Detection of rice seedlings rows' centerlines based on color model and nearest neighbor clustering algorithm

Centerline detection of rice seedling rows in paddy fields plays an important role in autonomous navigation of weeding robots. Since the paddy field environment varies from the different growth stages of rice seedling and the color of rice seedling is similar to cyanophytes and duckweed contained in the paddy field image in South China, the segmentation of rice seedlings and centerlines detection become very challenging. This paper presents a method to detect rice seedlings rows' centerlines based on color model and nearest neighbor clustering algorithm. The features of rice seedlings are firstly extracted based on 2G-R-B color model and S component in HSI space. The feature points of rice seedlings are then extracted without changing their shapes, and clustered based on nearest neighbor clustering algorithm according to their adjacent relationship. Finally, the centerlines can be correctly detected by applying a known point Hough transform in each cluster. The experimental results show that the feature of rice seedlings can be precisely extracted under a background with cyanophytes and duckweed, and the average processing time of a 1280960 resolution color image is less than 350 ms. In addition, the algorithms are adapt to the environmental variation and meet the real-time requirement for agricultural robots.