Fruit detection and yield estimation method and system based on machine vision

Abstract

The invention discloses a fruit detection and yield estimation method and system based on machine vision, and relates to the technical field of artificial intelligence target recognition and application. In order to solve the problem that fruit yield estimation is not accurate enough in the prior art, fruit video stream data sets of three view angles are collected; designing a CSP module based on a Transform encoder and an attention feature fusion mechanism, and integrating the CSP module into YOLOv5 to construct a target detection model YOLOv5-FF based on multi-level feature fusion neck; focal EIoU Loss is adopted as a frame loss function for training to obtain a fruit target detection model, a detection result of a YOLOv5-FF model for continuous frame images of a fruit video stream is input into a target tracking model Deep Sort, exclusive numbers are distributed for successfully matched fruits through cascade matching and an IoU matching mechanism, the number of fruits under multiple view angles is obtained, the fruit detection efficiency is improved, and the fruit quality is improved. The deployment is convenient, and the detection is efficient and accurate; the provided yield estimation model can realize high-precision measurement and calculation of the yield of the orchard.

Description

technical field [0001] The invention belongs to the technical field of artificial intelligence target recognition and application, and in particular relates to a method and system for fruit detection and yield estimation based on machine vision. Background technique [0002] China has a long history of producing pears and is one of the characteristic agricultural products in my country. The high yield per mu of pear trees requires a lot of human resources to monitor their growth conditions. For individual farmers and producer organizations, how to pick pears at the ripe stage in time and how to accurately estimate the yield of pears in the current year are important to improve the field management of pears. major challenges faced. In order to improve production efficiency and reduce labor consumption, the invention of an accurate and efficient target detection method and yield estimation method for crisp pears is crucial to the modern production of pear industry. The pear ...

AI Technical Summary

Problems solved by technology

[0006] Aiming at the problem that the fruit yield estimation in the prior art usually only collects target fruit tree images of a single perspective or single dimension, the fruit information is likely to be lost for densely distributed or high-yielding fruits, and the fruit yield estimation is not accurate enough. The present invention discloses a machine-based Visual fruit detection and yield estimation method and system, build a production capacity estimation model, realize intelligent estimation of orchard yield, rich collection dimensions, high accuracy

Images