夜间低照度条件下苹果采摘机器人的图像识别

o improve the environmental adaptability of apple picking robot and enable it to have the ability of recognizing images at night, the paper proposed a new method to recognize images acquired in artificial light source of low illumination at night. Based on all color information of pixels and clustering features in RGB color space, the method fits sampling pixels on the surface of apples to a spatial straight line by three-dimensional least square method firstly and then build a model of threshold segmentation to recognize images preliminarily according to the mean and standard deviation of the distance from sampling points to the straight line. In order to recognize shadow on the surface of apples, the same method combined with planar location information of pixels in shadow is adopted to make the recognition more accurate and more intact. Taking no account of sheltered apples, the recognition rate is over 90% in the light range. The method increases the recognition accuracy at night and work efficiency of apple harvesting robot.