Fruit visual collaborative searching method of harvesting robot

Abstract

The invention discloses a fruit visual collaborative searching method of a harvesting robot. According to the method, a large-small dual-view-field visual acquisition system is employed; a large-view-field camera is arranged at a mobile platform of a robot and is used for carrying out global imaging on a fruit tree in a orchard, carrying out salient region extraction on the image, measuring a fruit likelihood target by using a target likelihood value, removing a small region, determining an approximate region of a fruit target in a large-view-field image, and guiding the robot to approach thefruit tree; a small-view-field camera is arranged at a mechanical arm on which an end effector of the robot is arranged and is used for establishing a large-view-field image coordinate system and a small-view-field image coordinate system as well as a mapping relationship between the view-field coordinate systems and a world coordinate system and carrying out regional imaging based on fruit targetsearch region information obtained in the large-view-field current image and the coordinate system mapping relationship so as to search for a fruit cooperatively. The method is similar to human-eye target searching; blind out-of-order searching of the robot is avoided; and the foundation is laid for follow-up precise fruit identification.

Description

technical field [0001] The invention relates to the technical field of target search for picking robots, especially the visual search for fruits. Background technique [0002] When the picking robot enters the orchard, its primary task is to search for fruits. The so-called fruit search refers to the process of searching the fruit area for further identification during the working process of the picking robot. In the existing literature at home and abroad, the fruit search method is only involved in other related technical literature of picking robots. For example, Jiménez AR et al. used infrared laser ranging sensors to search and detect spherical targets in unstructured environments in the research of fruit recognition. objects, but this method has poor real-time performance and high cost. In the research of real-time obstacle avoidance algorithm, Li Zhankun and others set four different paths based on monocular vision to search for fruits, but due to the fixed path, the...

AI Technical Summary

Problems solved by technology

In the existing literature at home and abroad, the fruit search method is only involved in other related technical literature of picking robots. For example, Jiménez AR et al. used infrared laser ranging sensors to search and detect spherical targets in unstructured environments in the research of fruit recognition. objects, but this method has poor real-time performance and high cost

In the research of real-time obstacle avoidance algorithm, Li Zhankun and others set four different paths based on monocular vision to search for fruits, but due to the fixed path, the adaptability is not good

Generally speaking, there is still no suitable fruit search method at present, so that the picking robot can realize the effective search of fruits and further identify them.

Images