According to the gray-scale step change of the edge of the piston end image, the edge of the piston end face is detected by image processing, and the boundary is extracted to determine the position and size of the boundary feature. The detection of the image boundary circle includes round edge detection and center position detection. The edge error measurement is based on the center of the circle. First, the boundary function is fitted, and then the pixel function is compared with the standard function to obtain the circular position deviation and the circularity deviation. The determination of the center position is combined with the characteristics of the circle itself, and the least squares method is adopted. A circle fit is performed on the obtained coordinates to obtain a function of the center coordinates and the circle, thereby obtaining the position and size of the circle.
2 detection method is achieved.
The detection of the roundness error of the piston is based on the ideal digital image of the piston end face. In the image acquisition process, the image of the piston end face directly collected cannot meet the detection requirements due to the influence of illumination, equipment characteristics and various noises. . In order to achieve high-precision error detection, the image must be preprocessed to obtain a high-quality image, and then the boundary extraction operation is performed on the image. The detection system uses the composite Canny operator for boundary extraction to provide basic data for roundness analysis. The method of image preprocessing is described below.
2.1 Image enhancement.
In the process of acquisition, transmission and storage, images are often subject to various noise pollution and distortion, which seriously affects the quality of the image. Therefore, before the image is analyzed, the image should be improved in quality. The main purpose of Image Enhancement is to enhance the useful information in the image, to reduce the influence of noise and other interferences, to make the image clearer and more convenient for people or computers to analyze the image. Image enhancement and image restoration are different. It does not require the restoration of the original image information of the image, nor the specific reason for the image degradation. It only uses the experimental adjustment method to process the image to improve the image quality. Image enhancement techniques mainly include image grayscale transformation, image smoothing processing, and the like. In practical applications, it can be processed by a single method, or it can be combined in several ways to achieve the desired enhancement effect.
2.2 Image grayscale adjustment.
Some images have low contrast and the entire image looks blurry. At this time, the gray value of each pixel of the original image can be modified according to certain rules, thereby changing the dynamic range of the image grayscale. Let the original image be f(x, y) and its grayscale range is
The contrast enhancement effect of the piston end face image is realized by Matlab. In the figure, the vertical and horizontal coordinates are pixel coordinates, and (0, 0) is in the upper left corner of the image.
It can be seen from the experimental results that the results of image processing are often different when different gray levels are set. The smaller the gray scale range, the higher the image contrast and the clearer the boundary, but the greater the damage to the image background. Therefore, it is critical to select an appropriate grayscale range during actual image processing.
2.3 Image smoothing.
Digital images may have various noise pollution. Noise pollution can occur both in the image data transmission process and in the quantization process. In the image smoothing stage, an ideal smoothing method needs to be found. This method is required to eliminate image noise without blurring the edge contours and lines of the image.
In order to eliminate noise and maintain the details of the image, median filtering is an ideal processing method. The principle is: the template is exhausted in the image, and the center of the template is respectively coincident with a pixel position in the image, and the gray values ​​of the corresponding pixels under the template are sorted and found in the middle, and the middle The value will be the gray value of the pixel in the new image corresponding to the center of the template.
The median filtering is essentially to change the pixel having a larger difference from the gray value of the surrounding pixels to a gray value closer to the gray value of the surrounding pixels, thereby eliminating the isolated noise point. Since it is not a simple averaging, it is not easy to produce edge blur. The results of the median filtering are compared and the boundary information is further enhanced. Experiments show that the image processing of the piston end face adopts the median filtering method, which can obtain the ideal processing effect.
3 boundary fitting and error detection.
The ultimate goal of the circle fitting is to fit the end face of the piston out of the boundary circle, and obtain the characteristic parameters such as the center coordinate and the radius value of the fitted circle, so as to detect the machining precision of the piston part.
After pre-processing the acquired image of the piston end face, boundary extraction can be performed. In order to analyze the machining error, the boundary function must be obtained for the extracted boundary, and then the fitting error is compared to determine the magnitude and position of the error.
3.1 Mathematical tools for boundary circle fitting.
When performing boundary arc detection, you need to determine the function and center of the boundary circle.
3.2 Improvement of the round fitting algorithm.
In addition to achieving a certain degree of fitting accuracy, the round fitting algorithm must also meet the requirements of stability, efficiency, and real-time. In terms of circle fitting, a variety of algorithms have been proposed, first of all, the traditional statistical method, which is based on the least squares method, seeking the best fit point set, which is usually a nonlinear problem. Iterative method is needed to solve the problem. When the initial value is not suitable, it may cause the algorithm to fail. Secondly, the Hough transform method is used to obtain the characteristic parameters such as the center coordinate and the radius of the circular hole or the arc. The advantage is that the parameters of the circle can be directly obtained. The disadvantage is that the amount of calculation is relatively large, and the operation speed is relatively slow, which cannot meet the real-time requirements of online detection.
