A paper disease detection method and system using FFT

Abstract

The application discloses a paper disease detection method and system using FFT, and the method comprises the following specific steps: S1, image preprocessing, converting the acquired image into a gray image; S2, acquiring an image gray histogram; S3, subtracting the gray level corresponding to the highest frequency of the gray level in the gray histogram from the original gray value and squaring; S4, performing fast Fourier transform, and then taking an absolute value and normalizing; S5, summing up the FFT of each point to obtain a judgment value Z, and using an experience threshold to judge whether the paper is a disease paper. The application adopts the fast Fourier transform algorithm (FFT), is fast in calculation, and greatly reduces the missing detection rate of the paper disease through the signal processing mode of converting the signal from the spatial domain to the frequency domain.

Description

Technical Field

[0001] This invention relates to the field of paper image processing technology, and in particular to a paper defect detection method and system using FFT. Background Technology

[0002] As we all know, paper is closely related to people's daily lives, and the paper industry ranks third among the world's important industries. Most developed countries in the world attach great importance to the development of the paper industry, and it has even become an important industry in some developed countries (such as the United States and Germany). While paper production has increased, it has also brought related quality problems, especially the surface quality of the paper. However, for certain special papers, such as bank drafts, graduation certificates, national invoices, and property certificates, appearance defects are unacceptable. Ensuring quality while maintaining production volume has become a challenge for paper companies. With the continuous improvement of modern papermaking technology, paper machine speeds have increased, and paper widths have widened, leading to a rapid increase in paper production. In the modern papermaking industry, there are paper machines that can operate continuously around the clock, with speeds reaching 1800 meters per minute and paper widths of 10 meters. While production volume has increased, so have quality problems. Paper quality inspection is generally done manually, but traditional manual inspection is not only inefficient but also has low accuracy and cannot be done in real time, making it completely unsuitable for the modern papermaking industry. If numerous paper defects occur during production and operators fail to detect them in time, a large quantity of paper products will be substandard, significantly increasing production costs and causing substantial economic losses for the company. Therefore, an intelligent online paper defect detection system will be crucial for ensuring high-quality paper production in the modern paper industry.

[0003] With the development of science and technology, machine vision-based inspection methods have emerged as a new approach to inspection. To adapt to modern industrial production, thanks to significant technological advancements and accumulation, modern machine vision technology has been widely applied in various industrial sectors, such as the inspection of large-scale integrated circuit boards, the inspection and identification of precision instruments and parts, inspection and identification in automobile production, and missile navigation. A complete machine vision system has become an important criterion for judging the modernization of industrial production. In the paper industry, convolutional neural network-based inspection methods have begun to appear extensively. However, due to the continuous increase in paper machine speeds, the accuracy of these methods cannot meet the demands of the paper industry. Furthermore, while using traditional machine vision's support vector machine (SVM) algorithm for paper defect detection addresses the speed issue, its low accuracy remains a problem.

[0004] The above technologies and analyses indicate that there is a technological gap in a paper defect detection system applicable to ultra-high-speed, ultra-wide-width paper defects. Summary of the Invention

[0005] To ensure the quality of industrial products leaving the factory, the present invention addresses the current traditional quality inspection methods that rely heavily on manual labor for multiple inspections, re-inspections, and spot checks. Each product requires multiple manual inspections, which not only wastes a lot of manpower and resources, but also makes it impossible to achieve high stability and high quality of industrial products due to the subjectivity, uncertainty, and inconsistency of quality standards in manual inspections. The present invention provides a paper defect detection method and system using FFT.

[0006] The objective of this invention is achieved through the following technical solution:

[0007] A paper defect detection method using FFT includes the following specific steps:

[0008] S1: Image preprocessing, converting the acquired image into a grayscale image;

[0009] S2: Obtain the image grayscale histogram;

[0010] S3: Subtract the gray level corresponding to the highest frequency of gray level in the gray level histogram from the original image's gray level value, and then square it.

