Garment quality judgment method and system based on image processing and analysis

Abstract

The invention belongs to the technical field of image data processing, and particularly relates to a garment quality judgment method and system based on image processing analysis, and the method comprises the steps: obtaining a garment image on a conveyor belt through an industrial camera, and executing the standardized spatial transformation through a standard Gaussian filtering smoothing and bilinear interpolation algorithm; extracting a gradient variance and a periodic texture feature mean value of a local region in the transformed image, and obtaining a texture suppression index based on a proportional relation between the gradient variance and the periodic texture feature mean value; obtaining a local gray scale deviation absolute value and gray scale consistency, and obtaining a stress distortion degree through algebraic saturation model mapping; calculating the product of the stress distortion degree and the texture suppression index, and obtaining a quality deviation degree in combination with the deviation degree of the product relative to a quality safety threshold value; and when the quality deviation degree exceeds a threshold value, automatic shutdown maintenance is realized by correcting the input frequency of the conveyor belt driving motor. According to the invention, the sensitivity and the automation degree of garment defect identification in an assembly line scene are improved.

Description

A method and system for judging clothing quality based on image processing analysis Technical Field

[0001] This invention relates to the field of image data processing technology. More specifically, this invention relates to a method and system for judging clothing quality based on image processing analysis. Background Technology

[0002] In the textile and apparel manufacturing industry, surface inspection of garments is a core step in ensuring the quality of finished products. Traditional inspection methods mainly rely on manual visual observation and subjective experience. This approach is not only limited by the physiological fatigue and visual fluctuations of inspectors, but also difficult to achieve standardized real-time monitoring on high-speed industrial conveyor belts. With the evolution of computer vision technology, using imaging equipment in conjunction with algorithms for defect identification has become a common choice in the manufacturing chain.

[0003] However, garment inspection faces severe challenges in real-world industrial environments. Firstly, the complex physical environment of the workshop presents numerous interferences. Industrial camera sensors accumulate heat during long-term operation, easily introducing random thermal noise into the image signal. Simultaneously, continuous vibrations from mechanical devices and uneven lighting distribution cause spatial distortion and contrast fluctuations in the acquired raw images. If these non-target interference signals are not removed during preprocessing, they will severely dilute the characteristic energy of the true defect signals, leading to a significant increase in the false alarm rate during subsequent identification processes.

[0004] Secondly, textile fabrics possess extremely complex periodic textures. Different types of fabrics develop specific spatial frequencies and light-dark patterns during the weaving process. These normal physical textures are often mathematically very similar to warp breaks or yarn pulls. Existing image processing methods struggle to remove background texture interference while preserving details, leading to decision conflicts when processing high-density fabrics.

[0005] Furthermore, existing quality analysis systems have shortcomings in distinguishing between subtle deformations and environmental fluctuations. When there is a slight shift in the stitching on the garment surface, the resulting grayscale change is extremely weak. However, simply increasing the global sensitivity would cause the system to overload with normal color patch fluctuations. As a result, it is difficult for the recognition logic to achieve an ideal balance between detection accuracy and robustness. Summary of the Invention

[0006] To address the technical problem of insufficient sensitivity and robustness in defect identification in clothing inspection due to interference from background textures and environmental noise, the present invention provides solutions in the following aspects.

[0007] In a first aspect, the present invention provides a method for judging clothing quality based on image processing analysis, comprising: acquiring clothing images on a conveyor belt using an industrial camera and performing a standardized spatial transformation on the clothing images to obtain transformed clothing images; extracting the gradient variance and mean value of periodic texture features of local regions from the transformed clothing images, and obtaining a texture suppression index based on the proportional relationship between the gradient variance and the mean value of periodic texture features; acquiring the absolute value of local grayscale deviation and grayscale consistency of the transformed clothing images, and obtaining stress distortion degree based on the mapping result of the absolute value of local grayscale deviation under an algebraic saturation model combined with grayscale consistency; obtaining a quality deviation degree based on the product of stress distortion degree and texture suppression index, and combined with the deviation of stress distortion degree from a quality safety threshold; and correcting the input frequency of the conveyor belt drive motor to achieve automatic shutdown and maintenance when the quality deviation degree exceeds the quality safety threshold.