According to the above analysis, based on the characteristics of the above two algorithms, combined with the nature of the circle, an improved algorithm for round fitting is proposed to improve the accuracy and effectiveness of the fitting algorithm. The implementation process is as follows: (1) Let E be a circle, and now scan E from top to bottom. Assuming that each horizontal scan line intersects E at 2 points and then finds the midpoint of the distance between 2 points, the midpoints of all horizontal line segments are obtained. These points are discrete points, which are linearly approximated by least squares method. Combine and find the fitted line.
(2) In the same way, scan in the vertical direction, find the midpoint separately, and perform straight line fitting.
(3) Find the intersection of the straight lines 1 and 2, which is the coordinates of the center of the boundary circle. When scanning horizontally and vertically, it is possible that there are no boundary pixels on some horizontal lines, and such a situation can be discarded.
(4) When the center of the circle is obtained, there will inevitably be some noise in the boundary image, which will greatly interfere with the scanning of (1) and (2) steps, and have a great influence on the accuracy of the scanning. The method is to improve the accuracy of the pixel points. To this end, you can set the center scan constraint, that is, after calculating the coordinates of the center of the circle, calculate the distance from each pixel to the center of the circle, and identify those unreasonable points and remove them. Let ri=|r1-ri| (r1 is the average radius of the boundary pixel, ri is the distance from the boundary pixel to the center of the circle), if the deviation ri>n (n is an integer), then the boundary pixel is discarded, called no The sample point that matches the scan value, and conversely, the sample point that is the scan value.
(5) After the noise is removed, the (1)-(3) step scan and the fitting operation are repeated to obtain a new center coordinate with high precision.
This method can extract the center coordinates of the circle quickly for images with small circle and ellipticity. For the other parameters of the circle, the least squares method can be used to obtain the circle. The method can improve the fitting speed and accuracy of the circle. The accuracy and real-time requirements of the piston roundness digital detection system.
An example of image processing is performed as described above. After the first round of the center of the circle, in order to filter out the noise and improve the accuracy of the pixel point of the center, set the center-of-center scanning constraint deviation ri>2, that is, take n=2, and discard the pixel with the deviation greater than 2 pixels. . As can be seen from the direct comparison between (c) and (d), the center point of the circle is adjusted from one position (356, 264) to the second position (358, 266), and the accuracy of the two pixels is improved in the vertical and horizontal directions. It shows that the method has good practicability, especially in the strong noise image.
For the digital image of the piston end face, the steps for fitting the boundary circle are as follows:
(1) The boundary contour of the circle is separated by processing the image.
(2) Using the improved boundary circle center method, find the center coordinates, and use the least squares method to fit the boundary pixels to find the independent parameters of the circle. Circle fitting the digital image of the piston end face.
3.3 Roundness error analysis.
After theoretical analysis and algorithm programming, the coordinate value of the center of the image of the end face of the piston is extracted quickly, and the other parameters of the boundary circle are processed by the least squares method. The detection of the boundary error of the piston end face can be evaluated according to the roundness error evaluation method. The center of the circle is translated to the origin of the pixel coordinate system, and the difference between the points on the circle corresponding to the pixel point of the boundary image of the piston end face can be obtained.
The center of the circle is the least squares, and the Cartesian coordinate system is established with the horizontal and vertical directions as the x-axis and the y-axis, respectively.
4 conclusion.
Using digital image detection method, using computer image processing technology, combined with MATLAB's powerful mathematical operation and image processing ability, it can detect the pixel-level error of the circularity of the piston end face, which can reduce the labor intensity and improve the accuracy under the premise of ensuring the detection accuracy. The level of error detection automation, which has a positive effect on mass production. The digital detection system also has certain practical significance for improving the technical level of the air compressor piston manufacturing industry.
The vacuum conveyor is comprised of vacuum (no oil,no water),feeding tube,PE filter,vacuum hopper ,automatic control device,pneumatic discharging device and compress air device,etc.
vacuum conveyor for powder,food powder conveyor,powder conveyor,powder conveyor machine
Jiangsu Hongda Powder Equipment Co., Ltd , https://www.hdpowderexpert.com