Double-threshold adaptive CFAR detection method for worsted fabric defects

Abstract

The invention discloses a double-threshold self-adaptive CFAR detection method for worsted fabric defects, and relates to the technical field of worsted fabric defect detection. The invention discloses a double-threshold self-adaptive CFAR detection method for defects of worsted fabric. The method comprises the following steps: shooting a color image of a worsted fabric fabric by adopting a CCD (Charge Coupled Device) camera; converting the acquired color image into a grayscale image to obtain a grayscale value of each pixel; and traversing the target grayscale image pixel by pixel by using a sliding window type CFAR detector. According to the method, the radar target CFAR detection technology is introduced into the defect detection of the wool fabric, the detection of the wool fabric with different textures, hues and flatness can be realized, an extremely low false detection rate can be adaptively provided, and the maximum probability detection of defects with different shapes and sizes is realized; according to the method, the reliability and the effectiveness of fabric defect detection are ensured, the self-adaptive CFAR detection of the fabric defects is realized, and the method has the advantages of low false detection rate, high defect detection rate, simple algorithm, high speed, self-adaption and the like.

Description

Technical Field

[0001] The invention relates to the technical field of worsted wool fabric defect detection, and in particular to a double-threshold self-adaptive CFAR detection method for worsted wool fabric defects. Background Art

[0002] Worsted wool fabrics generally require fabric defect inspection for product quality control during the production process, including blank inspection, intermediate inspection, and final inspection. Fabric defect detection is typically performed manually through visual inspection, followed by raw and finished fabric repairs before the finished product is stored. Manual visual inspection is associated with high costs, low efficiency, and a high rate of missed inspections.

[0003] In order to solve the various problems faced by manual visual inspection of fabric defects, people have proposed methods such as histogram, edge detection, threshold segmentation, Gabor transform, and wavelet transform to detect textile defects. However, these methods have problems such as low detection accuracy and high false detection rate.

[0004] In recent years, people have used convolutional neural networks or deep learning methods to detect textile defects. However, the performance of these methods is heavily dependent on the training sample data set, with a high rate of missed detection of small defects and the need for massive labeled data sets. In addition, the algorithms are complex, the number of parameters is large, and the hardware deployment requirements are high.

[0005] Radar target detection is currently the most mature technology in signal detection theory. Currently, radar target constant false alarm rate (CFAR) detection technology has been widely used to detect targets in various complex time-varying, non-stationary, and non-Gaussian clutter environments.

[0006] Radar target CFAR detection technology has the advantages of simple algorithm, self-adaptation, high detection accuracy, and good real-time performance. Radar target detection is similar to worsted fabric defect detection in that they both detect abnormal points from the background, but there are differences.

[0007] In this regard, the inventor believes that how to introduce radar target CFAR detection technology into the defect detection of woolen fabrics, so as to realize the automatic detection of woolen fabric defects and overcome the limitations of existing detection schemes such as histograms, edge detection, and deep learning, is a technical problem that technical personnel in this field urgently need to solve.

[0008] The information disclosed in this background technology section is only intended to enhance understanding of the overall background of the invention and should not be regarded as an admission or any form of suggestion that the information constitutes the prior art already known to a person skilled in the art. Summary of the Invention

[0009] In response to the above technical problems, an embodiment of the present invention provides a dual-threshold adaptive CFAR detection method for worsted fabric defects to solve the problems raised in the above background technology.

[0010] The present invention provides the following technical solution: a double-threshold adaptive CFAR detection method for worsted fabric defects, comprising the following steps: A CCD camera is used to capture color images of worsted fabrics; Convert the collected color image into a grayscale image to obtain the grayscale value of each pixel; A sliding window CFAR detector is used to traverse the target grayscale image pixel by pixel; Assuming that the grayscale value of worsted wool fabric follows Gaussian distribution, calculate the mean μ and variance σ of the wool fabric; The mean μ is replaced by the median of the reference sample, that is: ; The variance σ is calculated using the unbiased estimation formula, namely: ; Set the decision thresholds T1 and T2; ; ; If the pixel to be detected x satisfies , it is judged as normal fabric; otherwise, it is judged as defective; Then, the detection window moves to the next pixel for detection until the entire image is detected; Where x is the pixel sample in the detection unit; y i is the pixel sample in the reference unit; t is the threshold factor.