[0008] This invention acquires images of clothing on a conveyor belt and performs standardized spatial transformations. It then combines the acquired texture suppression index and stress distortion degree to assess the quality deviation of the clothing. When the deviation exceeds the quality safety threshold, it corrects the input frequency of the conveyor belt drive motor to achieve automatic shutdown and maintenance. This invention provides dynamic quality monitoring logic in a production line inspection scenario and reduces the uncertainty caused by manual judgment.

[0009] Preferably, the standardization of the spatial transformation of the clothing image includes: smoothing the clothing image using a standard Gaussian filtering algorithm, and mapping the resolution of the clothing image to a preset pixel size using a bilinear interpolation algorithm.

[0010] This invention uses a standard Gaussian filtering algorithm to smooth clothing images and combines it with a bilinear interpolation algorithm to uniformly map the resolution of clothing images to a preset pixel size. In complex production environments, by standardizing the image acquisition benchmark, it reduces data fluctuation deviations in the subsequent feature acquisition process.

[0011] Preferably, the texture suppression index satisfies the expression: In the formula, express Moment texture suppression index, Indicates time, express The local energy concentration of the clothing image at any given moment. express The gradient variance of the local region at time step. express The mean of periodic texture features extracted at each time step. This represents the zero-structure compensation constant. This represents the texture sensitivity factor.

[0012] This invention obtains the texture suppression index based on the ratio of gradient variance to the mean of periodic texture features, and uses algebraic logic to suppress the energy performance of regular fabric textures, thereby reducing the interference of inherent periodic textures of clothing on defect judgment in different fabric detection scenarios.

[0013] Preferably, the stress distortion degree satisfies the expression: In the formula, express The stress distortion of clothing at all times express Local contrast weight of clothing at any given time. express The consistency of grayscale in clothing at all times. express The absolute value of local grayscale deviation at any given time. Indicates the hysteresis dissipation coefficient. Indicates the sensitivity gain factor. Represents the natural constant.

[0014] This invention obtains stress distortion degree based on the mapping result of the absolute value of local gray-level deviation under the algebraic saturation model and the gray-level consistency. By enhancing the sharp defect signal, it reduces misjudgment caused by background color block fluctuation in clothing quality judgment and obtains the ability to identify minor fraying or damage.

[0015] Preferably, the quality deviation satisfies the expression: In the formula, express The quality deviation of the clothing at all times express The stress distortion of clothing at all times express Moment texture suppression index, Indicates the quality and safety threshold. This represents the sensitivity coefficient for judgment. Represents the natural constant.

[0016] Preferably, the automatic shutdown and maintenance achieved by correcting the input frequency of the conveyor belt drive motor includes: when the quality deviation exceeds the quality safety threshold, sending a frequency correction command to the conveyor belt drive motor to reduce the motor speed and trigger a maintenance alarm.

[0017] This invention sends a frequency correction command to the conveyor belt drive motor to reduce the motor speed and trigger a maintenance alarm when the quality deviation exceeds the quality safety threshold, thereby reducing the risk of defective garments flowing to subsequent processes in the automated control process.

[0018] Preferably, the quality and safety threshold is set to 45.

[0019] Preferably, the zero-prevention structure compensation constant is set to 1.2.

[0020] Preferably, the texture sensitivity factor is obtained by performing linear regression on a standard clothing sample.

[0021] Secondly, the present invention provides a clothing quality judgment system based on image processing analysis, including a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned clothing quality judgment method based on image processing analysis is implemented.

[0022] By adopting the above technical solution, a computer program for judging clothing quality based on image processing analysis is generated and stored in a memory so that it can be loaded and executed by a processor. A terminal device can then be made based on the memory and processor for convenient use.

[0023] The beneficial effects of this invention are as follows: This invention assesses the quality deviation of clothing by obtaining the texture suppression index and stress distortion degree, and combines multi-dimensional image feature extraction and logical mapping to obtain more reliable detection results than single grayscale comparison in industrial conveyor belt operation scenarios.

[0024] This invention utilizes an algebraic saturation model to process local grayscale deviations and suppress the energy representation of periodic fabric textures. Through adaptive attenuation of background interference, it reduces the judgment error caused by ambient lighting or regular patterns of fabrics during the detection of clothing of different materials.

