Environment-friendly handbag material surface defect image detection method and system

Abstract

The invention relates to the technical field of handbag material surface defect image detection, in particular to an environment-friendly handbag material surface defect image detection method and system, and the method comprises the following steps: obtaining image data of the surface of an environment-friendly handbag material; dividing the image data into a plurality of local areas; for each local area, extracting image features representing visual features of the local area; establishing a normal feature range of the local area, and continuously adjusting the normal feature range in combination with the environment information; comparing the current image feature with the continuously adjusted normal feature range to identify an abnormal region deviating from the continuously adjusted normal feature range; and according to the image feature mode of the abnormal region, natural textures and defects are distinguished, and defect information is output. According to the method, natural textures and real defects on the surface of the environment-friendly handbag material can be effectively distinguished, potential defects with low contrast and fuzzy edges can be identified, and the accuracy and reliability of detection are improved.

Description

A method and system for detecting surface defects in environmentally friendly handbag materials Technical Field

[0001] This invention relates to the technical field of surface defect image detection for handbag materials, and specifically to a method and system for surface defect image detection of environmentally friendly handbag materials. Background Technology

[0002] In the field of automated quality inspection, especially for surface defect detection of environmentally friendly materials (such as those used in handbag production), ensuring material integrity is crucial. Traditional methods relying on manual visual inspection often fail to detect minute imperfections, which can affect product durability and potentially damage brand image. To overcome these shortcomings, advanced image inspection systems have been developed and deployed. These systems typically utilize high-resolution cameras and sophisticated image processing techniques to identify various surface defects. However, the unique properties of environmentally friendly materials themselves, coupled with subtle variations that may occur during the production process, often pose significant challenges to conventional inspection methods. Summary of the Invention

[0003] The purpose of this invention is to address the aforementioned shortcomings by proposing a method and system for detecting surface defects in environmentally friendly handbag materials.

[0004] The present invention adopts the following technical solution: a method for detecting surface defects in environmentally friendly handbag materials, the method comprising the following steps: acquiring image data of the surface of the environmentally friendly handbag material; dividing the image data into multiple local regions; extracting image features characterizing the visual properties of each local region; establishing a normal feature range for the local region and continuously adjusting the normal feature range in conjunction with environmental information; comparing the current image features with the continuously adjusted normal feature range to identify abnormal regions that deviate from the continuously adjusted normal feature range; distinguishing between natural textures and defects based on the image feature patterns of the abnormal regions, and outputting defect information.

[0005] This technical solution effectively distinguishes between the natural texture and real defects on the surface of environmentally friendly handbag materials, avoiding false alarms caused by natural textures in traditional methods. It can also identify potential defects with low contrast and blurred edges, thereby improving the accuracy and reliability of detection.

[0006] This application also discloses an image detection system for surface defects of environmentally friendly handbag materials, applied to the aforementioned image detection method for surface defects of environmentally friendly handbag materials. The system includes: an acquisition module for acquiring image data of the surface of the environmentally friendly handbag material; a segmentation module for dividing the image data into multiple local regions; an extraction module for extracting image features characterizing the visual properties of each local region; an adjustment module for establishing a normal feature range for the local region and continuously adjusting the normal feature range in conjunction with environmental information; an identification module for comparing the current image features with the continuously adjusted normal feature range to identify abnormal regions that deviate from the continuously adjusted normal feature range; and a processing module for distinguishing between natural textures and defects based on the image feature patterns of the abnormal regions and outputting defect information.

[0007] This application provides a system capable of implementing the aforementioned defect detection method. Through its modular design, the system achieves automated and high-precision detection of surface defects in environmentally friendly handbag materials, effectively overcoming the limitations of manual inspection and improving production efficiency and product quality.

[0008] This application acquires image data of a material surface and divides it into multiple local regions, extracting image features for each region. Based on this, the application establishes a normal feature range for each local region and continuously adjusts this range in conjunction with environmental information, thereby adapting to changes in the production environment. By comparing the current image features with the continuously adjusted normal feature range, the application can identify abnormal regions that deviate from the normal range. More importantly, the application can effectively distinguish between natural textures and real defects based on the image feature patterns of abnormal regions and output defect information. This method overcomes the false alarm problem caused by the similarity between natural textures and defect features in existing technologies, avoiding the misjudgment of harmless natural textures as defects. Simultaneously, by continuously adjusting the normal feature range and combining it with environmental information, the application can effectively identify low-contrast, blurred-edge "water stain" defects, solving the problem that existing systems cannot identify such critical defects. Therefore, this application significantly improves the accuracy and reliability of surface defect detection for environmentally friendly tote bag materials, reduces the false alarm rate, and effectively prevents materials with potential quality hazards from entering the market, thereby protecting product quality and brand reputation.

[0009] To further understand the features and technical content of the present invention, please refer to the following detailed description and drawings of the present invention. However, the drawings provided are for reference and illustration only and are not intended to limit the present invention. Attached Figure Description

[0010] Figure 1 is a flowchart of a method for detecting surface defects in environmentally friendly handbag materials according to the present invention; Figure 2 is a schematic diagram of a system for detecting surface defects in environmentally friendly handbag materials according to the present invention. Detailed Implementation

[0011] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can understand the advantages and effects of the present invention from the content disclosed in this specification. The present invention can be implemented or applied through other different specific embodiments, and various details in this specification can also be modified and changed based on different viewpoints and applications without departing from the spirit of the present invention. Furthermore, the accompanying drawings of the present invention are for simple illustrative purposes only and are not depictions of actual dimensions; this is stated in advance. The following embodiments will further describe the relevant technical content of the present invention in detail, but the disclosed content is not intended to limit the scope of protection of the present invention.

[0012] This embodiment provides a method and system for detecting surface defects in environmentally friendly handbag materials, as shown in Figures 1 and 2.

[0013] Referring to Figure 1, a method for detecting surface defects in environmentally friendly handbag materials includes the following steps: acquiring image data of the surface of the environmentally friendly handbag material; dividing the image data into multiple local regions; extracting image features characterizing the visual properties of each local region; establishing a normal feature range for the local region and continuously adjusting the normal feature range in conjunction with environmental information; comparing the current image features with the continuously adjusted normal feature range to identify abnormal regions that deviate from the continuously adjusted normal feature range; distinguishing between natural textures and defects based on the image feature patterns of the abnormal regions, and outputting defect information.

[0014] Here, "image data" refers to the visual information of the surface of environmentally friendly handbag materials acquired through optical equipment (such as high-resolution cameras), typically stored as digital images containing information such as pixel brightness and color. "Local region" refers to a smaller, independent image patch into which the entire image data is divided according to certain rules; each local region can be analyzed independently. "Image features" refer to numerical values ​​or vectors extracted from local regions that quantify their visual characteristics, such as texture features, color features, and shape features. "Normal feature range" refers to the statistical range within which the image features of local regions on the surface of environmentally friendly handbag materials should fall under normal production conditions; this range is dynamically adjusted based on environmental information. "Environmental information" can include external environmental parameters such as temperature, humidity, and light intensity in the production workshop, as well as internal production parameters such as material batch and production speed. "Abnormal region" refers to a local region whose image features deviate from the continuously adjusted normal feature range, potentially indicating defects or unusual textures. "Natural texture" refers to the inherent, harmless surface structure of the environmentally friendly handbag material itself, such as the natural texture formed by interwoven fibers. "Defects" refer to abnormalities on the surface of a material that affect its performance or appearance, such as scratches, stains, uneven coatings, etc.

[0015] The core of the environmentally friendly tote bag material surface defect image detection method proposed in this application lies in achieving intelligent analysis and defect identification of material surface image data through a series of refined steps.

[0016] First, image data of the surface of the eco-friendly tote bag material needs to be acquired. This can be achieved in several ways. For example, a high-resolution industrial camera can be used to continuously photograph the material on the production line, converting the acquired analog signals into digital image data using an analog-to-digital converter. Alternatively, a line-scan camera can be used for scanning imaging, acquiring image information of the material surface line by line and stitching it together to form a complete image. Furthermore, multispectral imaging equipment can be used to acquire image data of the material surface at different wavelengths to capture richer information about the material's properties.

