Aluminum laminated film surface defect online detection method and system based on FPGA image processing

Abstract

The invention belongs to the technical field of image processing, and particularly relates to an aluminum laminated film surface defect online detection method and system based on FPGA image processing. Comprising the following steps: collecting image data of the surface of an aluminum laminated film through a line scanning CMOS sensor in a high-speed line scanning camera, and installing the high-speed line scanning camera in a stainless steel enclosed hood in a vacuum cavity in a sealing manner; performing Fourier transform on the acquired image data to remove high-frequency noise, performing flat field processing to enable a defect-free part to be at a 128-gray-scale position, and performing 5 * 5 filtering or 7 * 7 filtering processing; performing ternary processing on the filtered image data, and setting 128 positive and negative threshold values to enable the area above the positive threshold value to be white defects and the area below the negative threshold value to be black defects; through the high-speed line scanning camera, the high-brightness LED strip-shaped light source and the image processing LSI based on the FPGA, the aluminum laminated film surface defect detection precision can be improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of image processing, and particularly relates to an on-line detection method and system for surface defects of aluminized films based on FPGA image processing, and particularly relates to the heat dissipation design, compact installation structure of a high-speed line-scan camera in a vacuum environment, and a high-speed image processing algorithm based on FPGA. Background Art

[0002] Aluminizing of a vacuum coater is to heat and melt aluminum metal to evaporation in a vacuum state, and aluminum atoms condense on the surface of a polymer material to form an extremely thin aluminum layer. Vacuum aluminizing requires the substrate surface to be smooth, flat, and of uniform thickness; with appropriate stiffness and coefficient of friction; large surface tension, good thermal performance, and being able to withstand the thermal radiation of the evaporation source and the condensation heat; common aluminized substrates include films such as polyester (PET), polypropylene (PP), polyamide (PA), polyethylene (PE), polyvinyl chloride (PVC), etc. If the vacuum degree in the coater is too low, brown stripes or uneven aluminum layer thickness will occur; if the system tension is not well controlled, or there are problems with the cooling system, defects such as stretching and deformation of the film due to heat will also occur; therefore, the equipment must precisely control the winding speed (generally above 900 - 1000 meters per minute), aluminum feeding speed, and heating current of the evaporation boat to obtain the required aluminum layer thickness of the product.

[0003] Generally, during the aluminizing process of a vacuum coater, surface defects such as bubbles, pinholes, and uneven stripes will be generated due to improper various controls. Therefore, in order to timely feedback the control parameters of the coater through the defects, it is very important to monitor the surface defects that occur during the aluminizing process of the material in real time.

[0004] However, there are still the following difficulties in realizing real-time on-line detection in the vacuum coater cavity: 1. Due to the high-speed operation of the aluminized film, a conventional image processing system cannot complete the defect detection of the aluminized film surface in real time; 2. The problem of heat dissipation in the vacuum cavity: The high-speed camera and the high-brightness light source will generate a large amount of heat during the detection process. Since the air flow speed in the vacuum cavity is basically zero, the air-cooling condition through an electric fan does not exist. How to reduce the temperature of the high-speed camera and the high-brightness light source in the vacuum cavity is a great challenge; 3. The installation problem of the detection system in a compact cavity: The current vacuum coater is not designed considering the installation of on-line detection equipment during the design process. Therefore, how to use the narrow space to complete the detection work is a relatively difficult problem. Summary of the Invention

[0005] The purpose of the present invention is to provide an on-line detection method and system for surface defects of aluminized films based on FPGA image processing, which can stably operate in a vacuum environment and have the on-line system with high-speed and high-precision detection capabilities.

[0006] To solve the above technical problems, the present invention provides an on-line detection method for surface defects of aluminized films based on FPGA image processing. The aluminized film is formed in real time online in the vacuum chamber of a vacuum aluminizing machine, and the method includes the following steps:

[0007] Step S1: Collect the image data of the surface of the aluminized film through the line scan CMOS sensor in the high-speed line scan camera, and the high-speed line scan camera is sealed and installed in the stainless steel enclosure in the vacuum chamber;

[0008] Step S2: Remove high-frequency noise from the collected image data through Fourier transform, make the defect-free part at the position of 128 gray levels through flat-field processing, and perform 5×5 filtering or 7×7 filtering processing;

[0009] Step S3: Binarize the filtered image data. By setting the positive and negative thresholds of 128, the area above the positive threshold is the white defect, and the area below the negative threshold is the black defect;

[0010] Step S4: Calculate the image feature quantities, and perform labeling connectivity processing on the binarized image to obtain image feature quantities including the areas, lengths, widths, and gray-scale volume of the white and black defects;

[0011] Step S5: Process the defects segmented by multiple frames of images, and crop a 256×256 pixel defect image with the defect centroid as the center;

[0012] Step S6: Pack the defect image and the corresponding defect information, and upload them to the host computer in the form of TCP / IP;

[0013] Step S7: Draw a defect map according to the image coordinate information;

[0014] Step S8: Use a deep learning network to classify and recognize the defect images.

