Fruit surface defect detection method based on visual saliency

Abstract

The invention discloses a fruit surface defect detection method based on visual saliency. The method comprises the following steps: acquiring an original image of a fruit to be tested through a charge coupled device (CCD) camera; preprocessing the image so as to eliminate small texture details and noise and encoding distortion; performing downsampling on the preprocessed image to obtain two pictures, performing upsampling to obtain two pictures, and forming five feature source images of different scales with the original image; calculating the five saliency maps under different scales, and performing spatial domain enhancement on the five maps; finally, fusing the saliency map models under the five scales, and segmenting the fruit surface defects according to the fused saliency map model. The interference brought by the texture and color of the surface of the fruit can be eliminated, the surface defect detection of the fruit is realized, and the problem that the harvested fruit cannot be accurately graded can be well solved, so that the automation degree of the fruit industry is improved, and the manual operation is reduced.

Description

technical field [0001] The invention relates to a method for realizing fruit non-destructive detection by using digital image processing technology, in particular to a method for detecting fruit surface defects based on visual salience. Background technique [0002] Fruit surface defects are the main discriminant factors of fruit grading standards, and are crucial to fruit post-harvest processing and commercial production. The traditional manual picking and hole mechanical screening and grading methods usually take up a lot of manpower, and the grading speed is slow, subjectivity varies greatly, and it is easy to cause mechanical damage. Now more research is focusing on using high-precision cameras to obtain fruit images. Using Machine vision method to judge fruit quality. [0003] According to the low gray value of the defect center, there are some apple defect detection methods based on gray images, such as using the "flood method" and its improved algorithm "snake method...

AI Technical Summary

Problems solved by technology

According to the characteristics of fruit surface defects, the above methods use pure classical image processing methods for defect segmentation. Compared with direct threshold segmentation, the effect of the above methods is greatly improved, but under the interference of fruit surface texture, color, etc., the effect is often not good, and Large amount of calculation

Images