[0017] Secondly, the acquired image data is divided into multiple local regions. Various strategies can be employed for image data division. For example, a fixed-size grid method can be used to uniformly divide the entire image into several rectangular or square local regions. Alternatively, content-adaptive division methods can be used, such as edge detection or region growing algorithms, to segment the image into irregular local regions with similar visual characteristics. Furthermore, overlapping division can be employed, ensuring some overlap between adjacent local regions and guaranteeing the continuity of feature extraction and analysis at region boundaries.

[0018] Next, for each local region, image features characterizing the visual properties of that region are extracted. Image feature extraction is a crucial step in defect identification. For example, gray-level co-occurrence matrix features, such as energy, contrast, correlation, and entropy, can be extracted to quantify the texture information of the local region. Color features such as color histograms and color moments can also be extracted to describe the color distribution of the local region. Furthermore, more complex texture features can be extracted using methods such as local binary patterns and Gabor filters, or the frequency characteristics of the local region can be analyzed through Fourier transform.

[0019] Then, a normal feature range for a local area is established, and this range is continuously adjusted based on environmental information. The establishment of the normal feature range can be achieved through the analysis of a large number of images of qualified materials. For example, statistical analysis can be performed on the image features of local areas of qualified materials to calculate their mean and standard deviation, and a confidence interval can be set as the normal feature range. The integration and continuous adjustment of environmental information is one of the key innovations of this application. For example, environmental parameters such as temperature and humidity in the production workshop can be monitored in real time. When these parameters change, the normal feature range can be fine-tuned according to a pre-defined mapping relationship. Furthermore, the normal feature range for the corresponding batch can be retrieved from a pre-established database based on the production batch information of the materials, to accommodate the differences in characteristics between different batches of materials.

[0020] Subsequently, the current image features are compared with the continuously adjusted normal feature range to identify anomalous regions that deviate from the adjusted normal feature range. This comparison process can be implemented using various statistical methods. For example, the image features of the current local region can be compared with the boundaries of the normal feature range; if the feature value exceeds preset upper and lower limits, it is judged as an anomaly. Alternatively, the distance between the current feature and the center value of the normal feature range can be calculated; if the distance exceeds a certain threshold, it is identified as an anomaly. Furthermore, machine learning-based anomaly detection algorithms, such as support vector machines or isolated forests, can be used to classify the current features and determine whether they belong to anomalies.

[0021] Finally, based on the image feature patterns of the abnormal regions, natural textures and defects are distinguished, and defect information is output. This step is crucial for addressing false positives and false negatives. For example, a classification model can be built that learns from a large number of labeled natural texture and defect samples to classify abnormal regions based on their image feature patterns (such as shape, size, contrast, and texture complexity). When a defect is identified, the system outputs corresponding defect information, such as the defect type, location, size, and severity, for subsequent processing and quality control.

[0022] The aforementioned method for detecting surface defects in environmentally friendly handbag materials further includes the following steps: establishing a first environmental state evaluator and outputting a first environmental state judgment based on external sensor data; establishing a second environmental state evaluator and inferring and outputting a second environmental state judgment by analyzing macroscopic changes in the material image; comparing the first environmental state judgment and the second environmental state judgment to identify whether there is a continuous difference between them; when there is a continuous difference and the continuous difference exceeds a preset difference range, making an adjustment plan: taking the second environmental state judgment output by the second environmental state evaluator as the dominant factor, or correcting the external sensor data; and updating the normal feature range of the local area based on the adjustment plan.

[0023] Specifically, the first environmental state assessor can be understood as a module used to monitor key parameters in the production environment, such as temperature, humidity, light intensity, and airflow speed. These parameters are typically collected in real time by external sensors (such as temperature and humidity sensors, light sensors, and anemometers) and converted into a first environmental state judgment, which reflects the current physical environmental conditions. Its purpose is to provide a direct and quantitative reference for the environmental state.

[0024] The second environmental state evaluator can be understood as a module that indirectly infers the environmental state by analyzing the macroscopic visual characteristics of images of eco-friendly handbag materials. For example, macroscopic features such as the overall brightness, tone uniformity, background noise level, or large-area texture changes in the material image are often related to environmental factors such as ambient lighting, dust accumulation, or production line vibration. By analyzing these macroscopic changes, a second environmental state judgment can be inferred and output. Its purpose is to provide an environmental state reference based on the material's own visual performance, as a supplement or verification to external sensor data.

[0025] In practical applications, the first and second environmental state assessments are continuously compared to identify any persistent discrepancies. This comparison aims to uncover potential inconsistencies between external sensor data and the actual visual representation of the material. For example, a discrepancy may exist if an external illumination sensor shows stable light intensity, but the material image consistently exhibits brightness fluctuations.

[0026] Furthermore, when a persistent difference is identified that exceeds a preset difference range, the system will implement corresponding adjustment measures. The preset difference range can be set based on historical data or expert experience to define acceptable environmental state judgment bias. Adjustment measures can include two strategies: one is to prioritize the second environmental state judgment output by the second environmental state evaluator, assuming that judgments based on macroscopic changes in material images are more reliable, in which case external sensor data may be problematic; the other is to correct the external sensor data, for example, through algorithmic compensation or prompting manual sensor inspection. The aim is to ensure the accuracy of the environmental information sources used to adjust the normal feature range.

[0027] Therefore, based on the adjusted plan, the normal characteristic range of the local area will be updated. This means that if the source of the environmental condition assessment is adjusted or corrected, subsequent continuous adjustments to the normal characteristic range will be based on more reliable environmental information, thereby improving the accuracy of the adjustments.

[0028] This application's solution effectively addresses the reliability issues that may arise from a single source of environmental information by introducing a dual environmental state assessment mechanism. Specifically, the first environmental state evaluator provides a direct environmental judgment based on physical sensor data, while the second environmental state evaluator infers the environmental state from macroscopic visual changes in material images, providing an indirect and independent judgment. By continuously comparing these two independent judgments, the system can promptly detect inconsistencies between them. When such inconsistencies persist and exceed acceptable limits, it indicates that one of the environmental information sources may be biased. At this point, the system can intelligently select the more reliable judgment as the dominant one, or correct the questionable sensor data, thereby ensuring that the environmental information used to adjust the normal characteristic range of a local area is verified and optimized. It is precisely because of this cross-validation and adaptive adjustment mechanism that the continuous adjustment of the normal characteristic range becomes more accurate and robust, avoiding misjudgments or omissions caused by inaccurate environmental information.

[0029] In some preferred embodiments, it is assumed that defect detection of the material surface is required on the production line of environmentally friendly handbag materials.

[0030] First, an environmental condition evaluator is established. This evaluator acquires real-time data on the ambient temperature, humidity, and light intensity of the workshop by connecting to temperature, humidity, and light sensors on the production line, and integrates this data into a first environmental condition judgment. For example, the current temperature is 25°C, the humidity is 60%, and the light intensity is 500 lux.

[0031] Simultaneously, a second environmental state evaluator is established, which continuously analyzes the macroscopic characteristics of the environmentally friendly tote bag material images acquired by the image acquisition device. For example, it monitors the overall average brightness of the image, the background noise level, and the texture consistency of large areas of the material. If the overall brightness of the image is found to be consistently low, or the background noise suddenly increases significantly, the second environmental state evaluator will infer that the current ambient light may be insufficient or there may be interference, and output a corresponding second environmental state judgment.

[0032] The system continuously compares the first environmental state assessment with the second environmental state assessment. For example, if the first environmental state evaluator reports that the light intensity is stable at 500 lux, but the second environmental state evaluator continuously infers insufficient light (e.g., the average image brightness is consistently below the normal threshold), and this difference exceeds a preset range (e.g., the average brightness deviation exceeds 10% and lasts for 5 minutes), then the system will recognize a persistent difference between the two.

