Self-adaptive harvesting method and device of unmanned harvester based on vision

The invention discloses a vision-based self-adaptive harvesting method and device for an unmanned harvester, and the method comprises the steps: 1, selecting a working region, a starting point and an ending point according to a to-be-harvested farmland, determining a global planning algorithm through a selected route mode, and calculating an initial global path; step 2, after the harvester runs to a starting point, determining a single ridge and an inter-ridge channel according to an image of a to-be-harvested farmland collected by the harvester at the starting point, finding a ridge closest to the position of the harvester, determining the closest ridge as a to-be-harvested ridge if the closest ridge is identified to have crops, and if the closest ridge is identified to have crops, determining the closest ridge as a to-be-harvested ridge; taking a field ridge channel adjacent to the nearest field ridge as a target field ridge channel, and entering the step 3; 3, the harvester is controlled to run to a turnin