[0025] This invention establishes a linkage system between image recognition results and conveyor belt drive control. When an abnormality in clothing quality is detected, the system automatically stops the machine by correcting the input frequency, thus providing an automated monitoring process with feedback adjustment capabilities in the clothing manufacturing field. Attached Figure Description

[0026] Figure 1 is a flowchart illustrating a method for judging clothing quality based on image processing analysis according to the present invention; Figure 2 is a schematic diagram illustrating the change of texture suppression index based on dynamic feature gradient; Figure 3 is a schematic diagram illustrating the evaluation of surface stress distortion of clothing; Figure 4 is a schematic diagram illustrating the dynamic determination of quality deviation. Detailed Implementation

[0027] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0029] This invention discloses a method for judging the quality of clothing based on image processing analysis. Referring to Figure 1, the method includes steps S1-S4: S1, acquiring clothing images on a conveyor belt using an industrial camera and performing a standardized spatial transformation of the clothing images to obtain the transformed clothing images.

[0030] It should be noted that in order to eliminate the interference caused by the heating of the industrial camera sensor and suppress random interference from the environmental background, this invention must perform a standardized spatial transformation of the clothing image to ensure the consistency of the benchmark for subsequent feature extraction. During the production process, clothing is prone to visual interference caused by environmental factors due to conveyor belt vibration and uneven lighting. Direct analysis will lead to deviations in the judgment results.

[0031] Specifically, this invention uses an industrial camera mounted on a production line support to acquire images of clothing on a conveyor belt at a fixed sampling frequency. The invention employs a standard Gaussian filtering algorithm to smooth the clothing images, thereby performing a standardized spatial transformation of the images, and combines this with a bilinear interpolation algorithm to uniformly map the resolution of the clothing images to a preset pixel size. This invention enhances the contrast of the clothing images through standard histogram equalization technology, improving the recognizability of clothing features in dark areas; and by eliminating interference caused by sensor heating, it provides clean data input for subsequent acquisition of the texture suppression index.

[0032] S2. Extract the gradient variance and mean of periodic texture features of the local region from the transformed clothing image, and obtain the texture suppression index based on the ratio of gradient variance to mean of periodic texture features.

[0033] It should be noted that the inherent periodic textures on the surface of clothing have specific energy characteristics in their regular distribution. In order to prevent regular textures from being misjudged as defects, this invention needs to obtain an algebraic model that can adaptively attenuate background energy. The normal textures on the surface of clothing have regular changes in brightness. If they are not suppressed in a targeted manner, these normal texture features will be mistaken for edges or cracks.

[0034] Specifically, the present invention extracts the gradient variance of local regions in the clothing image after standardized spatial transformation of the clothing image, and obtains the texture suppression index based on nonlinear bias logic.

[0035] The texture suppression index satisfies the expression:

[0036] In the formula, express Moment texture suppression index; Indicates time; express Local energy concentration in the clothing image at any given moment; express Gradient variance of a local region at time step; express The mean of periodic texture features extracted at each time step; This represents the zero-compensation constant for the anti-zero structure; This represents the texture sensitivity factor.

[0037] In the formula, the numerator uses the cross-multiplication of energy and gradient to characterize the complexity of the clothing surface; with the mean of the periodic texture features in the local area... An increase in the value of the denominator leads to an increase in the value of the texture suppression index. The smaller size effectively suppresses the response intensity of regular fabric textures.

[0038] It should be further noted that the texture sensitivity factor was obtained by performing linear regression on standard clothing samples, establishing the slope of the linear regression equation between the texture mean and the background energy residual; the zero-structure compensation constant was also mentioned. The value range is typically between 0.8 and 1.5; in this example of the invention, the zero-compensation constant for the anti-zero structure is used. Setting it to 1.2 ensures robust computation under complex textures. If this constant is set too small, the denominator will approach zero, causing computational oscillations; if the constant is too large, it will weaken the suppression effect on textures.

[0039] For example, Figure 2 is a schematic diagram of the texture suppression index variation based on dynamic feature gradients. The figure shows the fluctuation of the texture suppression index with the sample sequence in a complex industrial production line environment. Due to the introduction of compensation logic for sensor thermal noise and ambient light fluctuations, the index can adaptively attenuate the background energy generated by regular fabric textures. When the fabric texture density changes, the index shows a clear downward trend, demonstrating the effective ability of this invention to remove background interference signals.

[0040] S3. Obtain the absolute value of local grayscale deviation and grayscale consistency of the transformed clothing image. Based on the mapping result of the absolute value of local grayscale deviation under the algebraic saturation model and the grayscale consistency, obtain the stress distortion degree.

