Bayberry maturity estimation algorithm based on multi-feature fusion

Abstract

The rapid development of smart orchards is conducive to scientific planting and management, the estimation of fruit maturity is the key to harvest in orchards. Nowadays, research on the maturity of bayberry is almost nothing, in order to quickly and accurately estimate the maturity of bayberry in orchards, a bayberry maturity estimation algorithm is proposed based on multi-feature fusion by machine vision. Firstly, considering the local and global texture characteristics of bayberry appearance, bayberry image of texture features were extracted based on GLCM and LBP. Simultaneously the algorithm extracted R, G, B, H and S components based on RGB and HSV color space, the color components were transformed by histogram to obtain the color features of bayberry. Then the color and texture features were fused in series to accurately describe the surface features of bayberry with different maturity. Finally, an SVM-based bayberry maturity estimation model was constructed, the linear kernel function was selected to estimate bayberry maturity based on the sample features. Through experimental verification, the algorithm takes into account the accuracy and real-time performance, the average accuracy rate on the test set reaches 91.2%, and the reasoning time is only 5 ms, which has high practical value.