[0033] Based on the preset adjustment scheme, the system may determine that, since macroscopic changes in the image directly reflect the actual impact of the environment on the material, the "insufficient illumination" judgment output by the second environmental state evaluator will be the primary indicator. Simultaneously, the system may trigger a calibration command, prompting the system to check or calibrate the external illumination sensor.

[0034] Ultimately, based on this adjusted and more reliable environmental condition assessment (i.e., "insufficient illumination"), the normal feature range of the local area will be updated accordingly. For example, the system may appropriately lower the upper limit of brightness in the normal feature range to adapt to the current actual low-light environment, thereby avoiding misjudging the behavior of normal materials under low light as defects. In this way, even if external sensors deviate, the system can self-correct through internal cross-validation mechanisms to ensure the accuracy of defect detection.

[0035] To address this, this application further proposes steps for distinguishing defects, including: maintaining a feature history record for each local region, which stores the image feature sequence of the local region within a continuous time period; when the current image feature is detected to deviate from the continuously adjusted normal feature range, retrieving the feature history record of the corresponding abnormal region; performing stability analysis on the image feature sequence in the feature history record of the abnormal region to obtain the feature fluctuation degree and feature change rate that deviate from the continuously adjusted normal feature range; and distinguishing between transient optical artifacts and defects based on the feature fluctuation degree and feature change rate. Specifically, when the feature fluctuation degree is higher than a preset fluctuation threshold and the feature change rate is high, and both changes appear and recover to the continuously adjusted normal feature range within a short period of time, it is judged as a transient optical artifact; when the feature fluctuation degree is low and the feature change rate is low, it is judged as a defect.

[0036] Specifically, maintaining a feature history record for each local region refers to continuously collecting and storing the extracted image features of each predefined local region in the image of the surface of the eco-friendly tote bag material over a period of time. This feature history record can be a time-series database or a sliding window buffer, containing the feature vectors of that local region across multiple consecutive image frames. For example, it can store the sequence of changes in features such as brightness, contrast, texture complexity, and color distribution of the local region over time. When the current image features are detected to deviate from the continuously adjusted normal feature range, it means that the real-time feature value of a certain local region exceeds its pre-established and dynamically updated normal behavior pattern. At this time, the system will retrieve the feature history record corresponding to the abnormal region for deeper analysis.

[0037] Stability analysis of image feature sequences in the historical feature records of anomaly regions aims to assess the persistence and dynamic characteristics of the anomaly. Stability analysis can employ various statistical or signal processing methods, such as calculating the variance, standard deviation, autocorrelation function, and Fourier transform of the feature sequence, to quantify its fluctuation and rate of change. This allows us to obtain the degree of feature fluctuation and the rate of change of features deviating from the continuously adjusted normal feature range. The degree of feature fluctuation can be understood as the dispersion or amplitude of the feature value on the time axis, while the rate of change reflects how quickly the feature value changes over time.

[0038] Based on the acquired characteristic fluctuation level and characteristic change rate, anomalies can be finely classified. Specifically, when the characteristic fluctuation level exceeds a preset fluctuation threshold and the characteristic change rate is high, and both changes occur within a short period and quickly recover to the normal characteristic range after continuous adjustment, the anomaly is judged as a transient optical artifact. Transient optical artifacts are usually brief visual disturbances caused by changes in ambient light, temporary adhesion of dust particles, or camera shake, which are not inherent defects in the material. Conversely, when the characteristic fluctuation level and characteristic change rate are low, it indicates that the anomaly has strong persistence and stability, and is thus judged as a defect. Defects are usually structural or surface integrity problems inherent in the material itself, and their characteristics are relatively stable and do not easily recover on their own within a short period of time.

[0039] This application's solution effectively addresses the limitations of the aforementioned basic scheme in distinguishing between transient optical artifacts and real defects by introducing the maintenance and stability analysis of local region feature history. The basic scheme primarily relies on comparing the current image features with the normal feature range, making it difficult to capture the temporal dynamics of anomalies. Because transient optical artifacts and real defects exhibit significant differences in their temporal behavior—transient optical artifacts typically manifest as brief, dramatic feature fluctuations and changes that recover quickly, while real defects exhibit continuous, relatively stable feature deviations—this application, by performing stability analysis on feature history to obtain the degree of feature fluctuation and the rate of feature change, can quantitatively describe anomalies in the temporal dimension. Therefore, the system can more accurately classify anomaly regions based on these dynamic indicators, rather than just static feature deviations, thereby avoiding misjudging transient, non-defective disturbances as defects and significantly improving detection accuracy and robustness.

[0040] In some preferred embodiments, it is assumed that a localized area is monitored in consecutive image frames on a production line for environmentally friendly handbag materials. At a certain moment, due to a brief airflow disturbance in the production workshop, a small patch of dust briefly adheres to the surface of this localized area and quickly detaches in the next frame. At this time, the image features of this localized area (e.g., brightness, texture complexity) suddenly exhibit a significant peak, i.e., the feature value rises sharply in a short period of time, and then quickly falls back to the normal feature range after continuous adjustment. By performing stability analysis on the feature history of this localized area, it can be calculated that its feature fluctuation is high (e.g., variance far exceeds a preset threshold) and its feature change rate is high (e.g., large feature differences between adjacent frames). Because this phenomenon of high fluctuation and high change rate appears and recovers in a very short time, the system will judge it as a transient optical artifact rather than a real material defect. Conversely, if a minor scratch appears in the local area, its image features may continuously deviate from the adjusted normal feature range, but its feature fluctuation and rate of change are relatively low. That is, although the feature value deviates from the normal range, its changes over time are relatively stable, without sharp instantaneous peaks. In this case, the system identifies it as a genuine defect based on the criteria of low fluctuation and low rate of change. In this way, the proposed solution can effectively distinguish between these two different types of anomalies, thereby improving the accuracy of defect detection.

[0041] This application further proposes a step of maintaining a feature history record for each local area. This step includes: real-time monitoring of the conveying speed and image acquisition parameters of the environmentally friendly handbag material; adjusting the number of image frames stored in the feature history record according to the conveying speed to ensure that the time span covered by the feature history record remains constant; adjusting the number of image frames stored in the feature history record according to the image acquisition parameters to ensure that the material movement distance covered by the feature history record remains constant; and ensuring that the feature history record can reflect the image feature sequence of the local area within a fixed time span or a fixed material movement distance by adjusting the number of image frames stored in the feature history record.

[0042] Specifically, real-time monitoring of the conveying speed and image acquisition parameters of environmentally friendly handbag materials refers to continuously acquiring the real-time conveying speed of the material and various parameters of the image acquisition equipment, such as frame rate, exposure time, and resolution, through sensors or control systems integrated on the production line. The purpose is to provide accurate input data for subsequent dynamic adjustments. Adjusting the number of image frames stored in the feature history record based on the conveying speed to ensure a constant time span covered by the feature history record can be understood as the system dynamically calculating and adjusting the number of image frames to be stored according to preset time span requirements when the material conveying speed changes. For example, if the conveying speed increases, more image frames need to be stored to cover the same time span; conversely, if the conveying speed decreases, fewer image frames need to be stored. In practical applications, adjusting the number of image frames stored in the feature history record based on image acquisition parameters to ensure a constant material movement distance covered by the feature history record means that when the parameters of the image acquisition equipment (such as the frame rate) change, the system adjusts the number of stored image frames accordingly to ensure that the physical length of the material represented by the feature history record remains consistent. For example, if the frame rate decreases, fewer image frames need to be stored to cover the same material movement distance. Therefore, by adjusting the number of image frames stored in the feature history record as described above, it can be ensured that the feature history record always reflects the image feature sequence of the local area within a fixed time span or a fixed material movement distance, thereby providing a standardized and comparable data basis for subsequent defect analysis.

