Fruit identification method and device under same-color background and fruit picking robot

Abstract

The invention provides a fruit identification method under the same color system background, and belongs to the technical field of fruit picking robots, and the method comprises the steps: obtaining an environment photo in an orchard environment; inputting the environment photo into a recognition model, and determining whether a fruit exists in the environment photo, wherein the recognition model is obtained by using a plurality of groups of data through machine learning training based on a constructed transformer model and a neural feedforward network (FFN), each group of data in the plurality of groups of data comprises environment photos with fruits and marking information for marking the photos with the fruits. According to the method, a transform encoder-decoder model is used, parallel processing can be carried out, and the target recognition speed is increased; and the method is stable, high in speed and relatively high in precision, and can well meet agricultural requirements of a fruit picking robot, yield prediction and the like. A sample space is expanded by using a resampling method, so that the method well adapts to a small sample data set, is high in generalization ability, and can be applied to various robot vision systems for picking or preproducing fruits.

Description

technical field [0001] The invention relates to the technical field of fruit picking robots, in particular to a fruit recognition method and device based on a transformer model under the background of the same color system, and a fruit picking robot. Background technique [0002] In the body structure of agricultural robots, the machine vision system is like human eyes, helping the robot understand the surrounding environment and identify and locate the target. Machine vision has been widely used in fruit and vegetable yield prediction and target recognition for picking robots. For the predicted yield of fruits and vegetables, growers can be provided with more refined and perfect scientific management methods according to the results; and the accurate and fast positioning of target fruits has a huge impact on the real-time work of picking robots. [0003] When performing machine recognition, the accuracy of target recognition and target positioning is the key to the vision ...

AI Technical Summary

Problems solved by technology

Among them, the recognition method based on machine learning is usually accompanied by preprocessing, feature selection and other operations, which cannot realize the end-to-end detection process, and the recognition effect is easily affected by various disturbances in the natural environment

Although the recognition method based on deep learning has significantly improved the accuracy and can realize the end-to-end detection process, due to operations such as convolution and the dependence of the model on the anchor box, it requires a lot of computing and storage resources. The recognition speed is not yet up to the real-time requirements

Images