[0011] S4: Perform a fast Fourier transform, then normalize the absolute value;

[0012] S5: Sum the FFT values ​​at each point to obtain the judgment value Z, and use an empirical threshold to determine whether it is a defective paper.

[0013] Furthermore, step S1 specifically involves: reducing the RGB image captured by the camera to a grayscale image.

[0014] Furthermore, step S2 specifically involves: acquiring a grayscale histogram of the paper image, analyzing and obtaining the grayscale levels of the paper defect-free area and the paper defect area, wherein the grayscale level corresponding to the highest frequency of grayscale level occurrence is the paper defect-free area in the original image.

[0015] Furthermore, step S4 specifically involves: using Fast Fourier Transform to transform the grayscale distribution function of the paper image into the frequency distribution function of the image, and then performing absolute value normalization analysis.

[0016] Furthermore, step S5 specifically involves: traversing each row of the paper defect image to obtain the FFT spectrum of each row; summing the amplitudes of each point in the FFT spectrum corresponding to the row to obtain the judgment value Z corresponding to the row; and finally determining whether the row is a paper defect row by using a pre-set empirical threshold. If Z is greater than the empirical threshold, then the row is a paper defect row; otherwise, it is not a paper defect row.

[0017] A paper defect detection system using FFT includes a line scan camera, an image acquisition card, and a defect paper detection computing terminal. The output of the line scan camera is connected to the input of the image acquisition card, and the output of the image acquisition card is connected to the defect paper detection computing terminal. The image acquisition card receives the analog video signal acquired by the line scan camera, acquires and quantizes the analog video signal, and finally converts it into a digital signal, which is then input and stored in the defect paper detection computing terminal.

[0018] Furthermore, a paper defect detection system employing FFT also includes LED lights and a paper conveyor belt, with the paper conveyor belt located directly below the lens of the line scan camera and the LED lights located directly below the paper conveyor belt.

[0019] The beneficial effects of this invention are:

[0020] This invention utilizes machine vision. Due to the precision of machine vision, the detection accuracy of products can be greatly increased, significantly reducing the probability of defective paper products leaving the factory. This is very attractive in the specialty paper manufacturing industry, which has very strict product requirements, and is also the aspect that the specialty paper manufacturing industry pays the most attention to.

[0021] This invention employs a "CCD line scan camera + image acquisition card + PC" model, which is fully applicable to modern high-speed paper production workshops. Because the CCD line scan camera can directly acquire pixels from each line of the image, compared to traditional cameras, it eliminates the need for image stitching during acquisition. This not only significantly improves the image acquisition rate and reduces preprocessing workload but also indirectly improves production efficiency and timeliness.

[0022] The Fast Fourier Transform (FFT) algorithm used in this invention is not only fast in computation, but also significantly reduces the false negative rate of paper defects by transforming the signal from the spatial domain to the frequency domain. Furthermore, watermarks from the manufacturer are often printed on paper during production, which can be confused with paper defects, leading to misjudgments during paper inspection. The FFT algorithm of this invention can adjust the empirical threshold for FFT discrimination to classify paper with watermarks as defect-free, eliminating the interference of watermarks and thus improving the accuracy of identification. The algorithm of this invention is applicable to the detection of all types of paper defects and can bring significant economic benefits. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0024] Figure 1 This is a flowchart of the method of the present invention;

[0025] Figure 2 This is a schematic diagram of the system structure of the present invention;

[0026] Figure 3 These are schematic diagrams illustrating the effects of each step in this invention;

[0027] Figure 4 These are various types of paper defects;

[0028] Figure 5 This is a Z-value diagram of various paper defect types according to the present invention. Detailed Implementation

[0029] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0030] 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 a part of the embodiments of the present invention, and not all of them. 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. In addition, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.

[0031] like Figure 2 As shown, this invention discloses a paper defect detection system using FFT, including a line scan camera, an image acquisition card, and a defect paper detection computing terminal; the output end of the line scan camera is connected to the input end of the image acquisition card, and the output end of the image acquisition card is connected to the defect paper detection computing terminal. The image acquisition card receives the analog video signal acquired by the line scan camera, acquires and quantizes the analog video signal, and finally converts it into a digital signal, which is then input and stored in the defect paper detection computing terminal.

