Navel Orange Recognition Method Based on Improved Otsu Algorithm

An improved Otsu threshold segmentation algorithm with filling a circle method based on the center of mass(FCCM) is proposed to solve the problems of missing regional feature information and unsatisfactory effect in navel orange recognition by traditional image processing method. We first choose the appropriate color model. By analyzing the data, YCbCr model has better contrast compared with other color models, and the histogram of its Cr component has obvious peaks and troughs, which is conducive to image segmentation. We then segmented the extracted Cr component images by Otsu threshold segmentation algorithm. And further processing is carried out in combination with morphology and hole filling algorithm to make the image coherent, so as to ensure that the feature information of navel orange is not easy to be missing. And next, the center of mass of the image is determined and the radius of the detection circle is calculated by stepping method. Finally, the noise points outside the radius of the detection circle are eliminated and the recognition results are displayed in the original image. The experimental results show that the proposed method is effective in the recognition of low pixel images with both different light conditions and obscured condition.