[0043] This application's solution monitors the conveying speed and image acquisition parameters of the environmentally friendly tote bag material in real time, and intelligently adjusts the number of image frames stored in the feature history record based on this dynamic information. Specifically, when the conveying speed changes, adjusting the number of image frames ensures that the feature history record always covers a preset, constant time span, making the feature history records of different times or different batches of material comparable in the time dimension. Simultaneously, when the image acquisition parameters change, adjusting the number of image frames ensures that the feature history record always covers a preset, constant material movement distance, making the feature history records of different acquisition conditions consistent in the spatial dimension. It is precisely because of this dynamic adjustment mechanism that the maintained feature history record can accurately and stably reflect the image feature sequence of a local area within a fixed time span or a fixed material movement distance, providing a reliable and standardized data foundation for subsequent stability analysis and defect differentiation.

[0044] In some preferred embodiments, it is assumed that the production line for environmentally friendly tote bag materials requires the feature history of each local area to cover a 2-second material movement process. When the system monitors the material conveying speed in real time as 1 m / s and the image acquisition frame rate is 10 frames / s, the feature history will be configured to store 20 images (10 frames / s * 2 seconds) to cover the 2-second time span. If the conveying speed suddenly increases to 2 m / s, in order to maintain the 2-second time span, the system will calculate and adjust the number of image frames stored in the feature history based on the new speed and frame rate (assuming the frame rate remains constant). For example, it may be necessary to store 40 images (20 frames / s * 2 seconds, if the frame rate is also adjusted accordingly). Similarly, if the image acquisition frame rate is adjusted due to changes in lighting conditions, for example, from 10 frames / s to 5 frames / s, in order to ensure that the material movement distance covered by the feature history remains constant (e.g., corresponding to 1 meter of material), the system will recalculate and adjust the number of stored image frames based on the new frame rate and conveying speed (assuming the conveying speed is 1 m / s). Through this dynamic adjustment, regardless of changes in production conditions, the feature history record can always provide standardized and comparable image feature sequences, thus providing a solid data foundation for subsequent defect judgment.

[0045] This application further proposes steps for distinguishing between transient optical artifacts and defects, including: segmenting the image feature sequence in the feature history of the abnormal region into multiple consecutive time periods; calculating the local fluctuation degree and local change rate of the image feature sequence for each time period; comparing the local fluctuation degree and local change rate of adjacent time periods to identify whether there are drastic and discontinuous change points in the image feature sequence; when there are drastic and discontinuous change points, distinguishing between transient optical artifacts and defects based on the location and amplitude of the change point, as well as the time span required for the change point to appear and recover to the continuously adjusted normal feature range.

[0046] Specifically, segmenting the image feature sequences in the feature history of anomaly regions involves dividing the image feature sequences stored in the feature history into multiple consecutive, shorter time periods according to a preset time interval or number of frames. For example, if the feature history contains a feature sequence of 100 frames, it can be divided into 10 consecutive time periods, each containing features from 10 frames. The purpose of this is to capture the dynamic changes of image features more precisely within different time periods.

[0047] Specifically, for each time period, the local fluctuation level and local rate of change of the image feature sequence are calculated. This can be understood as independently calculating the statistical characteristics of the image features within each defined time period, such as mean, variance, and standard deviation, to characterize the local fluctuation level, and the average or maximum value of the feature differences between adjacent frames to characterize the local rate of change. The aim is to obtain more granular information on feature changes, rather than relying solely on the macroscopic statistics of the entire sequence.

[0048] In practical applications, comparing the degree of local fluctuation and the rate of local change between adjacent time periods aims to identify whether there are drastic or discontinuous points of change in the image feature sequence. For example, the difference between the degree of local fluctuation or the rate of local change between adjacent time periods can be calculated and compared with a preset threshold. If the difference exceeds the threshold, it indicates that there may be a drastic or discontinuous point of change at the boundary of that time period. The purpose is to accurately locate the start and end points of abnormal events.

[0049] When abrupt and discontinuous changes occur, transient optical artifacts and defects can be distinguished based on the location and amplitude of the change, as well as the time span required for the change to appear and recover to the continuously adjusted normal characteristic range. Specifically, the location of the change indicates the point in time when the anomalous event occurred, the amplitude reflects the severity of the anomalous event, and the recovery time span measures the persistence of the anomalous event. For example, abrupt changes that appear quickly and recover rapidly are more likely to be transient optical artifacts, while changes that last longer and have relatively stable or slowly changing amplitudes are more likely to indicate defects.

[0050] This application's solution segments the image feature sequences in the feature history of anomaly regions and calculates the local fluctuation degree and local change rate for each time period, decomposing complex feature change patterns into more easily analyzable local segments. It is precisely this fine-grained analysis that enables the system to accurately identify drastic and discontinuous change points in the image feature sequence by comparing the local characteristics of adjacent time periods. These change points are often key indicators of transient optical artifacts or transient perturbations. By further combining the location and amplitude of the change points with the time span required for them to recover to the continuously adjusted normal feature range, this solution can more accurately capture the "transient" and "recoverable" characteristics of transient optical artifacts, thereby effectively distinguishing them from genuine defects. Compared to relying solely on overall stability analysis, this segmented and local comparison method significantly improves the accuracy and robustness of identifying dynamic, transient anomalies.

[0051] In some preferred embodiments, it is assumed that the feature history of a local region contains a 10-second sequence of image features. For more refined analysis, this sequence is divided into 100 consecutive 0.1-second time intervals. For each 0.1-second time interval, the local fluctuation and local rate of change of image features such as average brightness and texture complexity are calculated. Subsequently, the system compares these local metrics between adjacent 0.1-second time intervals. For example, if the local brightness fluctuation suddenly rises from 0.05 to 0.8 in a certain time interval (e.g., the 50th 0.1-second interval) and then rapidly recovers to around 0.05 in the following few time intervals, while the local rate of change also exhibits a similar sharp rise and fall trend, this will be identified as a sharp, discontinuous change point. Further, if this change point occurs within 0.5 seconds and completely recovers to the continuously adjusted normal feature range, and its amplitude is within the normal range, the system will judge it as a transient optical artifact, such as caused by a brief flash of light or dust particles rapidly passing through the detection area. Conversely, if the degree of local fluctuation and the rate of change remain at a low level over a certain period of time, but deviate from the normal characteristic range, and there are no obvious points of drastic change or signs of recovery, it will be judged as a defect.

[0052] The steps described above for comparing the degree of local fluctuation and the rate of local change between adjacent time periods to identify whether there are drastic and discontinuous change points in the image feature sequence include: obtaining the production batch information of the material and the current production environment parameters; extracting the expected fluctuation range and rate of change characteristics of the periodic texture corresponding to the current material batch and environmental conditions from a pre-stored material property database based on the production batch information and the current production environment parameters; comparing the degree of local fluctuation and the rate of local change between adjacent time periods, and combining the expected fluctuation range and rate of change characteristics to identify whether there are drastic and discontinuous change points in the image feature sequence, wherein when the difference between the degree of local fluctuation and the rate of local change between adjacent time periods exceeds the range allowed by the expected fluctuation range and rate of change characteristics, it is determined that there are drastic and discontinuous change points.

[0053] Specifically, acquiring material production batch information and current production environment parameters means that when performing surface defect image inspection on environmentally friendly tote bag materials, the system will acquire, in real-time or in advance, the production batch information of the material being inspected, such as batch number, material type, and supplier, as well as the parameters of the current inspection environment, such as ambient temperature, humidity, light intensity, and airflow speed. This information can be obtained through manual input, automatic acquisition through integration with the Production Management System (MES), or real-time monitoring by external sensors. The purpose is to provide crucial contextual information for subsequent feature analysis, enabling the system to make personalized judgments based on specific circumstances.

