Design and application of image processing technology in part surface damage detection

The problem of surface damage detection

Accurate and rapid detection of surface defects of parts is directly related to product quality. If unqualified products are not removed in time, it will bring quality risks. However, in industries such as automobiles, motorcycles, and internal combustion engines characterized by mass production, the identification and detection of surface defects in key parts of important parts are still mainly based on manual visual inspection. Considering the complexity of the process execution (especially after the use of advanced connecting rod breaking process with separated reducers), standardized evaluation standards must also be proposed. For example, there are specific regulations for possible damage on the connecting rod reducer joint surface: the crack area is less than 3mm2; the linear length of the crack in any direction is less than 2.5mm. As long as one of the conditions is met, it will be judged as unqualified and removed.

According to the characteristics of the parts, the area where the crack may appear is outside the joint surface (line), and its range is in the shape of an "eight". In this case, relying on manual visual inspection and estimation is not only inefficient and labor-intensive, but also unable to accurately implement the provisions of the above standards. On the other hand, even if other conventional measurement methods are used, it is difficult to achieve the above purpose.

The principle of using image processing technology for surface defect detection

Image processing technology, also known as "machine vision", uses the image of the object to be measured as a carrier of information and extracts useful information from it to achieve the purpose of measurement. It has the advantages of non-contact, high speed, large measurement range, and rich information. Different detection requirements can be achieved through the combination of CCD (Charge Coupled Device) camera, optical system, and processing system. The reflection method shown in Figure 1 can be adopted for the identification of the above-mentioned surface defects of the workpiece.

As shown in the figure, this system illuminates the crack area to be detected through a square LED diffuse reflection light source. After the light is irradiated on the surface of the object, it is reflected on the photoelectric coupling CCD component in the camera and converted into a corresponding electrical signal. The CCD component can be understood as a dot matrix composed of photosensitive pixels. Each pixel of the CCD corresponds to the two-dimensional image characteristics of the object to be measured. That is, the image characteristics of the object can be indirectly analyzed by analyzing the "pixel imaging result". For example, by calculating the number of imaging pixels in the binary image, the length value and area value of the corresponding object can be obtained. After the image processing system performs binarization processing on the obtained image according to the electrical quantity signal, the binary image is taken as the object for further calculation and analysis.

In actual use, the image processing system adopts the comparison method to realize the setting of the grayscale binarization threshold and the light source. The specific method of comparison is: using a known sample as a reference for calibration (comparison), dividing the known reference measurement value by the pixel value corresponding to the reference, and the corresponding ratio value between the pixel and the actual value can be obtained. By adjusting the brightness of the light source and the binarization threshold of the system, the grayscale binarization threshold is optimized to ensure that the system has a relatively high resolution for the object boundary, that is, the optimized binarization threshold and light source can make the change of the boundary produce the largest possible pixel value change. As

a novel and practical sensing technology, the image detection unit has been commercialized in recent years. Some well-known manufacturers, such as Panasonic in Japan and Siemens in Germany, have launched a series of products with complete varieties and specifications, including light sources, cameras, image processors, etc., which has created very favorable conditions for the promotion and application of image detection technology. At the same time, the relevant enterprise standards promulgated not only standardize production, but also provide a basis for users to select appropriate detection units in different situations and design systems faster and better.

According to the characteristics of the object to be measured (workpiece, measured part), referring to the relevant standards, it will be easy to select the appropriate image detection unit. Taking the connecting rod as an example, since the defect area of ​​the joint surface cannot be greater than 15×15 mm2, it is more appropriate to take the "field of view" of 20×21.4 mm2 from the corresponding standard. Relative to the field of view and depth of field of each level, users can choose cameras with different focal lengths, such as 8, 16, 25 and 50. Each focal length corresponds to two parameters, such as the distance al from the characteristic lens to the measured surface and the distance ba from the characteristic lens to the CCD photosensitive surface. According to the situation of the workpiece to be measured, a camera with a focal length of f=25mm is selected. At this time, the above two parameters are 137mm and 9mm respectively. This example uses a small image detection unit from Panasonic, in which the core component CCD photosensitive film has a pixel of 512×480. When the field of view is determined, the measurement resolution of the selected detection unit can be calculated based on this:

X-axis resolution:
21.4/512=0.0417mm

Y-axis resolution:
20.0/480=0.0417mm

Area resolution:
0.0417×0.0417=0.00174mm2

As shown in Figure 1, the reflective image measurement has two forms: light source combined in the camera and split arrangement. The light source itself has fluorescent lamps, halogen lamps, lasers, LED light sources, etc. According to the specific situation of this example, a split-arrangement LED light source solution is adopted, which is easy to adjust.