[0015] Preferably, in the step S1, the stainless steel enclosure is hoisted on the cavity reinforcing rib, the cavity reinforcing rib is located above the inner side of the transparent plexiglass window of the vacuum aluminizing machine, and the stainless steel enclosure further includes a highly transparent glass provided thereon, which is convenient for providing a shooting window for the high-speed line scan camera disposed perpendicular to the aluminized film.

[0016] Preferably, it further includes a high-brightness LED strip light source disposed perpendicular to the aluminized film, which can provide uniform, stable, and high-intensity light source illumination to ensure the clarity and contrast of the images captured by the high-speed line scan camera; and the high-brightness LED strip light source is installed at the position of the original fluorescent lamp of the vacuum aluminizing machine.

[0017] Preferably, a semiconductor water cooling module is mounted on both the high-speed line scan camera and the high-brightness LED bar light source through thermal conductive adhesive for heat dissipation.

[0018] The present invention also provides an on-line detection system for surface defects of aluminized film based on FPGA image processing, which adopts an on-line detection method for surface defects of aluminized film based on FPGA image processing as described above, and includes: a high-speed line scan camera, a high-brightness LED bar light source, a semiconductor water cooling module, and an FPGA-based image processing LSI; the FPGA-based image processing LSI includes a preprocessing unit, a line buffer, a processing unit, and a postprocessing unit connected in sequence; the processing unit is composed of 16 PE units with multiplication functions and 15 gray-scale search engine units GSEU.

[0019] Preferably, the FPGA-based image processing LSI further includes a smoothing filter, an edge enhancement, and a rank filter circuit to realize the processing of general images.

[0020] Preferably, the FPGA-based image processing LSI further includes a block matching hardware based on the sequential similarity detection algorithm SDA to realize the gray-scale search function.

[0021] Preferably, the FPGA-based image processing LSI further includes: an SH-4 RISC processing engine, a memory controller, a Flash ROM, a RAM, an SH bus, and 49 parallel processing units; the SH-4 RISC processing engine, the memory controller, the Flash ROM, and the RAM are interconnected through the SH bus, the memory controller is connected to the preprocessing unit and the postprocessing unit, and 49 parallel processing units are connected to the line buffer and the postprocessing unit.

[0022] Preferably, the FPGA-based image processing LSI further includes a PCI expansion interface and a video input / output interface. The PCI expansion interface is used to connect to the PCI bus, and the video input / output interface is used to connect to the high-speed line scan camera and a display device, and the video input / output interface is interconnected with the memory controller.

[0023] Compared with the prior art, the present invention has the following beneficial effects:

[0024] 1. Through the high-speed line scan camera, the high-brightness LED bar light source, and the FPGA-based image processing LSI, the present invention can improve the accuracy of detecting surface defects of aluminized film, enabling the FPGA-based pipeline processing speed to reach 16 MHz, the processing time of a single frame (1024×1024 pixels) ≤ 1.4 ms, the system can detect defects with a diameter ≥ 0.1 mm in real time, and the defect recognition rate ≥ 99.5%.

[0025] 2. The present invention solves the problem of reducing the temperatures of a high-speed line-scan camera and a high-brightness LED bar light source in a vacuum chamber by adding a semiconductor water-cooling module for water-cooling heat dissipation. The water-cooling system reduces the camera temperature by 30% and the light source temperature by 25%, ensuring continuous operation in a vacuum environment.

[0026] 3. The present invention installs a high-brightness LED bar light source at the original position of a fluorescent lamp by removing the fluorescent lamp from an original vacuum aluminizing machine. A dust-proof cover of the high-speed line-scan camera is suspended above the inner side of an observation window of transparent plexiglass, i.e., under a cavity stiffener. The dust-proof cover is of a sealed structure, and a high-speed camera and its semiconductor water-cooling module are installed inside, thus solving the installation problem of a detection system in a compact cavity and achieving the installation compatibility design index of compact design adapted to the existing aluminizing machine cavity without modifying the equipment structure.