[0054] The process of extracting the expected fluctuation range and rate of change of periodic textures corresponding to the current material batch and environmental conditions from a pre-stored material property database, based on production batch information and current production environment parameters, can be understood as the system maintaining a material property database containing a large amount of historical data and predefined rules. This database stores the normal fluctuation range and rate of change of surface image features (such as texture, color, brightness, etc.) of different production batches and different material types under various environmental conditions. For example, some natural fiber materials may exhibit greater texture fluctuations under specific humidity levels, and these fluctuations are normal and non-defective. Upon obtaining the current production batch information and environmental parameters, the system queries this database to extract the range of local fluctuations and local rates of change that best match the current conditions and are expected to be present in the material. These expected characteristics can be expressed as statistical means, standard deviations, confidence intervals, or more complex predicted values ​​from machine learning models. The purpose is to provide a dynamic, context-based benchmark for subsequent anomaly detection.

[0055] In practical applications, comparing the degree of local fluctuation and the rate of change between adjacent time periods, and combining this with the expected fluctuation range and rate of change characteristics, identifies whether there are drastic and discontinuous change points in the image feature sequence. Specifically, this involves comparing the difference in the degree of local fluctuation and the rate of change between adjacent time periods, calculated at the moment, with the expected fluctuation range and rate of change characteristics extracted from the material property database. For example, the deviation between the current difference and the expected range can be calculated, or the current difference can be input into a classifier trained based on the expected characteristics. When the difference in the degree of local fluctuation and the rate of change between adjacent time periods exceeds the range allowed by the expected fluctuation range and rate of change characteristics, a drastic and discontinuous change point is identified. This means that only when the observed change exceeds the normal fluctuation range acceptable to the material under the current environmental conditions will it be considered a true abnormal change point, thus avoiding misjudging normal fluctuations as defects.

[0056] This application's solution, by incorporating material production batch information and current production environment parameters, combined with a pre-stored material property database, provides a more refined and accurate basis for identifying whether there are drastic and discontinuous change points in image feature sequences. Specifically, traditional comparison methods may rely solely on fixed thresholds to judge differences in local fluctuation levels and rates of change, which can easily lead to misjudgments when faced with material diversity and environmental variability. In contrast, this application's solution first acquires production batch information and environmental parameters, enabling the system to understand the specific "background" of the object being detected. Subsequently, based on this background information, the system can dynamically extract the "expected" fluctuation range and rate of change characteristics that match the current material and environmental conditions from a pre-established material property database containing a large amount of historical experience data. These expected characteristics reflect the fluctuation pattern that the specific material should exhibit under normal conditions in the current environment. Finally, when comparing the local fluctuation level and local rate of change between adjacent time periods, the comparison is no longer made with a static, universal threshold, but with a dynamic, personalized "expected" range. Only when the observed fluctuations or changes exceed this "expected" range are they identified as drastic and discontinuous points of change. This mechanism effectively distinguishes fluctuations in the inherent properties of the material and normal changes caused by the environment from genuine defect signals, thereby improving the accuracy of identification.

[0057] In some preferred embodiments, suppose a batch of material made from recycled polyester fiber is being tested on an environmentally friendly tote bag material production line. The batch information shows "RPF-202305-A," and the workshop environmental sensors report a current humidity of 75% and a temperature of 28°C. Without the solution described in this application, if the recycled polyester fiber material naturally exhibits higher local texture fluctuations at 75% humidity than in a dry environment, traditional detection methods might misinterpret it as "drastic and discontinuous change points."

[0058] However, when using the scheme of this application, the system first obtains the production batch information of "RPF-202305-A" and the environmental parameters of "75% humidity and 28℃". Subsequently, the system queries a pre-stored material property database. This database may store historical data indicating that the expected range of local fluctuation degree for the recycled polyester fiber material of batch "RPF-202305-A" under humidity conditions of 70%-80% is X to Y, and the expected range of local change rate is A to B. When the system calculates the difference in local fluctuation degree and local change rate between adjacent time periods, it compares these differences with the expected ranges of X to Y and A to B extracted from the database. Only when the actual observed difference exceeds these expected ranges, for example, the difference in local fluctuation degree reaches Z (Z>Y), or the difference in local change rate reaches C (C>B), will the system determine that there is a drastic and discontinuous change point. In this way, the system can accurately distinguish between normal texture fluctuations and true surface defects of the recycled polyester fiber material under specific humidity conditions, thereby avoiding false alarms and improving the accuracy of detection.

[0059] This application further proposes a method to distinguish between transient optical artifacts, transient optical disturbances, and complex defects based on the location and amplitude of the change point, and the time span required for the change point to appear and recover to the continuously adjusted normal characteristic range. This includes: acquiring the spatial distribution characteristics of the abnormal region, including the morphology, size, and edge characteristics of the abnormal region; acquiring the evolution trajectory of the spatial distribution characteristics of the abnormal region over different time periods; and distinguishing between transient optical artifacts, transient optical disturbances, and complex defects based on the location and amplitude of the change point, the spatial distribution characteristics, the evolution trajectory of the spatial distribution characteristics, and the time span required for the change point to appear and recover to the continuously adjusted normal characteristic range. Specifically, when the characteristic fluctuation level is higher than a preset fluctuation threshold, and the... When the characteristic change rate is high, and both changes appear and recover to the normal characteristic range after continuous adjustment within a short period of time, and the spatial distribution characteristics show no fixed shape and random distribution, it is judged as a transient optical artifact; when the characteristic fluctuation degree is higher than the preset fluctuation threshold, and the characteristic change rate is high, and both changes appear and recover to the normal characteristic range after continuous adjustment within a short period of time, and the spatial distribution characteristics show the characteristics of moving with local airflow or electrostatic field changes, it is judged as a transient optical disturbance; when the characteristic fluctuation degree is low, the characteristic change rate is low, and the time span required for the change point to appear and recover to the normal characteristic range after continuous adjustment is long, and the spatial distribution characteristics show multiple, superimposed or specific geometric shapes, it is judged as a composite defect.

[0060] Specifically, acquiring the spatial distribution characteristics of anomaly regions refers to using image processing techniques, such as image segmentation, edge detection, and morphological analysis, to extract information such as the geometric shape, size, and boundary sharpness of the anomaly regions. Here, morphology can be understood as the overall outline and structure of the anomaly region, such as circular, linear, or irregular shapes; size refers to quantitative indicators such as the area, perimeter, and aspect ratio of the anomaly region; and edge characteristics describe the smoothness, sharpness, or blurriness of the anomaly region's boundaries. These features provide static spatial information about the anomaly region at a given moment.

[0061] Furthermore, acquiring the evolution trajectory of the spatial distribution characteristics of anomaly regions over different time periods refers to continuously monitoring the spatial characteristics of anomaly regions and recording their changes over time. For example, a dynamic evolution path can be constructed by comparing the changes in the shape, size, and location of anomaly regions in consecutive image frames. This evolution trajectory can reveal whether the anomaly is fixed, randomly moving, periodically changing, or gradually evolving, thus providing crucial dynamic spatial information for distinguishing different types of anomalies.

[0062] Therefore, the solution in this application constructs a multi-dimensional and more refined anomaly discrimination model by combining the instantaneous optical characteristics of the anomalous region (such as the degree of characteristic fluctuation and the rate of characteristic change) with more comprehensive spatial distribution characteristics (such as morphology, size, and edge characteristics) and their evolution trajectory in the time dimension. This combination enables the system not only to identify the occurrence and duration of anomalies, but also to deeply understand the spatial manifestations of anomalies and their dynamic change patterns. For example, instantaneous optical artifacts usually exhibit random, non-fixed-shape, and short-lived spatial characteristics; transient optical disturbances may exhibit specific spatial morphologies and evolution trajectories that change with the external environment (such as airflow and electrostatic fields); while composite defects often have stable, specific, or superimposed geometric shapes and last for a relatively long time. By comprehensively analyzing this information, misjudgments that may occur when judging solely based on time characteristics can be effectively avoided.

