| In the process of non-destructive quality inspecting and automatic grading of fruits with machine vision and image processing technology, the quality of fruit image captured by cameras would have great effects on the accuracy of inspecting and grading. So, it is very important to try to decrease the image distortion while capturing images. However, it is impossible to completely get rid of the image distortions, such as the distortions of grey level and color, caused by hardware of machine vision system and the projection of three-dimensional object to a two-dimensional image. These unavoidable distortions could only be reduced by correction. The methods and algorithms of image correction were studied, and the main research contents and results were as follows:1. The research advancements and achievements in the area of non-destructive quality inspecting and automatic grading of fruits with machine vision was reviewed, and the existing problems were put forwarded.2. In the light of the properties of the machine vision system adopted in this research, the calibrating method of camera was established. Distortion patterns of color were studied with standard color cards, and the color correction model for the camera in this machine vision system was set up: R = 2.903/Z'-1.124G'-140.903 , G = 0.369#'+1.846G'-0.3085'-150.777, and B = 0.297tf+0.466G'+1.2885'-151.384. Here, R', G', B' and R, G, B were the color primes before and after correction, and the correlation coefficients R2 of the model were 0.980, 0.973, and 0.981 respectively.3. The images of 57 balls, which were divided into three groups with the diameters of 75mm, 60mm, and 40mm respectively, and every group consisted of 19 balls with different colors, were captured at 6 different positions. The grey level distribution characteristics of images were analyzed. It was found that images of balls with different sizes and different colors had the same grey level distortion patterns, and the grey level iistortion would increase with the increasing of distance from the center of ball images. Additional, with the increasing of distance from the ball center to the center of camera view field, the grey level values of would decrease. At last, the grey level correction model of ball image captured at the center of camera view field was set up as follows:Here, AG was the gray level difference between the center pixel of ball image and the pixel to be corrected, and r and R were the distance from the pixel to be corrected to the center pixel of ball image and the ball radius respectively. The correlation coefficient of the model R2 was 0.796.The grey level correction model of ball image deviated from the center of camera view field was also set up as follows:AG = 0.04975d- 0.467Here, AG was the gray level difference between ball images at the center of camera view field and deviated from the center, and d was the distance from the ball center to the center of camera view field. The correlation coefficient the model R2 was 0.842.4. After analyzing the distribution characteristics of brightness, hue and saturation of ball images, it was found that the image color distortion was only related to brightness, but hue and saturation of images. The color correction model of ball images at HSV color space was established as follows:Here, AF was the distorted value of brightness of color. The correlation coefficient model R2 was 0.846.5. Based on the grey level and color correction models established, the method for correcting the grey level and color of orange images was researched, and the corresponding correction algorithm was proposed and programmed with MATLAB programmers.6. The grey level correction model and the color correction model were validated by tests. These results laid a solid foundation for the development of quality inspecting and grading machine of agricultural products.
|
PDF Full Text Request