Potato Shape Detection Based On Machine Vision

Potato domain is lord industry in Inner Mongolia country region.The quality examination of the potato and the automation of processing techniques, fine turn is the important way that raises its grade and wealth.However, the potato shape is one of the important factors of the potato quality examines, the task of manual graduates is heavy while the efficiency is lower.Filter system of excellent shape potato based on machine vision was studied in this paper,and the potato shape’s automatic detecting system was established though two potato shape screening technique based on Fourier transformation.1.There were some color potato pictures which are classified as circular, oval and abnormal,and each was180. These pictures were grayscale processed, filtered, thresheld in segmentation and edge detected via functions in MATLAB2008a software IPT.2.Filter method of excellent shape potato based on machine vision was studied systematically in this paper. On the basis of Fourier transformation, this paper created potato’s shape examining evaluation system.(1) Proposed the potato shape examine method based on Fourier descriptors of boundary points Moment Features, Obtained boundary points on the edge of the potato’s image, and calculated value of the moment features, then got the information of the potato shape feature. Though discrete Fourier transformation, the Fourier descriptor which has the features of the translation, rotation and scale invariance was got. The round, oval and deformity potatoes were classified via Euclidean distance classification. Detection accuracy were90%,91.75%and98.75%respectively when the training sample was80pictures. Detection accuracy were92.5%,93.75%and100%respectively when the training sample was100pictures.(2) This paper quoted Optical correlation transform theory, created non-linear processing methods of joint power spectrum, and stipulated sidelobe peak which formed in the related joint transform indicates potato similarity. Then the test results of the potato shape can be gotten vividly. Detection accuracy were87.5%,85%and92.5%respectively when the training sample was80pictures. Detection accuracy were88.75%,91.25%and95%respectively when the training sample was100pictures.3.This paper contrasted experiment result of above two kinds of potato shape examination method, and analysised their accuracy, velocity, ease and practical innovation, elaborated merit and shortcoming of each one respectively,which provided basis for accurate graduation of potato shape and the theory research of agricultural products quality examination though the machine vision.