[0063] Through the above technical solution, this application can significantly improve the accuracy and robustness of surface defect detection for environmentally friendly tote bag materials. By introducing spatial distribution characteristics and their evolution trajectories, the system can more precisely distinguish between transient optical artifacts, transient optical disturbances, and composite defects, thereby reducing the misjudgment of harmless transient phenomena as real defects and avoiding the omission of complex or dynamically changing defects. This not only optimizes the accuracy of defect classification but also provides a more reliable basis for subsequent quality control and production process adjustments, helping to improve product quality and reduce production costs.

[0064] In some preferred embodiments, it is assumed that on the production line of environmentally friendly handbag materials, the image detection system identifies an image feature in a local area that deviates from the normal feature range.

[0065] First, the system obtains the historical feature records of the abnormal area and performs stability analysis. It finds that the feature fluctuation is high and the feature change rate is also high, and it appears and recovers to the normal range in a short period of time.

[0066] Next, the system further acquires the spatial distribution characteristics of the abnormal region, including its shape, size and edge characteristics, and tracks its evolution trajectory in consecutive image frames.

[0067] If the spatial distribution characteristics of the abnormal area show that it has no fixed shape and is randomly distributed, and its evolution trajectory shows that it appears briefly and then disappears quickly, the system judges it as a transient optical artifact, such as tiny dust particles that are briefly attached to the surface of a material.

[0068] If the spatial distribution characteristics of the abnormal area show that it moves with changes in local airflow or electrostatic field, such as its shape or position swinging or drifting slightly in a regular manner in a short period of time, the system judges it as a transient optical disturbance, such as changes in the reflectivity of the material surface caused by local airflow in the production environment.

[0069] If the abnormal region has low characteristic fluctuation and low characteristic change rate, but its spatial distribution features exhibit multiple, superimposed or specific geometric characteristics, such as a long and thin scratch and a dot-shaped stain coexisting, and its evolution trajectory shows that it lasts for a long time, then the system judges it as a composite defect, such as structural damage that exists in the material itself.

[0070] Through this multi-dimensional analysis, the system can accurately classify different types of anomalies, thereby achieving more intelligent and reliable defect detection.

[0071] This application further proposes a step for obtaining the evolution trajectory of the spatial distribution characteristics of the aforementioned abnormal area in different time periods, including: real-time acquisition of the conveying speed and image acquisition frequency of the environmentally friendly handbag material; adjusting the image acquisition frame rate according to the conveying speed to ensure that the number of image frames per unit material movement distance remains constant, thus obtaining the adjusted image acquisition frame rate; adjusting the length of the image frame sequence used to construct the evolution trajectory according to the image acquisition frequency to ensure that the time span covered by the evolution trajectory remains constant, thus obtaining the adjusted image frame sequence length; and obtaining the evolution trajectory of the spatial distribution characteristics of the abnormal area in different time periods based on the adjusted image acquisition frame rate and the adjusted image frame sequence length.

[0072] Specifically, real-time acquisition of the conveying speed of eco-friendly tote bag materials refers to continuously monitoring the material's movement rate on the production line using speed sensors or encoders installed on the line. Image acquisition frequency can be understood as the number of images captured per second by the image acquisition device, which is typically determined by the settings of the camera or image processing system.

[0073] The purpose of adjusting the image acquisition frame rate according to the conveying speed is to ensure that the number of image frames acquired over the same physical distance the material travels remains consistent, regardless of changes in the material's movement speed. For example, when the conveying speed increases, the image acquisition frame rate is increased accordingly to avoid acquiring too few image frames over the same distance, which could lead to the loss of detail in the evolution trajectory. Conversely, when the conveying speed decreases, the image acquisition frame rate is decreased to avoid acquiring redundant image data.

[0074] In practical applications, the length of the image frame sequence used to construct the evolution trajectory is adjusted according to the image acquisition frequency. The purpose is to ensure that the analyzed evolution trajectory always covers a fixed time span. For example, if the image acquisition frequency changes, the system will adjust the length of the image frame sequence used for analysis accordingly to ensure that, regardless of the acquisition frequency, the analyzed image sequence represents the evolution of the anomalous region within the same time period.

[0075] Therefore, by adjusting the image acquisition frame rate and the image frame sequence length as described above, the evolution trajectory of the spatial distribution characteristics of standardized and consistent abnormal regions in different time periods can be obtained.

[0076] This application's solution addresses the inconsistency in evolution trajectory data caused by fluctuations in production line speed or changes in acquisition parameters by real-time monitoring of the conveying speed and image acquisition frequency of the environmentally friendly tote bag material, and dynamically adjusting the image acquisition frame rate and image frame sequence length accordingly. Specifically, real-time acquisition of the conveying speed allows the system to adjust the image acquisition frame rate based on the actual movement of the material, ensuring a constant number of image frames acquired per unit material movement distance. This is crucial for analyzing the spatial evolution of abnormal areas. Simultaneously, real-time acquisition of the image acquisition frequency and adjustment of the image frame sequence length ensure that the evolution trajectory always covers a fixed time span, which is significant for analyzing the dynamic changes of abnormal areas over time. It is precisely this dual adjustment mechanism that ensures the high consistency and accuracy of the spatial distribution characteristics of the acquired abnormal areas, providing a reliable data foundation for subsequent defect classification.

[0077] Through the above technical solution, this application effectively overcomes the limitations of traditional methods in acquiring inaccurate evolution trajectory data in dynamic production environments. By intelligently adjusting the image acquisition frame rate and image frame sequence length, it ensures that the evolution trajectory of the acquired spatial distribution characteristics of abnormal areas has a high degree of standardization and consistency in both time and space dimensions. This significantly improves the accuracy and robustness in distinguishing between transient optical artifacts, transient optical disturbances, and composite defects, reduces false alarm and false negative rates, and thus enhances the overall performance and reliability of the environmentally friendly tote bag material surface defect detection system.

[0078] In some preferred embodiments, it is assumed that the eco-friendly tote bag material is conveyed on the production line at a speed of 10 centimeters per second, and the image acquisition system is initially set to acquire 10 frames per second. In this case, to ensure that one frame is acquired for every 1 centimeter of material movement, the image acquisition frame rate is set to 10 frames per second. Simultaneously, to analyze the evolution trajectory of abnormal areas within 0.5 seconds, the image frame sequence length is set to 5 frames (10 frames per second * 0.5 seconds).

[0079] Specifically, when the conveying speed of the eco-friendly tote bag material suddenly increases to 20 centimeters per second, the system will detect this change in real time. To maintain a constant number of image frames per unit material movement distance (i.e., 1 frame per 1 centimeter), the system will automatically adjust the image acquisition frame rate to 20 frames per second. Simultaneously, to ensure the evolution trajectory still covers a 0.5-second time span, the system will adjust the image frame sequence length according to the new image acquisition frequency (20 frames per second), setting it to 10 frames (20 frames per second * 0.5 seconds).

[0080] Conversely, if the conveying speed is slowed to 5 centimeters per second, the image acquisition frame rate will be adjusted to 5 frames per second, and the image frame sequence length will be adjusted to 2.5 frames (in practical applications, this will be rounded down, for example, to 3 frames, or processed according to a more refined strategy). Through this dynamic adjustment, regardless of changes in the material conveying speed, the evolution trajectory of the spatial distribution characteristics of the acquired abnormal areas over different time periods can maintain its uniformity in the physical space or time dimension, thus providing stable and accurate data input for subsequent defect classification.

[0081] This application further proposes a step for dividing image data into multiple local regions, including: denoising the image data; and dividing the denoised image data into multiple local regions based on a grid partitioning method.

[0082] Specifically, image denoising refers to preprocessing the raw image data of the surface of the eco-friendly handbag material to eliminate or reduce random noise, salt-and-pepper noise, Gaussian noise, and other noise present in the image. The aim is to improve image quality and reduce the interference of noise on subsequent image analysis and processing. In practical applications, denoising can be implemented using various algorithms, such as median filtering, Gaussian filtering, bilateral filtering, wavelet denoising, or non-local mean denoising. These methods smooth the image and remove outlier pixel values, thereby highlighting the effective information in the image and providing a clearer data foundation for subsequent local region segmentation.