[0011] Preferably, the sliding window structure of the sliding window CFAR detector includes: a detection unit, a protection unit and a reference unit; the reference unit is used to set the detection threshold; the protection unit is used to prevent larger defects from affecting the detection threshold.

[0012] Preferably, according to the preset It is required to determine t; where, is the false alarm probability (false detection rate).

[0013] Preferably, the calculation formula of t is: ; Among them, e is a natural constant.

[0014] Preferably, the detection method is suitable for automatic detection of points, lines, balls, holes, and other irregularly shaped defects in woolen fabrics.

[0015] The embodiment of the present invention provides a dual-threshold adaptive CFAR detection method for worsted fabric defects, which has the following beneficial effects: (1) The present invention introduces radar target CFAR detection technology into the defect detection of woolen fabrics, which can realize the detection of woolen fabrics with different textures, tones, and flatness, and can adaptively provide an extremely low false detection rate, while achieving the maximum probability of detecting defects of different shapes and sizes; (2) The present invention not only ensures the reliability and effectiveness of fabric defect detection and realizes adaptive CFAR detection of fabric defects, but also has the advantages of low false detection rate, high defect detection rate, simple algorithm, fast speed and adaptability. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 Schematic diagram of a sliding window for woolen fabric defect detection in the present invention; Figure 2 This is a flow chart of a specific implementation method of CFAR detection of woolen fabric defects in the present invention; in, Figure 2 The judgment criteria (4) and (6) in the patent refer to formulas (4) and (6) in this patent. DETAILED DESCRIPTION

[0017] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts shall fall within the scope of protection of the present invention.

[0018] The following is combined with Figure 1-2 , and specific implementation methods are further described to illustrate the present invention.

[0019] In response to the problems mentioned in the above background technology, an embodiment of the present invention provides a dual-threshold adaptive CFAR detection method for worsted fabric defects to solve the above technical problems. The specific technical solution is as follows.

[0020] 1. Design concept of a dual-threshold adaptive CFAR detection method for worsted fabric defects provided by the present invention Radar detection technology is primarily used for detecting targets such as aircraft and ships. Its detection principle is based on the difference in electromagnetic wave reflection between the target and the surrounding environment. The radar system receives the echo signal reflected by the target and compares it with a pre-set or dynamically calculated threshold. When the echo intensity exceeds the threshold, the target is determined to be present, achieving adaptive detection and accurately identifying and locating the target object.

[0021] Woolen fabric defects appear as irregular shapes, such as dots, lines, spheres, and holes. These defects appear bright or dark in grayscale images, distinguishing them from the grayscale of normal fabric. Therefore, the CFAR method can be used to detect woolen fabric defects based on the differences between the grayscale images of woolen fabric defects and those of normal fabrics.

[0022] The structure and decision criteria of the wool fabric defect detector are different from those of the radar target CFAR detector. The method proposed in this invention is applicable to the automatic detection of irregular shaped defects such as points, lines, balls, and holes in worsted wool fabrics. The specific detection principle is as follows: (1) The present invention adopts sliding window detection technology to detect fabric defects. Figure 1 As shown; Figure 1 The red unit in the image is the detection unit (pixel), and the surrounding green units are called reference units, which are used to set the detection threshold. The part between the detection unit and the reference unit is called the protection unit, which is used to prevent larger defects from affecting the detection threshold. represents a pixel sample in a reference cell, and Represents a pixel sample in the detection unit.

[0023] Assume that the grayscale of wool fabric pixels obeys Gaussian distribution, that is: ; (1) In formula (1), Represents the average value of woolen fabrics, represents the variance of the woolen fabric; in order to overcome the interference in the reference sample, the median of the reference sample is used instead of the mean μ, that is: ; (2) Variance of woolen fabrics The following unbiased estimation formula is used: ; (3) (2) If the grayscale value x of the pixel sample in the detection unit meets the judgment criteria: or ; (4) If the pixel in the detection unit is defective, it is judged as normal fabric.

[0024] According to the false alarm probability (false detection rate) requirement P fa , the threshold factor t is determined by the following formula: ; (5) make ,but ; (6) From formula (6), we can see that the false positive rate Average value of woolen fabric and variance This means that the detector has a false detection rate of 0.01% for woolen fabrics with different textures, tones, and flatness. Can be adaptively maintained at a low level superior.

