Sunshine rose grape fruit big and small grain image identification method

Abstract

The invention discloses a sunshine rose grape fruit big and small grain image identification method, which comprises the following steps of (1) picture data set production: collecting sunshine rose grape cluster picture data, performing instance segmentation and labeling on fruit grains in the sunshine rose grape cluster picture data, and dividing the fruit grains into a training set, a verification set and a test set according to a certain proportion; (2) machine learning model data processing; (3) machine learning model training; (4) constructing an instance segmentation network; (5) constructing a loss function; (6) training an instance segmentation model; and (7) model reasoning: obtaining a big and small particle classification result. According to the method, the instance segmentation and machine learning algorithm is utilized, the large-scale rapid measurement requirement of the sunshine rose grapes is met, meanwhile, the fruit indexes can be quantified, and the recognition accuracy is improved.

Description

technical field [0001] The present invention relates to the technical field of grape image recognition, and more specifically, to a method for recognizing the size and grain image of sunshine rose grapes. Background technique [0002] Sunshine Rose grapes have been favored by consumers for their sweet taste and unique rose fragrance in recent years. With the continuous growth of market demand, the planting scale of Sunshine Rose grapes is also expanding rapidly. Quality control is especially important. [0003] Grape fruit size and grain size is a common physiological disease in the cultivation of Sunshine Rose grapes. It is specifically manifested as uneven fruit size and obvious size grain phenomenon. If the problem of fruit size and grain is not dealt with in time, it will lead to a decrease in the yield and quality of Sunshine Rose grapes, which will have an important impact on product value and economic benefits. [0004] The traditional method of judging the size of ...

AI Technical Summary

Problems solved by technology

The method based on image classification usually divides grape images into two categories: normal and large and small, but the characteristics of large and small grains are scattered, and different labelers have different standards, so it is difficult to unify the labeling standards, which will easily cause ambiguity in the labeling data, and eventually lead to identification. low rate

Images