Abstract: This paper puts forward a real time extraction method of rice and wheat harvest boundary line which is suitable for shadow environment according to the problems that the existing harvest boundary extraction is easily affected by light and shadow, and the speed of extraction is slow. The Cb component in YCrCb color space is selected to carry on the subsequent image processing in order to reduce the light effect. The images are processed by filtering, thresholding, morphological operation and edge detection. The cumulative progressive probabilistic hough transform algorithm is used to identify the harvest boundary line, and the method of twice screening is used during the line extraction to improve the accuracy of rice and wheat harvest boundary detection in the shadow environment. On this basis, combined with Kalman filter and the boundary line prediction method based on projection, this paper improves the process of image processing by selecting two dynamic regions of interest, making the average tim of each image processing is reduced from 0.176 087 s to 0.064 547 s. This method provides feasible demonstration and technical support for reducing the influence of shadow environment on extracting rice and wheat boundary line and improving the speed of image processing.