[0025] At the same time, it can detect fabric defects with the highest probability. In this way, this patented detection method provides reliability and effectiveness for fabric defect detection, and realizes adaptive CFAR detection of fabric defects.

[0026] It's important to note that the texture, color, and smoothness of the worsted fabrics being inspected in production workshops vary greatly, and defects on these fabrics also vary in size and shape. Existing methods typically suffer from high false detection rates, high rates of missed detection of small objects, and an inability to adaptively adjust detection thresholds in such time-varying, non-uniform backgrounds. 2. Specific embodiments (1) Detection method The wool fabric image captured by the CCD camera is converted from color to grayscale to obtain a grayscale image, which is then sent to the sliding window CFAR detector.

[0028] The sliding window CFAR detector uses the reference samples in the reference sliding window to estimate the mean μ and variance σ of the woolen fabric, and then calculates the error rate P according to the set error rate P. fa The threshold factor t is determined to give the decision thresholds T1 and T2; ; .

[0029] If the pixel to be detected x satisfies , it is judged as normal fabric. Otherwise, it is judged as a defect. Then, the detection window moves to the next pixel for detection until the entire image is detected.

[0030] (2) Test results In the specific implementation, 100 grayscale images of worsted wool fabrics containing different types of defects were selected for testing, with an image resolution of 1920×1080. The parameters of the sliding window CFAR detector were set as follows: the reference unit width was 3 pixels, the protection unit width g was 10 pixels, and the false alarm probability was 0. Set as .

[0031] According to statistics, there are 865 defects in 100 images, and the method of the present invention successfully detects 842 of them, with a defect detection rate of 97.34%. In the normal area without defects in the 100 images, only 1.80 pixels in each picture are misjudged as defects on average, with an average false positive rate as low as .

[0032] The results show that the present invention effectively achieves high-precision detection of worsted wool fabric defects through technical means such as sliding window structure design, median estimation mean and double threshold judgment, while controlling the false detection rate to an extremely low level, which fully demonstrates the feasibility and effectiveness of the technical solution in terms of high defect detection rate and low false detection rate.

[0033] In addition, the present invention breaks through the limitations of deep learning detection. It does not require massive labeled data or retraining, and can achieve rapid detection of a single image on ordinary hardware, effectively solving the problems of high data dependence, high hardware requirements, and unstable detection in traditional methods.

[0034] Although the above describes the specific embodiments of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without any creative work are still within the scope of protection of the present invention.

Claims

1. A double-threshold adaptive CFAR detection method for worsted fabric defects, characterized in that: The following steps are involved: A CCD camera is used to capture color images of worsted fabrics; Convert the collected color image into a grayscale image to obtain the grayscale value of each pixel; A sliding window CFAR detector is used to traverse the target grayscale image pixel by pixel; Assuming that the grayscale value of worsted wool fabric follows Gaussian distribution, calculate the mean μ and variance σ of the wool fabric; The mean μ is replaced by the median of the reference sample, that is: ; The variance σ is calculated using the unbiased estimation formula, namely: ; Set the decision thresholds T1 and T2; ； ； If the pixel x to be detected satisfies , it is judged as normal fabric; otherwise, it is judged as defective; Then, the detection window moves to the next pixel for detection until the entire image is detected; Where x is the pixel sample in the detection unit; y i is the pixel sample in the reference unit; t is the threshold factor.

2. The double-threshold adaptive CFAR detection method for worsted fabric defects according to claim 1, characterized in that: The sliding window structure of the sliding window CFAR detector includes: a detection unit, a protection unit and a reference unit; The reference unit is used to set the detection threshold; the protection unit is used to prevent larger defects from affecting the detection threshold.

3. The double-threshold adaptive CFAR detection method for worsted fabric defects according to claim 1, characterized in that: Determine t based on the preset Pfa requirement, where Pfa is the false alarm probability.

4. The double-threshold adaptive CFAR detection method for worsted fabric defects according to claim 2, characterized in that: The formula for calculating t is: ； Among them, e is a natural constant.

5. The double-threshold adaptive CFAR detection method for worsted fabric defects according to claim 1, characterized in that: The detection method is suitable for automatically detecting points, lines, balls, holes, and other irregular-shaped defects in woolen fabrics.

Images