in real time, a change rate threshold is configured, and it is determined whether the real-time change rate of the production line speed is greater than the change rate threshold. If it is greater, it is marked as an abnormal speed change, and a sampling instruction is triggered to clear the sampling counter.

10. The thermally stimulated image analysis and recognition system as described in claim 9, characterized in that, The specific steps for determining whether a sampling instruction has been triggered also include: When a sampling command is triggered, the sampling interval threshold update process is initiated, i.e.: Obtain the quality inspection results of the sampled target film. If the quality of the target film sample is determined to be qualified, maintain the sampling inspection threshold. If the quality of the target film sample is deemed unqualified, the sampling interval threshold will be adjusted to the minimum sampling interval unit. Configure an adjustment threshold to measure the duration of monitoring. Within the adjustment threshold, calculate the pass rate of the sampled target film. Set a pass rate threshold and determine whether the pass rate within the adjustment threshold is greater than the pass rate threshold. If it is greater, restore the sampling detection threshold; otherwise, issue a production quality warning and stop the production line.

Images