[0027] 4. The present invention seals the camera and its related key devices with a fully sealed stainless-steel cover and uses vacuum aviation plugs to connect the power supply line and the signal line to eliminate the risk of component explosion. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 is a schematic flow chart of an on-line detection method for surface defects of aluminized films based on FPGA image processing according to the present invention.

[0029] Figure 2 is a schematic diagram of the internal structure of a vacuum aluminizing machine cavity after installation of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0030] The following further describes the present invention in detail with reference to the drawings and specific embodiments. The advantages and features of the present invention will be clearer according to the following description. It should be noted that the drawings are all in a very simplified form and use non-precise scales, only for the purpose of facilitating and clearly assisting in explaining the objectives of the embodiments of the present invention.

[0031] As Figure 1 shown, an embodiment of the present invention discloses an on-line detection method for surface defects of aluminized films based on FPGA image processing. The aluminized film is formed in real time on-line in a vacuum chamber of a vacuum aluminizing machine, and the method includes the following steps:

[0032] Step S1: Collect image data of the surface of the aluminized film through a line-scan CMOS sensor in a high-speed line-scan camera; and the high-speed line-scan camera is sealed and installed in a stainless-steel enclosure in the vacuum chamber;

[0033] Step S2: Remove high-frequency noise from the collected image data through Fourier transform, make the defect-free part at the position of 128 gray levels through flat-field processing, and perform 5×5 filtering or 7×7 filtering processing;

[0034] Step S3: Binarize the filtered image data. By setting the positive and negative thresholds of 128, the area above the positive threshold is the white defect, and the area below the negative threshold is the black defect;

[0035] Step S4: Calculate the image feature quantities. Based on the binarized image, perform the labeling connection process of the image to obtain the image feature quantities including the areas, lengths, widths, and gray volume of the white and black defects;

[0036] Step S5: Process the defects segmented by multiple frames of images. Crop a 256×256 pixel defect image with the defect centroid as the center;

[0037] Step S6: Package the defect image and the corresponding defect information and upload them to the host computer in the form of TCP / IP;

[0038] Step S7: Draw a defect map according to the image coordinate information;

[0039] Step S8: Use a deep learning network to perform classification training and recognition on the defect images.

[0040] As Figure 2 shown, a fluorescent lamp and a transparent organic glass observation window are installed in the cavity of the aluminizing machine, which was originally used for operators to observe the defects of the aluminized film in real time. Currently, our detection system has removed the fluorescent lamp and installed a high-brightness LED light source in its original position. Above the inner side of the transparent organic glass window, a dust-proof cover of a high-speed line-scan camera is suspended under the cavity stiffener. The dust-proof cover is of a sealed structure, and a high-speed camera and its water-cooling device are installed inside.

[0041] Line-scan cameras basically cannot be used in a vacuum environment. A large number of components such as capacitors have the risk of explosion in a vacuum environment. Also, the solder joints of the circuit board may cause problems such as poor soldering in a vacuum environment. Therefore, how to use the existing line-scan cameras in a vacuum environment is a feature of the present invention. The present invention seals the camera and its related key devices with a fully sealed stainless steel seal cover, and uses vacuum aviation plugs to connect the power supply line and the signal line. The camera acquires images through the highly transparent glass in front of the stainless steel seal cover.

[0042] In the step S1, the stainless steel seal cover is suspended on the cavity stiffener, the cavity stiffener is located above the inner side of the transparent organic glass window of the vacuum aluminizing machine, and the stainless steel seal cover also includes a highly transparent glass provided thereon, which is convenient for providing a shooting window for the high-speed line-scan camera.

[0043] It further includes a high-brightness LED bar light source, which can provide uniform, stable and high-intensity light source illumination to ensure the clarity and contrast of the images captured by the high-speed line-scan camera; and the high-brightness LED bar light source is installed at the position of the original fluorescent lamp of the vacuum aluminizing machine.

