Method for establishing fruit tree canopy feature map in orchard

Abstract

The invention discloses a method for establishing a fruit tree canopy feature map in an orchard. The method comprises: using an RGB-D camera for collecting color information and depth information of surrounding environment images in the moving process of a robot in real time, using a semantic segmentation network for segmenting fruit trees and other environment features, reserving information of the fruit trees in the depth images according to the segmented images, and therefore point cloud data of the fruit trees are generated; extracting characteristic angular points in the color image, calculating the motion of the camera of two adjacent frames of images, and fusing the motion with the IMU data so as to estimate the motion state of the camera; constructing an optimization problem for the multi-frame images and the camera pose estimation by adopting a factor graph, and optimizing the pose of the camera; and constructing an octree map of the fruit tree canopy according to the fruit tree point cloud data and the camera pose. Rich environmental information of an orchard is collected, the understanding of the robot to the orchard environment is improved, and the requirement of the robot for completing higher-level operation such as automatic navigation, variable spraying and fruit picking is met.

Description

technical field [0001] The invention relates to visual SLAM technology and semantic segmentation technology, and in particular provides a method for establishing a canopy feature map of fruit trees in an orchard. Background technique [0002] From soil analysis, to orchard management, to climate detection, and fruit picking, a variety of precision agricultural technologies are being applied to all aspects of orchard production. Precision agriculture is becoming the future of modern orchard development. Orchard autonomous operation robots are gradually replacing humans to complete the complicated labor in the orchard. One of the key technologies of autonomous operation robots is environmental perception technology, which creates digital maps based on collected data to guide robots to move autonomously. [0003] SLAM (Synchronous Localization and Mapping) is considered to be the key to realize the real autonomous movement of the robot. Most of the traditional orchard SLAM meth...

AI Technical Summary

Problems solved by technology

However, visual SLAM based on traditional machine vision methods cannot adapt to changes in the color and growth of fruit tree leaves in orchards.

[0005] When the camera moves too fast or the visual recognition is wrong, using the visual sensor alone to estimate the robot's motion will cause large errors, which cannot ensure the reliability of the orchard robot.

Images