A localization method of sticky citrus based on deep convolutional neural network model

Abstract

This application discloses a location method for sticky citrus based on a deep convolutional neural network model, including: obtaining the first visible light image; performing manual calibration to obtain the first training image, performing training to obtain the first citrus location model; performing the second 1. Picking and moving processing, so that the sticky citrus is the first proportion value; obtaining the second visible light image; performing manual calibration to obtain the second training image, performing training to obtain the second citrus positioning model; performing i-th picking And move the processing, so that the cohesive citrus is the i-th proportion value; obtain the i+1th visible light image; perform manual calibration to obtain the i+1th training image, and perform training to obtain the i+1th citrus positioning model; get The visible light image to be positioned obtains the preliminary citrus positioning result; obtains the proportion of cohesive citrus; if the proportion of cohesive citrus is greater than the proportion threshold, it is recorded as the final citrus positioning result, thereby realizing the accurate positioning of cohesive citrus.

Description

technical field [0001] The present application relates to a localization method, device, computer equipment and storage medium of sticky citrus based on a deep convolutional neural network model. Background technique [0002] The automatic intelligent and precise picking technology of mature citrus is an important part of agricultural automation. The citrus fruit positioning technology involved in the automatic intelligent and precise picking technology needs to be implemented in different wild fruit tree environments, such as the wild fruit tree environment with significant unstructured characteristics, and the multi-fruit clusters caused by fruit cluster growth at random spatial positions Adhesion occlusion affects the identification and positioning of citrus, and the variable light conditions of the citrus in the wild environment further affect the quality of the image of the citrus, thereby affecting the positioning accuracy. The traditional citrus positioning solutions...

AI Technical Summary

Problems solved by technology

The automatic intelligent and precise picking technology involves the positioning technology of citrus fruits, which needs to be implemented in different wild fruit tree environments, for example, in the wild fruit tree environment with significant unstructured characteristics, and the clustering of fruits at random spatial positions caused by multi-fruit Adhesive occlusion affects the identification and positioning of citrus, and the variable light conditions of the citrus in the wild will further affect the quality of the image of the citrus, thereby affecting the positioning accuracy

However, traditional citrus positioning solutions, such as trying to provide additional light sources through artificial lighting, still fail to solve the problem of sticking citrus that is difficult to locate.

Therefore, the prior art lacks a solution for accurately locating sticky citrus

Images