[0044] Due to the extremely large amount of image processing of the high-speed line-scan camera, a large amount of heat will be generated by the FPGA and power supply module inside the camera. Currently, the camera uses heat sinks and fans to dissipate heat through air flow. However, since the inside of the aluminizing machine cavity is a vacuum environment, we cannot use air flow to dissipate heat. The semiconductor water-cooling module is attached to both sides of the camera through thermal conductive adhesive to take away the heat of the camera with water cooling. For the bar-shaped LED light source used for detection, due to the shooting by the high-speed camera, the exposure time of the camera is very limited (less than 100μm), so a high-power LED light source is required. In this way, the heat of the light source also needs to be dissipated in time in the vacuum cavity. The semiconductor water-cooling module is mounted on both the high-speed line-scan camera and the high-brightness LED bar light source through thermal conductive adhesive for heat dissipation.

[0045] The image processing equipment for mass production needs to reduce the manufacturing cost and shrink the system scale. The present invention also provides an on-line detection system for surface defects of aluminized films based on FPGA image processing, adopting an on-line detection method for surface defects of aluminized films based on FPGA image processing as described above, including: a high-speed line-scan camera, a high-brightness LED bar light source, a semiconductor water-cooling module and an FPGA-based image processing LSI; the FPGA-based image processing LSI includes a preprocessing unit, a line buffer, a processing unit and a post-processing unit connected in sequence; the preprocessing unit is responsible for operations such as density conversion, bit masking, bit inversion, etc.; the line buffer provides image data of a 5×5 or 7×7 window to the processing unit (PE), and these PE units are responsible for performing operations such as convolution and morphological processing. Other operations between pixels (such as arithmetic operations, etc.) are performed by the multifunctional main processing unit PE, and binarization, absolute value calculation and gray-scale histogram processing are performed by the post-processing unit; the processing unit is composed of 16 PE units with multiplication functions (one of which is a multifunctional main processing unit PE) and 15 gray-scale search engine units GSEU. By combining 16 PEs, 5×5 and 7×7 window filtering can be performed. The PE can be programmed through parameters, and its processing speed can reach 16MHz, which is the same as the system clock speed, and the speed always remains unchanged regardless of the change of the processing window size. An image of 1024×1024 pixels, 8 bits per pixel, can be processed in only 1.4 milliseconds. The 15 gray-scale search engine units GSEU are used to calculate the correlation value between the reference data and the search data in local parallel processing. In the present invention, the local parallel processing of template matching means that the correlation value between one reference data and multiple search data located in the neighboring regions can be calculated simultaneously.

[0046] The FPGA-based image processing LSI further includes a smoothing filter, an edge enhancement, and a rank filter circuit to implement the processing of general images.

[0047] The FPGA-based image processing LSI further includes block matching hardware based on the Sequential Similarity Detection Algorithm (SDA) to implement the gray-scale search function. This method is applicable to motion tracking, such as target tracking and optical flow measurement, because it can significantly reduce the amount of computation.

[0048] The FPGA-based image processing LSI can build a smaller system because its architecture replaces the PC with a RISC processing engine and no longer requires five independent image memories. It enables the SH bus to serve as a common data bus between two processors. Therefore, the FPGA-based image processing LSI is suitable for programmable processing because it does not need to transfer data between the SH-4 and the memory used by the FPGA-based image processing LSI.

[0049] In the design of the FPGA-based image processing LSI, since there is only one memory channel (the SH bus), the memory controller uses time-division multiplexing technology to handle multiple functions such as input image data, output result data, and image processing simultaneously. The bus width of the SH bus is 32 bits, thus compensating for the impact of reducing the memory channel.

[0050] The FPGA-based image processing LSI further includes: an SH-4 RISC processing engine, a memory controller, a Flash ROM, a RAM, an SH bus, and 49 parallel processing units; the SH-4 RISC processing engine, the memory controller, the Flash ROM, and the RAM are interconnected via the SH bus, the memory controller is connected to the preprocessing unit and the postprocessing unit, and the 49 parallel processing units are connected to the line buffer and the postprocessing unit. The control program of the device is loaded from the Flash ROM to the RAM before processing and executed by the RISC processing engine (SH-4, 60 MHz).

[0051] The FPGA-based image processing LSI further includes a PCI expansion interface and a video input / output interface. The PCI expansion interface is used to connect to the PCI bus, and the video input / output interface is used to connect to a high-speed line-scan camera and a display device, and the video input / output interface is interconnected with the memory controller.

[0052] The above description is only a description of the preferred embodiments of the present invention and does not limit the scope of the present invention in any way. Any changes and modifications made by those of ordinary skill in the art of the present invention based on the above disclosure fall within the scope of protection of the claims.

