Quality classification of peanuts based on image processing

Abstract

To ensure the good quality of peanuts, the system of quality classification is established on the basis of computer image processing, thus realizing the identification of damaged and mildewed peanuts as well as the grading of the shape and size. The colour information R, G, B and texture information correlation are taken as the feature parameters to distinguish the mildewed, stale and normal peanuts. The colour features R, G, B of damaged area are extracted to recognize the damaged peanuts based on pattern matching. Fourier Transform and Inverse Fourier Transform are adopted to describe the shape of peanuts. Then the thirteen harmonics of Fourier descriptor are taken as the feature data to achieve the peanut shape classification as normal, triangular, elliptic and circular. Such geometric feature parameters as peanut area and circumference, etc. are applied to recognize peanuts of different size. The results show that this method achieves an accuracy of 93.33% for mildewed peanut, 80.12% for damaged peanut, and also an average accuracy over 86.67% for shape and over 90% for size.