[0083] The grid-based partitioning method, which divides denoised image data into multiple local regions, can be understood as follows: after denoising, the entire image is structurally segmented into several rectangular or square regions of equal or predetermined size. The aim is to decompose the complex overall image into smaller, more manageable and analyzable blocks, allowing each local region to be independently processed for feature extraction and analysis, thereby improving detection efficiency and accuracy. For example, an image can be divided into an M-row, N-column grid, with each grid representing a local region. This partitioning method is simple, efficient, and ensures that each local region has similar size and structure, facilitating subsequent parallel processing and feature comparison.

[0084] The proposed solution first denoises the image data, effectively removing potential interference information from the original image and ensuring data quality for subsequent processing. Based on this, a grid partitioning method is used to structurally segment the denoised image data, ensuring that each local region is formed based on clear and accurate image information. The introduction of denoising eliminates the random influence of noise on local region segmentation, guaranteeing accuracy and stability. Simultaneously, the standardized nature of grid partitioning ensures clear boundaries for each local region, facilitating independent image feature extraction and analysis for each region. This avoids misclassification or feature extraction bias caused by noise, thus laying a solid foundation for subsequent anomaly identification and defect differentiation.

[0085] Through the above technical solutions, this application can significantly improve the accuracy and robustness of surface defect image detection for environmentally friendly handbag materials. The introduction of denoising effectively avoids the interference of noise on local region segmentation and subsequent feature extraction, enabling the system to more accurately identify the true texture and potential defects of the material surface. Based on a mesh partitioning method, the uniformity and controllability of local region segmentation are ensured, simplifying subsequent parallel processing and improving detection efficiency. Compared to directly segmenting the original image, this scheme, through preprocessing and structured segmentation, effectively reduces the false alarm rate and false negative rate, improving the reliability and practicality of the entire detection method.

[0086] In some preferred embodiments, this application is implemented as follows: First, raw image data of the surface of the environmentally friendly handbag material is acquired using an industrial camera. It is assumed that some randomly distributed salt-and-pepper noise exists in the raw image data. To eliminate this noise, the image data is input to a median filter for denoising. This filter effectively smooths the image and removes the salt-and-pepper noise by replacing the gray value of each pixel with the median gray value of its neighboring pixels. After denoising, the resulting clear image data is fed into a grid partitioning module. This module divides the image data according to a preset 10x10 grid, thereby dividing the entire image into 100 equally sized local regions. The size of each local region is set to 50x50 pixels, ensuring that each region is small enough for detailed analysis while being large enough to contain meaningful visual information. In this way, each local region is formed based on the denoised clear image data, providing high-quality input for subsequent image feature extraction and defect identification.

[0087] Referring to Figure 2, a specific embodiment of this application also discloses an image detection system for surface defects of environmentally friendly handbag materials, applied to the aforementioned image detection method for surface defects of environmentally friendly handbag materials. The system includes: an acquisition module for acquiring image data of the surface of the environmentally friendly handbag material; a segmentation module for dividing the image data into multiple local regions; an extraction module for extracting image features characterizing the visual properties of each local region; an adjustment module for establishing a normal feature range for the local region and continuously adjusting the normal feature range in conjunction with environmental information; an identification module for comparing the current image features with the continuously adjusted normal feature range to identify abnormal regions that deviate from the continuously adjusted normal feature range; and a processing module for distinguishing between natural textures and defects based on the image feature patterns of the abnormal regions and outputting defect information.

[0088] The environmentally friendly tote bag material surface defect image detection system proposed in this application aims to effectively address the challenges faced by traditional detection systems in detecting surface defects in environmentally friendly tote bag materials through modular design and intelligent processing. These challenges include misidentifying natural textures as defects and failing to identify potential defects that highly overlap with natural texture features. Through the close collaboration of its various functional modules, the system achieves accurate acquisition, detailed analysis, dynamic adaptation, and intelligent judgment of material surface image data, thereby significantly improving the accuracy and reliability of defect detection.

[0089] To better understand the technical solution proposed in this application, the various modules in the system are described in detail below. The functions (i.e., the corresponding method steps) performed by each module have already been described in the above embodiments, and will not be repeated here. It should be emphasized that these modules, as system components, can be implemented in various ways.

[0090] Specifically, the acquisition module can be configured to utilize one or more high-resolution industrial cameras capable of capturing image data of the surface of the eco-friendly tote bag material in real time. For example, the acquisition module may include a line scan camera that acquires image information of the material through continuous scanning and transmits it to a subsequent processing unit. Alternatively, the acquisition module can also consist of an array of area scan cameras to acquire large-area image data in parallel.

[0091] The segmentation module can be configured as a software program running on the processor, responsible for dividing the image data output by the acquisition module into multiple local regions according to a preset grid or adaptive algorithm. For example, the segmentation module can use a fixed-size rectangular grid to uniformly segment the image. As another implementation, the segmentation module can use image processing algorithms (such as edge detection or region growing) to dynamically identify salient features in the image and thereby segment semantically meaningful local regions.

[0092] The extraction module can be configured to utilize image processing libraries or dedicated hardware accelerators to extract image features characterizing the visual properties of each local region. For example, the extraction module can calculate gray-level co-occurrence matrix features (such as energy, contrast, entropy, etc.) of the local region to quantify texture information. As an alternative implementation, the extraction module can apply Gabor filters or local binary mode algorithms to extract more complex texture features, or calculate color histograms and color moments to describe color distribution.

[0093] The adjustment module can be configured to include a data storage unit and a processing unit. The data storage unit stores the initial model of the normal feature range of the local area and an environmental information database. The processing unit is responsible for receiving environmental information in real time (e.g., from temperature sensors, humidity sensors, material batch information input interfaces, etc.) and continuously adjusting the normal feature range based on this environmental information. For example, the adjustment module can dynamically update the upper and lower limits of the normal feature range based on a pre-trained machine learning model and current environmental parameters.

[0094] The recognition module can be configured to use a comparator and a decision logic unit. The comparator compares the current image features with the normal feature range after continuous adjustment by the adjustment module. The decision logic unit then identifies abnormal regions that deviate from the normal feature range based on the comparison results. For example, the recognition module can calculate the statistical distance between the current feature and the center value of the normal feature range; when this distance exceeds a preset threshold, it is determined to be an abnormal region.

[0095] The processing module can be configured to include a classifier and an information output interface. The classifier is responsible for distinguishing between natural textures and defects based on the image feature patterns of the abnormal region. The information output interface is used to output the identified defect information. For example, the processing module can use a support vector machine as a classifier to classify the features of the abnormal region. When a defect is identified, the processing module sends information such as the type, location, size, and severity of the defect to the operator or a subsequent automated processing system through the information output interface.

[0096] The surface defect image detection system for environmentally friendly handbag materials proposed in this application represents a significant improvement over traditional automated quality inspection systems. Traditional detection systems often employ fixed detection thresholds or simple image processing algorithms, making it difficult to effectively handle the complexity of the natural textures on the surface of environmentally friendly materials and the dynamic changes in the production environment, resulting in high false alarm and false negative rates. For example, traditional systems often fail to accurately identify minute defects such as "water stain-like" patches that highly overlap with natural texture features.

[0097] This application's system overcomes the aforementioned limitations by introducing a dynamically adjusted normal feature range and an intelligent differentiation mechanism based on image feature patterns. The collaborative work of the acquisition, segmentation, and extraction modules ensures high-quality image data acquisition and refined feature extraction. The adjustment module continuously optimizes the normal feature range based on real-time environmental information and material batch differences, enabling the system to adapt to minor fluctuations in the production process and avoid misjudging normal natural textures as defects. The recognition and processing modules further achieve accurate differentiation between natural textures and real defects through in-depth analysis of image feature patterns in abnormal areas, particularly effectively identifying defects that are macroscopically similar to natural textures but have potential quality risks. Therefore, this application's system significantly improves the accuracy and reliability of defect detection, reduces the risk of false alarms and false negatives, thereby ensuring product quality and improving production efficiency and brand image.

