Picking point detection model construction method and picking point positioning method based on machine vision

Abstract

The invention discloses a picking point detection model construction method and a picking point positioning method based on machine vision, and belongs to the technical field of artificial intelligence detection. Collecting a multi-angle RGB image and a depth image of a picking target, generating a first point cloud set of the picking target through visual angle mapping, and cooperating with the first point cloud set as a training set; a three-dimensional reconstruction method is adopted to reconstruct a second point cloud set of a picking body tree canopy in real time to cooperate as a verification set and a test set, and a picking point detection model Alpha-VoteNet is trained. During picking, a third point cloud set of a picking object tree canopy layer is subjected to three-dimensional reconstruction in real time by adopting a three-dimensional reconstruction method, picking area marking frames of all picking targets of the picking object tree canopy layer are obtained by utilizing the picking point detection model, and the bottom surface center is calculated according to the top points of the bottom surfaces of the marking frames to obtain the three-dimensional coordinates of the picking points. The method overcomes the problems that in a traditional method, three-dimensional information of a picking target cannot be obtained, and the recognition precision is not high due to obstacle shielding, and has the advantages of being high in detection precision, high in recognition speed, high in generalization ability and the like.

Description

technical field [0001] The invention belongs to the technical field of artificial intelligence identification, and in particular relates to a method for constructing a picking point detection model and a picking point positioning method based on machine vision, which are used for picking tea tree buds. Background technique [0002] Tea originated in China and is a characteristic agricultural product of our country. For a long time, tea picking has been mainly carried out by traditional manual methods, and the timing and grade of picking can be judged by observing the shape of young shoots and leaves. Processing costs. For this reason, people invented mechanical tea picking machines, which are divided into single-person self-propelled and multi-person lifting picking. Although it can improve picking efficiency and reduce labor costs, due to the unevenness of the tea tree canopy buds, there is a "one size fits all" This method increases the difficulty of tea quality grading,...

AI Technical Summary

Problems solved by technology

[0006] Aiming at the problems of low detection accuracy and poor anti-interference performance of the existing picking point detection model, the present invention provides a machine vision-based picking point detection model construction method and a picking point positioning method; The first point cloud and the dense point cloud of the tea tree canopy based on the 3D reconstruction are used to train the picking point detection model. This model can realize the complete and accurate detection of tea tree canopy shoots, and at the same time, use the point cloud data to contain the characteristics of the target geometric information. Locate the bud picking point according to the detected three-dimensional information of the picking target, which has the advantages of strong anti-interference and high accuracy; in addition, it also simplifies the complexity of model design and enhances robustness

Images