Fruit and vegetable growth form visual identification method in natural environment

Abstract

The invention discloses a fruit and vegetable growth form visual identification method in a natural environment. The method comprises an image acquisition step of acquiring the fruit and vegetable images in real time based on monocular vision; a target area extraction step of processing the acquired image, and extracting a fruit and vegetable area with shielding branches and leaves in the image, wherein the method for processing the acquired image comprises the following steps of detecting and semantically segmenting a fruit and vegetable region in a fruit and vegetable image by adopting a convolutional neural network of a mask based on a region; an overlapping form identification step of dividing the fruits and vegetables into single-fruit-form fruits and vegetables and multi-fruit-form overlapping fruits and vegetables; and a branch and stem shielding form identification step of further subdividing the single-fruit-form fruits and vegetables and the multi-fruit-form fruits and vegetables into non-branch-stem shielding single fruits and vegetables, branch and stem shielding single fruits and vegetables and non-branch-stem shielding overlapped fruits and vegetables, and branch andstem shielding overlapped fruits and vegetables. According to the method, the picking robot can automatically classify the fruits and vegetables with different growth forms in the collected image, anda basis is provided for the subsequent selection of a corresponding picking mechanism.

Description

technical field [0001] The invention relates to the field of fruit and vegetable shape pattern recognition for a picking robot, in particular to the visual recognition of the growth shape of fruits and vegetables. Background technique [0002] The growth forms of fruits and vegetables in the natural environment are various, and the picking mechanism of fruits and vegetables with different growth forms should be different for picking robots. Therefore, in the operation of fruit and vegetable picking robots, the primary task of its work is to be able to visually identify different growth forms. Shaped fruits and vegetables. However, at present, there are only a few literatures that are slightly involved in the target recognition research, and most of them are only the judgment of a single growth form of fruits and vegetables, and they are too simplistic. Adaptive adopts corresponding methods to complete the accurate identification and subsequent picking of fruits and vegetabl...

AI Technical Summary

Problems solved by technology

However, at present, there are only a few literatures that are slightly involved in the target recognition research, and most of them are only the judgment of a single growth form of fruits and vegetables, and they are too simplistic. Self-adaption adopts corresponding methods to complete the accurate identification and subsequent picking of fruits and vegetables in different growth forms, which delays the practical progress of picking robots

Images