Solar panel segmentation under low contrast condition

How to accurately segment a solar panel is an intractable problem when the infrared images are captured under a low contrast condition. In this paper, an effective segmentation method is proposed by combining low-light enhancement of dark images, bias correction of level set segmentation, and local entropy segmentation. First, the dark images are enhanced via the histogram equalization and V-channel of HSV(Hue, Saturation, Value) color space. Secondly, the bias field of enhanced image is corrected by the level set method to improve the intensity of the image. Then, the image local entropy is derived to obtain the outline profile of the solar panel. Finally, the solar panel is fitted with the large plate to obtain the final segmentation result. The experimental results demonstrate the effectiveness and feasibility of the proposed method.