[0032] Furthermore, a paper defect detection system employing FFT also includes LED lights and a paper conveyor belt, with the paper conveyor belt located directly below the lens of the line scan camera and the LED lights located directly below the paper conveyor belt.

[0033] Specifically:

[0034] (1) Selection of image acquisition card

[0035] With the advent of the information and automation era, machine vision technology is widely used for signal processing to improve the intelligence level of production processes and quality inspection and supervision. We know that machine vision systems generally acquire analog signals. To recognize or process these signals, they need to be converted into digital signals. An image acquisition card connects the image acquisition section and the processing section. Through the image acquisition card, it receives analog video signals from the camera, acquires and quantizes these signals, and finally converts them into digital signals, which are then input and stored in the output device. Generally, image signal transmission requires high speed, which ordinary transmission interfaces cannot meet. Therefore, an image acquisition card is needed to assist in this process. Acquisition cards come in various types and specifications. Although their designs and characteristics differ, most acquisition cards share the same basic principle.

[0036] Based on this, we selected the Matrox image acquisition card through extensive comparative experiments. The Matrox image acquisition card not only has extremely low pixel jitter and grayscale noise, but its PCI bus has an average transfer rate of 60-100MB / s, which meets both the requirements for transmission speed and the high quality requirements of the acquired image signal.

[0037] (2) Camera selection

[0038] Considering the practical scenarios in paper production, we initially focused on linear scan cameras. As the name suggests, linear scan cameras capture images linearly and utilize linear image sensors. Linear scan image sensors are primarily CCDs, although some linear scan CMOS image sensors have also appeared on the market in recent years. However, considering cost-effectiveness and the question of whether camera performance is excessive, we chose to select from CCD cameras.

[0039] To meet the demands of ultra-high speed and ultra-wide format in paper production, the first consideration should be the line scan frequency of the CCD camera, i.e., the maximum number of lines of image that the camera can capture per second. This frequency must be greater than 29150Hz and compatible with Camera Link or Gigabit Ethernet interfaces. Therefore, cameras meeting these requirements are mainly found in models from Dalsa, Basla, and Atmel, and their main parameters are shown in Table 1.

[0040] Table 1 Comparison of Camera Performance

[0041]

[0042]

[0043] Based on comparison, the camera selected for this project is the Dalsa S3-24-02K40. The Dalsa 02K40 features a more compact design and utilizes a gigabit dual serial high-speed standard, achieving up to 3 times the response time and 2 times the transmission speed (80 megapixels per second) compared to the Spyder2 series cameras. Furthermore, its interface has been simplified, and it provides a low-noise, unified output using correlated double sampling and embedded correction. In addition to its high performance and cost-effectiveness, the Dalsa S3-24-02K40 industrial digital camera also possesses advanced features such as flat-field correction and CDS (correlated double sampling).

[0044] (3) Selection of optical lenses

[0045] Based on the lens specifications analysis above, this system uses a SCHNEIDER high-performance lens, whose minimal aberrations ensure excellent image quality. Combined with a DALSA 02K40 monochrome CCD camera, this configuration produces very clear images with excellent uniformity. Furthermore, the software allows for further white balance correction, fully meeting the inspection requirements.

[0046] (4) Selection of light source

[0047] In online scanning projects, commonly used light sources include LED lights, halogen lamps (fiber optic lamps), and high-frequency fluorescent lamps. Based on the above analysis and extensive experimental data, the best choice for the light source in this project is a high-brightness white LED light. The main characteristics of a high-brightness white LED light are:

[0048] ① Due to their small size, LED lights are easy to form various light sources, thus creating different types of lighting such as lines, surfaces, and rings. One-dimensional linear CCD lighting sources, because their physical size is smaller than fiber optic sources, can provide tilted illumination for detection from any angle;

[0049] ② It also features low power consumption and low operating voltage.