[0098] The content disclosed above is only a preferred and feasible embodiment of the present invention, and is not intended to limit the scope of protection of the present invention. Therefore, all equivalent technical changes made based on the content of the present invention specification and drawings are included within the scope of protection of the present invention. Furthermore, the elements therein can be updated as technology develops.

Claims

1. A method for detecting surface defects in environmentally friendly handbag materials using images, characterized in that, The method includes the following steps: acquiring image data of the surface of the environmentally friendly handbag material; dividing the image data into multiple local regions; extracting image features that characterize the visual properties of each local region; establishing the normal feature range of the local region and continuously adjusting the normal feature range in combination with environmental information; comparing the current image features with the continuously adjusted normal feature range to identify abnormal regions that deviate from the continuously adjusted normal feature range; distinguishing between natural textures and defects based on the image feature patterns of the abnormal regions, and outputting defect information.

2. The method for detecting surface defects in environmentally friendly handbag materials as described in claim 1, characterized in that, The method further includes the following steps: establishing a first environmental state evaluator and outputting a first environmental state judgment based on external sensor data; establishing a second environmental state evaluator and inferring and outputting a second environmental state judgment by analyzing macroscopic changes in material images; comparing the first environmental state judgment and the second environmental state judgment to identify whether there is a continuous difference between them; when there is a continuous difference and the continuous difference exceeds a preset difference range, making an adjustment plan: taking the second environmental state judgment output by the second environmental state evaluator as the dominant factor, or correcting the external sensor data; and updating the normal feature range of the local area based on the adjustment plan.

3. The method for detecting surface defects in environmentally friendly handbag materials as described in claim 1, characterized in that, The steps for distinguishing defects include: maintaining a feature history record for each local region, which stores the image feature sequence of the local region over a continuous time period; when the current image feature is detected to deviate from the continuously adjusted normal feature range, retrieving the feature history record of the corresponding abnormal region; performing stability analysis on the image feature sequence in the feature history record of the abnormal region to obtain the feature fluctuation degree and feature change rate that deviate from the continuously adjusted normal feature range; and distinguishing between transient optical artifacts and defects based on the feature fluctuation degree and feature change rate. Specifically, when the feature fluctuation degree is higher than a preset fluctuation threshold and the feature change rate is high, and both changes appear and recover to the continuously adjusted normal feature range within a short period of time, it is judged as a transient optical artifact; when the feature fluctuation degree is low and the feature change rate is low, it is judged as a defect.

4. The method for detecting surface defects in environmentally friendly handbag materials as described in claim 3, characterized in that, The steps for maintaining a feature history record for each local area include: real-time monitoring of the conveying speed and image acquisition parameters of the eco-friendly handbag material; adjusting the number of image frames stored in the feature history record according to the conveying speed to ensure that the time span covered by the feature history record remains constant; adjusting the number of image frames stored in the feature history record according to the image acquisition parameters to ensure that the material movement distance covered by the feature history record remains constant; and ensuring that the feature history record can reflect the image feature sequence of the local area within a fixed time span or a fixed material movement distance by adjusting the number of image frames stored in the feature history record.

5. The method for detecting surface defects in environmentally friendly handbag materials as described in claim 3, characterized in that, Distinguishing between transient optical artifacts and defects also includes the following steps: segmenting the image feature sequence in the feature history of the abnormal region into multiple consecutive time periods; calculating the local fluctuation degree and local rate of change of the image feature sequence for each time period; comparing the local fluctuation degree and local rate of change of adjacent time periods to identify whether there are drastic and discontinuous change points in the image feature sequence; when there are drastic and discontinuous change points, distinguishing between transient optical artifacts and defects based on the location and amplitude of the change point, as well as the time span required for the change point to appear and recover to the normal feature range after continuous adjustment.

6. The method for detecting surface defects in environmentally friendly handbag materials as described in claim 5, characterized in that, The steps for identifying whether there are drastic and discontinuous change points in the image feature sequence by comparing the local fluctuation degree and local change rate of adjacent time periods include: obtaining the production batch information of the material and the current production environment parameters; extracting the expected fluctuation range and change rate characteristics of the periodic texture corresponding to the current material batch and environmental conditions from a pre-stored material property database based on the production batch information and the current production environment parameters; comparing the local fluctuation degree and local change rate of adjacent time periods, and combining the expected fluctuation range and change rate characteristics to identify whether there are drastic and discontinuous change points in the image feature sequence, wherein when the difference between the local fluctuation degree and local change rate of adjacent time periods exceeds the range allowed by the expected fluctuation range and change rate characteristics, it is determined that there are drastic and discontinuous change points.

7. The method for detecting surface defects in environmentally friendly handbag materials as described in claim 5, characterized in that, The steps for distinguishing between transient optical artifacts and defects based on the location and amplitude of the change point, and the time span required for the change point to appear and recover to the continuously adjusted normal feature range, include: acquiring the spatial distribution characteristics of the abnormal region, including the shape, size, and edge characteristics of the abnormal region; acquiring the evolution trajectory of the spatial distribution characteristics of the abnormal region in different time periods; and distinguishing between transient optical artifacts, transient optical disturbances, and composite defects based on the location and amplitude of the change point, the spatial distribution characteristics and their evolution trajectory, and the time span required for the change point to appear and recover to the continuously adjusted normal feature range. Specifically, when the feature fluctuation level is higher than a preset fluctuation threshold and the feature change rate is high, and both... When changes occur and recover to the normal characteristic range after continuous adjustment within a short period of time, and the spatial distribution features exhibit characteristics of no fixed shape and random distribution, it is judged as a transient optical artifact; when the feature fluctuation level is higher than the preset fluctuation threshold, and the feature change rate is high, and both changes occur and recover to the normal characteristic range after continuous adjustment within a short period of time, and the spatial distribution features exhibit characteristics of moving with local airflow or electrostatic field changes, it is judged as a transient optical disturbance; when the feature fluctuation level is low, the feature change rate is low, and the time span required for the change point to appear and recover to the normal characteristic range after continuous adjustment is long, and the spatial distribution features exhibit characteristics of multiple, superimposed, or specific geometric shapes, it is judged as a composite defect.

8. The method for detecting surface defects in environmentally friendly handbag materials as described in claim 7, characterized in that, The steps for obtaining the evolution trajectory of the spatial distribution characteristics of an abnormal area over different time periods include: real-time acquisition of the conveying speed and image acquisition frequency of the environmentally friendly handbag material; adjusting the image acquisition frame rate according to the conveying speed to ensure that the number of image frames per unit material movement distance remains constant, resulting in an adjusted image acquisition frame rate; adjusting the length of the image frame sequence used to construct the evolution trajectory according to the image acquisition frequency to ensure that the time span covered by the evolution trajectory remains constant, resulting in an adjusted image frame sequence length; and obtaining the evolution trajectory of the spatial distribution characteristics of the abnormal area over different time periods based on the adjusted image acquisition frame rate and the adjusted image frame sequence length.

9. The method for detecting surface defects in environmentally friendly handbag materials as described in claim 1, characterized in that, The steps of dividing image data into multiple local regions include: denoising the image data; and dividing the denoised image data into multiple local regions based on a grid partitioning method.

10. A surface defect image detection system for environmentally friendly handbag materials, applied to the surface defect image detection method for environmentally friendly handbag materials as described in claim 1, characterized in that, The system includes: an acquisition module for acquiring image data of the surface of the environmentally friendly handbag material; a segmentation module for dividing the image data into multiple local regions; an extraction module for extracting image features that characterize the visual properties of each local region; an adjustment module for establishing the normal feature range of the local region and continuously adjusting the normal feature range in conjunction with environmental information; an identification module for comparing the current image features with the continuously adjusted normal feature range to identify abnormal regions that deviate from the continuously adjusted normal feature range; and a processing module for distinguishing between natural textures and defects based on the image feature patterns of abnormal regions and outputting defect information.

Images