Method for fruit shape feature acquisition based on multidirectional vision

In view of the drawbacks of few fruit feature information in traditional fruit grade identification, which relies on acquiring one fruit image, an algorithm to realize fruit shape feature based on multidirectional vision is presented in this paper. This system can realize synchronic acquisition of multidirectional images of fruit surface, and it can give the features of multidirectional fruit images after image processing, which include image segmentation based on OHTA space and denoising based on improved BLOB aglorithm. A linear classfier is proposed according to statistics theory. A judging rule was given after complementary fusion on multidirectional image features. The performance of this algorithm was tested with the developed intelligent fruit sorting experimental prototype, and the experimental results show that it is effective and the success rate is up to 97 percent, which can satisfy the requirements of real time fruit sorting system.