[0050] ③ DC power supply, no flicker.

[0051] (5) Computer

[0052] Disease detection computing terminal.

[0053] (6) Conveyor belt device

[0054] The paper is transported at high speed, and LED lights are used under the conveyor belt to project onto the line scan camera to improve the clarity of the image acquisition.

[0055] like Figure 1As shown, this invention discloses a paper defect detection method using FFT, comprising the following specific steps:

[0056] S1: Image preprocessing, converting the acquired image into a grayscale image;

[0057] S2: Obtain the image grayscale histogram;

[0058] S3: Subtract the gray level corresponding to the highest frequency of gray level in the gray level histogram from the original image's gray level value, and then square it.

[0059] S4: Perform a fast Fourier transform, then normalize the absolute value;

[0060] S5: Sum the FFT values ​​at each point to obtain the judgment value Z, and use an empirical threshold to determine whether it is a defective paper.

[0061] Specifically:

[0062] (1) Algorithm Flow

[0063] ① Image preprocessing: convert the acquired image into a grayscale image.

[0064] Since the images captured by the camera are RGB images with three dimensions, subsequent processing is relatively complex. Therefore, the images are first reduced to grayscale.

[0065] ② Obtain the image grayscale histogram

[0066] A grayscale histogram is a function of gray levels, describing the number of pixels at each gray level in an image, reflecting the frequency of each gray level. The horizontal axis represents the gray level, and the vertical axis represents the frequency of each gray level. The gray levels range from 0 to 255 (minimum 0 for black, maximum 255 for white). Figure 3 As shown in (b), we selected dark spot paper disease as a sample to obtain its grayscale histogram. The figure shows that the grayscale levels in the range of 60 to 80 account for the vast majority of the dark spot paper disease, while from... Figure 3 As can be seen from (a), the area without paper defects in the image occupies most of the entire image. Therefore, it can be inferred that the gray level corresponding to the highest frequency of gray level in the gray-level histogram is approximately the area without paper defects in the original image, while 20 to 80 correspond to the area with paper defects.

[0067] ③ Subtract the gray level corresponding to the highest frequency of gray level in the gray level histogram from the original image's gray level value, and then square the result.

[0068] This step allows differences such as dark spots or bright spots to be magnified equally, which is more conducive to the processing of fast Fourier transform in subsequent steps.

[0069] ④ Perform FFT (Fast Fourier Transform), then normalize the absolute value.

[0070] The frequency of an image is an indicator of the degree of drastic change in grayscale levels. It represents the gradient of grayscale levels in a spatial plane; a faster change in grayscale results in a higher frequency, and vice versa. For example, a large desert area in an image is a region with slow grayscale changes, corresponding to a low frequency value; conversely, an edge region with drastic changes in surface properties is a region with drastic grayscale changes, corresponding to a higher frequency value. From a physical perspective, the Fourier transform converts an image from the spatial domain to the frequency domain, and its inverse transform converts it back to the spatial domain. Since information in the spatiotemporal domain is difficult to process in images, the Fourier transform can be used to transform the image's grayscale distribution function into its frequency distribution function for analysis and processing. However, because the Fourier transform involves convolution calculations, which are computationally intensive and relatively slow for computers, the Fast Fourier Transform (FFT) was developed, significantly reducing the computational load and making signal processing by computers possible. The Fast Fourier Transform (FFT) is a fast algorithm for the Fourier Transform, which can reduce computation and storage overhead, making it particularly advantageous for hardware implementation. Therefore, we adopted FFT to process images.

[0071] ⑤ Sum the FFT values ​​at each point to obtain the judgment value – Z. Use an empirical threshold to determine whether the paper is defective.

[0072] like Figure 3 As shown in (d), by traversing each row of the paper defect image, the FFT spectrum of each row is obtained (each colored line represents the FFT spectrum of each row). The summation of the amplitudes of each point in the FFT spectrum corresponding to that row yields the judgment value Z corresponding to that row. Figure 3 As shown in (e), the row is then judged to be a paper defect row by a pre-set empirical threshold. If Z is greater than the empirical threshold, the row is a paper defect row; otherwise, it is not a paper defect row.

