Study On Recognizing Of Crop Row Centre Line With Computer Vision

The guidance signal of vision guidance system consists of the lateral deviation and the orientation deviation between the guidance directrix and the vehicle axis. The centre lines include the guidance signal for row crops. How to recognize the centre lines is one of the most important parts of the vision guidance system.A new method was studied to segment and recognize the centre lines of crop row from images using computer vision and digital image processing. In addition, the translation from the image coordinate system to the crop row coordinate system was described under monocular vision. The contents of the study could be briefly summarized as follows:1. The segmentation of crop image was studied. The Excess Green value, which could distinguish the crop with soil effectively, was used to segment crop rows from color to grayscale image. Thresholding method was used to convert grayscale to binary image. Considering the thresholding result and time, bimodal thresholding was selected to segment the crop grayscale image.2. According to the blob area distribution the improved morphology operation was put forward to remove small blob in the binary images.3. The recognition algorithm of crop rows was studied. Firstly the crop image was divided into a number of horizontal strips according to the geometric distribution of crop rows. Secondly, the recognition of crop rows was converted into achieving anchor points of crop rows. Finally, the centre lines of crop rows were regressed using the robust regression.4. The translation model between image and crop row coordinate under monocular vision was set up, ignored the height of row crop. The experiment was tested indoor according the model and the maximum error was 43.1mm when the camera height and inclined angle was 1140 mm and 54.1 degrees respectively. The result confirmed the effectiveness of the translation model.