Abstract:Traditional vision detection methods generally have better image segmentation effect when image foreground and background have obvious chromaticity difference. However, for the maize stubble field harvested by combine harvester, there are other backgrounds besides maize stubble row such as maize residues, naked land surface, and their color are very similar. Therefore, traditional image processing methods are not suitable for the segmentation of maize stubble row. In order to achieve precise and rapid segmentation of maize stubble row, a segmentation method for maize stubble row based on hyperspectral imaging technology was put forward. Firstly, the original hyperspectral image of maize stubble row was corrected by black and white correction algorithm. Then, the principal components analysis algorithm (PCA) was used to analyze the hyperspectral image. And the feature wavelengths (1260nm, 1658nm and 2131nm), which could maximum highlight the stubble tip incision and lighten the backgrounds, was selected according to the weight coefficient curve. In addition, the PCA algorithm was widely used in hyperspectral image analysis because of its effective dimension reduction effect and convenience. Secondly, images at three wavelengths were extracted and analyzed by PCA. After that, the PC2 image was convert into binarization image via single threshold method. Thirdly, the median filtering algorithm, morphological open operation, and edge noise removing algorithm were applied to ensure the precision and integrity of the maize stubble row. Totally 50 test images were collected to verify the segmentation effect of the presented method. At the same time, the segmentation precision rate, recall rate, and F1 value were calculated. The results revealed that the method proposed had good segmentation effect, and the segmentation precision rate, recall rate, and F1 value were 91.85%, 90.49%, and 91.16%, respectively. Therefore, the developed method realized good performance in maize stubble row segmentation and can provide great help for detection of navigation line in maize stubble cropland harvested by combine harvester.