[0073] This invention employs a "CCD line scan camera + image acquisition card + PC" model, which is fully applicable to modern high-speed paper production workshops. Because the CCD line scan camera can directly acquire pixels from each line of the image, compared to traditional cameras, it eliminates the need for image stitching during acquisition. This not only significantly improves the image acquisition rate and reduces preprocessing workload but also indirectly improves production efficiency and timeliness.

[0074] like Figure 4 and Figure 5As shown, paper defects are diverse in type, with complex textures and varied shapes. Therefore, using traditional machine learning or convolutional neural network algorithms would significantly increase the false negative rate of paper defects. Furthermore, because convolutional neural networks require substantial computational resources, this method cannot meet the demands of today's ultra-high-speed, ultra-wide-format paper production. The Fast Fourier Transform (FFT) algorithm we use not only offers high computational speed but also significantly reduces the false negative rate of paper defects by transforming the signal from the spatial domain to the frequency domain. Moreover, from... Figure 3 It is known that paper is imprinted with the manufacturer's watermark during the production process, which can be confused with paper defects, leading to misjudgments during paper inspection. Our FFT algorithm can adjust the empirical threshold for FFT discrimination to classify paper with watermarks as defect-free, eliminating the interference of watermarks on identification and thus improving the accuracy of identification. Our algorithm is applicable to the detection of all paper defects and can bring significant economic benefits.

[0075] The above description is merely a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and environments, and can be altered within the scope of the concept described herein through the above teachings or related technologies or knowledge. Modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the protection scope of the appended claims.

Claims

1. A paper disease detection method using FFT, characterized by, The specific steps include the following: S1: Image preprocessing, converting the acquired image into a grayscale image; S2: Obtain the image grayscale histogram; S3: Subtract the gray level corresponding to the highest frequency of gray level in the gray level histogram from the original image's gray level value, and then square the result. S4: Perform a fast Fourier transform, then normalize the absolute value; S5: Sum the FFT values ​​at each point to obtain the judgment value Z, and use an empirical threshold to determine whether it is a defective paper; Specifically, step S4 involves: using Fast Fourier Transform to transform the grayscale distribution function of the paper image into the frequency distribution function of the image, and then performing absolute value normalization analysis. Step S5 specifically involves: traversing each row of the paper defect image to obtain the FFT spectrum of each row; summing the amplitudes of each point in the FFT spectrum corresponding to the row to obtain the judgment value Z corresponding to the row; and finally determining whether the row is a paper defect row by using a pre-set empirical threshold. If Z is greater than the empirical threshold, then the row is a paper defect row; otherwise, it is not a paper defect row.

2. The paper defect detection method using FFT according to claim 1, characterized in that, Step S1 specifically involves: reducing the RGB image captured by the camera to a grayscale image.

3. The paper defect detection method using FFT according to claim 1, characterized in that, Step S2 specifically involves: acquiring a grayscale histogram of the paper image, analyzing the grayscale levels of the paper defect-free area and the defective area, wherein the grayscale level corresponding to the highest frequency of grayscale level occurrence is the defect-free area in the original image.

4. A paper defect detection system employing FFT, using the paper defect detection method employing FFT as described in any one of claims 1-3, characterized in that, It includes a line scan camera, an image acquisition card, and a defective paper detection computing terminal. The output end of the line scan camera is connected to the input end of the image acquisition card, and the output end of the image acquisition card is connected to the defective paper detection computing terminal. The image acquisition card receives the analog video signal acquired by the line scan camera, acquires and quantizes the analog video signal, and finally converts it into a digital signal, which is then input and stored in the defective paper detection computing terminal.

5. A paper defect detection system using FFT according to claim 4, characterized in that, It also includes LED lights and a paper conveyor belt, which is located directly below the lens of the line scan camera, and the LED lights are located directly below the paper conveyor belt.

Images