Vision-Based Guidance Line Detection in Row Crop Fields

A new method of vision-based guidance line detection is developed to find guidance line in row crop fields for agricultural machinery, which uses the mathematical morphology to process images and segment crops in images and to extract target crops. The guidance line is fit with the least-square method according to the centroids of individual target crops or/and the central points of several pixel rows of crops with a continuous region in image. The guidance line could be found easily in three kinds of row crop fields, such as individual crop plants with spacing in image, plants with continuous region in image, and plants with both features above. Then experiments were implemented by an agricultural robot in a vegetable field, and results showed that the algorithm enabled the robot to drive along crop row with a high accuracy ±29.5 mm, and had an expected speed of image processing between 0.7s and 1.3s, which meets the needs of agricultural production. Therefore, the proposed algorithm has a good performance in accuracy and speed of image processing, and has robust performance for different crops.