Detecting and Tracking Hair Impurities in Mushroom Semi-product Images

A new area operator was proposed to extract axes feature of hair impurity in mushroom semiproduct images. This operator was based on the local intensity comparison, and it had two advantages: 1) its outputs reached local maxima at the hair axes; 2) the accurate direction information of the hair axes could be derived by it. After the axes feature of the hairs was extracted using the proposed operator, an extended Kalman filter was applied to track the hairs. The searching path was established when a cost function was minimized. Starting points of the tracking path were found using dyadic wavelet analysis. The proposed algorithm had good adaptivity for the arbitrary orientation of the hairs. The experiment results showed that hairs could be tracked accurately.