Identification and location of corn seedling based on computer vision

A prerequisite for variable spray operations is to detect, identify and locate maize seedlings in the complex scenes of farmland rapidly and accurately. In this paper, Regular pattern in distribution and growth of the corn seedling by bunch planting but leaves of plant overlapped in field is carefully taken into account, with the image acquisition mode which the camera image plane parallelized to the ground, in order to obtain the center of maize seedling roots, over-segmentation block projection method based on position feature is employed to locate crown of maize seedlings with the image captured. So the coordinate of corn root can be indirectly obtained by means of principles of perspective in machine vision as well as growing characteristics of corn. Experiment shows that the positioning error by this method is less than 30mm, and the average time in processing an image with 640 × 512 pixels in MATLAB is less than 300ms.