Machine Vision for a Micro Weeding Robot in a Paddy Field

The study developed an image processing method to determine the travel directrix for a micro weeding robot in a paddy field. The processing samples, which included 24 images in each sample, were selected from the videotapes which were obtained on a weekly basis, when the camera manually revolved 360° eight times at each height and angle position in the space between the rice rows from the third day after transplanting to 7 weeks. The target image was abstracted from the 24 images by analysing the distribution of blue pixels in the colour image and the distribution of black pixels in its binary image. The candidate points for the directrix line in the target image were obtained by analysing each horizontal line profile of its binary image or the blue image according to the level of growth in the paddy field. The travel directrix line was detected by passing a known-point Hough transform. The directrix lines of all 69 samples used in the study were correctly detected. The processing times required to draw the directrix line after reading the 24 images were about 0·4–0·8 s.