Fruit growth form identification method based on image rendering

Abstract

The invention relates to the technical field of convolutional neural networks, in particular to a fruit growth form identification method based on image rendering, and the method comprises the steps: collecting a fruit image; enhancing the image data; constructing a backbone network, and performing feature fusion on a network layer of the backbone network through a top-down and bottom-up bidirectional fusion network so as to perform feature extraction on the image; an RPN network is constructed; the method comprises the following steps of: firstly, selecting a small number of true value points for prediction; secondly, point-by-point feature representation is extracted from each selected point, and finally an MLP prediction classification is constructed; and sending a test set image into the trained network model and then carrying out forward propagation. According to the method, the feature extraction network is built based on deep learning, and accurate identification of the fruit growth form is realized based on the idea of image rendering, so that the picking robot can automatically identify the fruit growth form, and a foundation is laid for further selecting a corresponding picking mechanism for the picking robot.

Description

technical field [0001] The invention relates to the technical field of convolutional neural networks, in particular to a method for identifying fruit growth shape based on image rendering. Background technique [0002] The planting scale and output of fruits are increasing year by year around the world, but at present, most of the fruit picking work is done manually, and manual labor is time-consuming and laborious. Therefore, the development of high-intelligence picking robot-related technologies has important practical significance and broad application prospects. As an important part of the picking robot, the visual system's recognition accuracy and speed have a great impact on the working efficiency of the picking robot. Although researchers have carried out research on fruit and vegetable vision systems over the years, few picking robots have reached commercial maturity so far. [0003] At present, deep learning methods have achieved high detection rate and fast detec...

AI Technical Summary

Problems solved by technology

However, it is difficult for the current picking robot prototypes to pick fruits and vegetables that are covered or overlapped by branches. If the picking robot forcibly picks off the fruits that are covered or overlapped by branches, it may cause damage to the end effector and damage to the fruit, resulting in unfavorable results. necessary economic loss

Images