[0041] It should be noted that damage or misalignment of clothing usually manifests as abrupt changes in local shape. This distortion signal needs to be enhanced by an algebraic saturation model to simulate the visual system's capture of sharp defects, while preventing signal overload caused by large background fluctuations. Normal clothing surfaces are smooth, while defects can cause drastic changes in light reflection, creating the visual effect of local protrusions or depressions.

[0042] Specifically, this invention introduces an algebraic saturation model to obtain the stress distortion degree, which satisfies the expression:

[0043] In the formula, express The stress distortion of clothing at all times express Local contrast weight of clothing at any given time. express The consistency of grayscale in clothing at all times. express The absolute value of local grayscale deviation at any given time. Indicates the hysteresis dissipation coefficient. Indicates the sensitivity gain factor. Represents the natural constant.

[0044] In the formula, the product term on the left side This constitutes the dynamic gain control logic; the larger the value of the local contrast weight, the more significant the contrast between light and dark areas in the local region, thus providing a basic weight for the distorted signal; the exponential term Nonlinear adjustment of the signal is achieved using grayscale uniformity. Increased grayscale uniformity reflects a smoother local surface and lower noise, causing the exponential term to approach 1, thus preserving the distorted signal completely. Conversely, when grayscale uniformity is low, the exponential term decays rapidly, suppressing pseudo-distortion caused by background clutter. The fractional term on the right... With the absolute value of local grayscale deviation As the value of the fractional term increases, it approaches 1, driving the final stress distortion. Achieving a stable saturation value ensures that the invention can effectively amplify minor imperfections while preventing signal saturation caused by large-area color block fluctuations.

[0045] It should be further noted that the sensitivity gain factor is obtained by calculating the reciprocal of the upper limit of the confidence interval of the gray-scale deviation between the defective area and the normal area through statistical analysis of a historical database of defective products.

[0046] It is necessary to further add the hysteresis dissipation coefficient. The value ranges from 0.01 to 0.05; in this example of the invention, the hysteresis dissipation coefficient is... Set to 0.03 to achieve good interference suppression. If this coefficient is set too small, it will not be able to filter out the jitter interference caused by conveyor belt vibration; if this coefficient is set too large, it will lead to sluggish recognition response.

[0047] For example, Figure 3 is a schematic diagram of stress distortion assessment on the surface of clothing. The figure depicts the real-time monitoring status of stress distortion under the interference of industrial conveyor belt vibration and thermal noise. In the early stage of the sample sequence, the value remains in a low steady-state range, which effectively suppresses false alarm signals caused by mechanical vibration; when the clothing has substantial stitching deviation or damage, the stress distortion responds rapidly and produces pulse-like jumps under the action of the algebraic saturation model, successfully achieving the separation of defect signals from environmental noise.

[0048] S4. Based on the product of stress distortion degree and texture suppression index, and combined with the deviation of stress distortion degree from the quality safety threshold, the quality deviation degree is obtained. When the quality deviation degree exceeds the quality safety threshold, the input frequency of the conveyor belt drive motor is corrected to achieve automatic shutdown and maintenance.

[0049] It should be noted that the final quality judgment requires the comprehensive superposition of the smoothed texture suppression index and the enhanced stress distortion degree to generate a logical judgment index with steep switching characteristics; in this way, the present invention can completely separate normal surface fluctuations of clothing from real physical defects.

[0050] Specifically, the present invention obtains the quality deviation degree through proportional relationship calculation, and the quality deviation degree satisfies the expression:

[0051] In the formula, express The quality deviation of the clothing at all times express The stress distortion of clothing at all times express Moment texture suppression index, Indicates the quality and safety threshold. This represents the sensitivity coefficient for judgment. Represents the natural constant.

[0052] In the formula, the exponential mapping term provides extremely strong numerical discrimination; as the stress distortion degree increases... Exceeding the quality and safety threshold This leads to an explosive increase in the value of the exponential term, driving the final quality deviation. By quickly exceeding the quality and safety threshold, clothing with serious quality defects can be identified.

[0053] It should be further added that, in this example of the present invention, the quality and safety threshold is... Set to 45; Decision sensitivity coefficient The value typically ranges from 5.0 to 15.0; in this example, the sensitivity coefficient will be determined. Set to 10.0 to ensure that the invention can generate sufficient decision gain when detecting minute filamentation.