Figure 1 Reflection image measurement principle

Composition, design features and operation process of special inspection equipment

System composition

The formation of the measurement system plan is based on the characteristics of the object to be measured. As mentioned above, the distribution range of the cracks and defects on the joint surface is in the shape of an "eight", which means that in order to complete an inspection, it is necessary to measure in three directions; on the other hand, the process and production departments have put forward the requirement of implementing full inspection. Therefore, it was decided to adopt a semi-automatic solution, that is, except for the manual loading and unloading of the workpiece, the entire measurement process is automatic to adapt to the faster work rhythm.

Figure 2a

Figure 2b

Figure 2a, the measuring system is mainly composed of a camera, an LED light source, an image processing unit, a programmable controller (PLC), a display and a mechanical part, among which the camera is used for image acquisition; the LED light source provides a stable and long-lasting light source to ensure the image quality; the programmable controller (PLC) controls the function execution of the measuring system; the image processing unit processes and analyzes the data and provides signal output; the display shows the image acquisition situation and the data analysis results; the mechanical part supports and implements the actions during the measurement process.

Operation process and characteristics of the measuring system

The detection device is a desktop instrument with a very compact structure. As can be seen from the two schematic diagrams in Figure 3, the camera 3 is fixed at one end of the swing arm 4, and the other end is supported on the bracket 7 through a pivot. The stepper motor 5 installed on one side of the frame 9 can drive the swing arm 4 to rotate with the help of a synchronous toothed belt 6 and a synchronous wheel on the pivot, and the rotation range is ±150, and the two proximity sensors 8 arranged on both sides of the frame assist in positioning. Another pair of photoelectric sensors are placed on both sides of the entrance of the workpiece 1 to ensure the accurate positioning of the workpiece on the fixture before measurement and to turn on the light source.

Figure 3 Structural diagram of special testing equipment

Example of measuring the crack defect on the connecting rod joint surface

The camera, that is, the original state of the swing arm is in the right position (A in Figure 2b). Driven by the swing arm drive mechanism, the camera sequentially samples from the right, middle and left positions with an angle of 150 degrees to each other, that is, ABC. At the same time, the image processing unit in the block diagram of Figure 2a transmits the numerical results of the camera sampling three times in a row at each position to the programmable controller PLC for comparison through the RS232 interface. If the above three consecutive measurement results are the same, this value is confirmed as a reliable value, and then stored in the stack of the PLC for the final comparison of the three orientation detection results and finding the maximum value.

If the camera's detection results at each position are different for three consecutive times, it is necessary to perform another three consecutive samplings and compare the results. If no reliable value can be obtained after five repeated cycles (15 samplings), the entire detection system will automatically reset, and the swing arm drive mechanism and the camera installed on it will return to the initial position on the right. At the same time, the device sends a "system failure" signal.

As a surface defect measurement, the above detection system has the characteristics of combining intuitive display of images with analysis and judgment of measurement results. For the area detection of the crack defect on the connecting rod joint surface, the crack image is separated by the optimization algorithm in image processing, wherein the image of the crack defect part is binarized into black, and the image of the other part is white. Then, by statistical calculation and unit conversion of black pixels, the area size of the crack defect part can be obtained, and then a qualified judgment can be made. As for the detection of the maximum linear length of the crack, the "boundary scanning method" is used to scan along the X and Y directions to find the boundary of the crack, and then the maximum linear length of the crack defect part is determined based on the calculation of the diagonal line of the rectangle surrounded by two sets of parallel lines, that is, the rectangle enveloping the crack boundary. After

each detection is completed, the measurement results presented on the display include: area CA01, X-direction length CA02, Y-direction length CA03 and maximum linear length CA04. Then, according to the evaluation indicators set in advance, the state of the crack defect on the joint surface is judged. To facilitate operators in mass production, there is a green (qualified) indicator light or a red (unqualified) indicator light on the upper part of the front of the machine body, which indicates the status of the inspected workpiece in a simpler way.