Claims

1. An on-line detection method for surface defects of aluminized film based on FPGA image processing. The aluminized film is formed in real time on-line in the vacuum chamber of a vacuum aluminizing machine. It is characterized in that It includes the following steps: Step S1: Collect the image data on the surface of the aluminized film through the line-scan CMOS sensor in the high-speed line-scan camera, and the high-speed line-scan camera is hermetically installed in the stainless-steel enclosure in the vacuum chamber; Step S2: Remove the high-frequency noise from the collected image data through Fourier transform, make the defect-free part at the 128 gray-level position through flat-field processing, and perform 5×5 filtering or 7×7 filtering processing; Step S3: Binarize the filtered image data. By setting the positive and negative thresholds of 128, the area above the positive threshold is the white defect, and the area below the negative threshold is the black defect; Step S4: Calculate the image feature quantities, and perform the labeling connection process of the image based on the binarized image to obtain the image feature quantities including the areas, lengths, widths, and gray-scale volume of the white and black defects; Step S5: Process the defects segmented by multiple frames of images, and crop the defect images of 256×256 pixels with the defect centroid as the center; Step S6: Package the defect images and the corresponding defect information, and upload them to the host computer in the form of TCP / IP; Step S7: Draw a defect map according to the image coordinate information; Step S8: Use the deep learning network to classify and recognize the defect images.

2. The on-line detection method for surface defects of aluminized film based on FPGA image processing according to claim 1, characterized in that, In the step S1, the stainless-steel enclosure is hoisted on the cavity reinforcing rib. The cavity reinforcing rib is located above the inner side of the transparent plexiglass window of the vacuum aluminizing machine, and the stainless-steel enclosure also includes a highly transparent glass provided to facilitate providing a shooting window for the high-speed line-scan camera arranged perpendicular and orthogonal to the aluminized film.

3. An on-line detection method for surface defects of aluminized film based on FPGA image processing according to claim 1, characterized in that, It also includes a high-brightness LED bar light source arranged perpendicular and orthogonal to the aluminized film, which can provide uniform, stable, and high-intensity light source illumination to ensure the clarity and contrast of the images captured by the high-speed line-scan camera; and the high-brightness LED bar light source is installed at the position of the original fluorescent lamp of the vacuum aluminizing machine.

4. The on-line detection method for surface defects of aluminized film based on FPGA image processing according to claim 3, characterized in that, Semiconductor water-cooling modules are mounted on both the high-speed line-scan camera and the high-brightness LED bar light source through thermal conductive adhesive for heat dissipation.

5. An on-line detection system for surface defects of aluminized film based on FPGA image processing, which adopts an on-line detection method for surface defects of aluminized film based on FPGA image processing according to any one of claims 1 to 4, characterized in that, It includes: A high-speed line-scan camera, a high-brightness LED bar light source, a semiconductor water-cooling module, and an FPGA-based image processing LSI; the FPGA-based image processing LSI includes a preprocessing unit, a line buffer, a processing unit, and a post-processing unit connected in sequence; the processing unit is composed of 16 PE units with multiplication functions and 15 gray-scale search engine units GSEU.

6. An on-line detection system for surface defects of aluminized films based on FPGA image processing according to claim 5, characterized in that, The FPGA-based image processing LSI also includes a smoothing filter, an edge enhancement, and a rank filter circuit to realize the processing of general images.

7. An on-line detection system for surface defects of aluminized film based on FPGA image processing according to claim 5, characterized in that, The FPGA-based image processing LSI also includes a block matching hardware based on the sequential similarity detection algorithm SDA to realize the gray-scale search function.

8. An on-line detection system for surface defects of aluminized film based on FPGA image processing according to claim 5, characterized in that, The FPGA-based image processing LSI further includes: an SH-4 RISC processing engine, a memory controller, a Flash ROM, a RAM, an SH bus, and 49 parallel processing units; the SH-4 RISC processing engine, the memory controller, the Flash ROM, and the RAM are interconnected through the SH bus, the memory controller is connected to the preprocessing unit and the postprocessing unit, and the 49 parallel processing units are connected to the line buffer and the postprocessing unit.

9. The on-line detection system for surface defects of aluminized film based on FPGA image processing according to claim 8, characterized in that, The FPGA-based image processing LSI further includes a PCI expansion interface and a video input / output interface, the PCI expansion interface is used to connect to the PCI bus, the video input / output interface is used to connect to a high-speed line scan camera and a display device, and the video input / output interface is interconnected with the memory controller.

Images