Crop row recognition method and device

Abstract

An embodiment of the present invention provides a crop row recognition method and device, the method comprising: converting the crop row perspective view into a crop row bird's-eye view through an inverse perspective transformation matrix, and the crop row perspective view is taken according to a camera at an elevation angle to the road surface Obtain; select the first region and the second region in the bird's-eye view of the crop row, obtain the crop row edge point of the first vertical projection image of the first region and the crop row edge point of the second vertical projection image of the second region; obtain the second The midpoint of the left and right edge points of each crop row in a vertical projection image is the first midpoint, and the midpoint of the left and right edge points of each crop row in the second vertical projection image is the second midpoint; The midpoint and the second midpoint are connected into a straight line to obtain the crop row in the bird's-eye view of the crop row. The crop row recognition method and device provided by the embodiments of the present invention use inverse perspective transformation to eliminate image geometric distortion, introduce structural constraints on crop rows, and improve the robustness of the method.

Description

technical field [0001] Embodiments of the present invention relate to the field of machine vision, and in particular to a crop row recognition method and device. Background technique [0002] In recent decades, agricultural environmental issues have received more and more attention. Excessive application of chemical fertilizers and pesticides has adverse effects on the environment and increases input. If chemical fertilizers and pesticides are applied to specific positions in crop rows, the use of chemical fertilizers and pesticides can be reduced. In order to achieve the precise positioning and application of chemical fertilizers and pesticides and the operation requirements of not damaging seedlings, it is particularly important to identify crop rows. In actual production and planting, many crops are usually planted neatly in a straight line and parallel to each other. Therefore, in precision agriculture, a research hotspot is to identify crop rows from graphics, detect c...

AI Technical Summary

Problems solved by technology

The existing crop row recognition methods are all recognized in perspective images, and cannot use the parallelism of crop rows to add constraints

The camera installed on the rack for cultivating and weeding is tilted downward, and its optical axis is not vertical to the ground. The crop rows captured will have certain geometric distortion in the picture, and the relative relationship between the cultivating and weeding machine and the crop rows cannot be judged only by the captured images. geometric relationship

[0004] The existing method of detecting pixel jumps directly on the perspective binarized image of crop rows is prone to misjudgment due to noise points, weeds, and the shape of the crop itself, which is not conducive to the segmentation of different crop rows

Images