[0054] Furthermore, when the quality deviation exceeds the quality safety threshold, the controller corrects the input frequency of the conveyor belt drive motor to achieve automatic shutdown and maintenance, thus realizing automated judgment and control of garment quality.

[0055] For example, Figure 4 is a schematic diagram of the dynamic determination of quality deviation. The figure shows the correspondence between the final determination index and the quality safety threshold. In the normal production process, constrained by both texture suppression and stress enhancement logic, the quality deviation always operates below the safety threshold. When a serious physical defect is detected, due to the numerical discrimination provided by the exponential mapping term, the quality deviation exhibits an explosive growth characteristic and quickly breaks through the threshold line. This steep numerical switching characteristic provides deterministic logical support for the system to correct the motor frequency and realize automatic shutdown maintenance.

[0056] This invention also discloses a clothing quality judgment system based on image processing analysis, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement a clothing quality judgment method based on image processing analysis according to this invention.

[0057] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

Claims

1. A method for judging the quality of clothing based on image processing analysis, characterized in that, include: An industrial camera is used to acquire images of clothing on a conveyor belt, and a standardized spatial transformation of the clothing images is performed to obtain transformed clothing images. The gradient variance and mean value of periodic texture features of local regions are extracted from the transformed clothing images. The texture suppression index is obtained based on the ratio of gradient variance to mean value of periodic texture features. The absolute value of local gray-level deviation and gray-level consistency of the transformed clothing images are obtained. The stress distortion degree is obtained based on the mapping result of the absolute value of local gray-level deviation under the algebraic saturation model and the gray-level consistency. The quality deviation is obtained by multiplying the stress distortion degree and the texture suppression index, and by combining the deviation of the stress distortion degree from the quality safety threshold. When the quality deviation degree exceeds the quality safety threshold, the input frequency of the conveyor belt drive motor is corrected to achieve automatic shutdown and maintenance.

2. The method for judging clothing quality based on image processing analysis according to claim 1, characterized in that, The standardization of the spatial transformation of the clothing image includes: smoothing the clothing image using a standard Gaussian filtering algorithm, and mapping the resolution of the clothing image to a preset pixel size using a bilinear interpolation algorithm.

3. The method for judging clothing quality based on image processing analysis according to claim 1, characterized in that, The texture suppression index satisfies the expression: In the formula, express Moment texture suppression index, Indicates time, express The local energy concentration of the clothing image at any given moment. express The gradient variance of the local region at time step. express The mean of periodic texture features extracted at each time step. This represents the zero-structure compensation constant. This represents the texture sensitivity factor.

4. The method for judging clothing quality based on image processing analysis according to claim 1, characterized in that, The stress distortion degree satisfies the expression: In the formula, express The stress distortion of clothing at all times express Local contrast weight of clothing at any given time. express The consistency of grayscale in clothing at all times. express The absolute value of local grayscale deviation at any given time. Indicates the hysteresis dissipation coefficient. Indicates the sensitivity gain factor. Represents the natural constant.

5. The method for judging clothing quality based on image processing analysis according to claim 1, characterized in that, The quality deviation satisfies the expression: In the formula, express The quality deviation of the clothing at all times express The stress distortion of clothing at all times express Moment texture suppression index, Indicates the quality and safety threshold. This represents the sensitivity coefficient for judgment. Represents the natural constant.

6. The method for judging clothing quality based on image processing analysis according to claim 1, characterized in that, The automatic shutdown and maintenance is achieved by correcting the input frequency of the conveyor belt drive motor, including: when the quality deviation exceeds the quality safety threshold, sending a frequency correction command to the conveyor belt drive motor to reduce the motor speed and trigger a maintenance alarm.

7. The method for judging clothing quality based on image processing analysis according to claim 1, characterized in that, The quality and safety threshold is set to 45.

8. The method for judging clothing quality based on image processing analysis according to claim 3, characterized in that, The zero-prevention structure compensation constant is set to 1.

2.

9. The method for judging clothing quality based on image processing analysis according to claim 3, characterized in that, The texture sensitivity factor was obtained by performing linear regression on standard clothing samples.

10. A garment quality judgment system based on image processing analysis, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement a method for judging clothing quality based on image processing analysis according to any one of claims 1-9.

Images