The visual identification technology of mature citrus fruits in the condition of disturbance had been studied. Firstly, analyzing disturbed citrus in ANSYS, so as to confirm the law of motion was the simple pendulum movement. Secondly, processing video image of the disturbed citrus,by using the improved K-means clustering method combined with Hough circle transform method to segment fruits. According to the image segmentation result, fitting the binary citrus image using Hough line method. With constraint of straight slope between -0.45~0.45, disturbed citrus’picking positions were determined ultimately. Test results show that this method could identify the disturbed citrus effectively in the natural environment, and the recognition rate of picking position was 85%, with processing time within 7.58s to 9.24s. The study provides the theoretical foundation and technical support for the visual localization of disturbed fruits.