Fresh Cut Flower Grading Algorithm Based On Deep Learning

Flowers are a kind of ornamental plant.In recent years,the fresh cut flowers industry in Yunnan Province of China has made certain achievements,such as improved seed breeding,green production,primary processing,and deep processing.During the process of flower sales,flower quality grade identification is an important and arduous task.The traditional classification of fresh cut flowers has been mainly relying on manual work,which has the disadvantages of low efficiency and poor accuracy.Manual grading cannot meet the needs of the insurance period of fresh flowers,the rapid growth of transportation,and market demand.In the highly competitive global flower market,it is very important for the development of the fresh cut flower industry in Yunnan to quickly and effectively evaluate the quality of fresh cut flower products and maintain high standards.This research uses machine vision technology combined with deep learning methods to perform non-destructive testing on the quality of fresh cut flowers,and comprehensively evaluate and grade the quality of fresh cut flowers by measuring the different characteristics of fresh cut flowers in all aspects.The main research content and results of this paper are as follows:(1)A classification method for the maturity status characteristics of fresh cut flowers is designed.A maturity grading method for fresh cut flowers based on deep learning and depth information is proposed.The color image and depth information of the fresh cut flower buds are collected as the maturity grading data set.After that,data preprocessing is performed to form RGBD flower bud data.A four-channel convolutional neural network model is constructed,so that the depth data of the surface of fresh cut flower buds can also be extracted through the convolutional neural network.The experiment compares the feature extraction capabilities of four different backbone networks.In the improved network with Inception V3 as the backbone network,the best classification accuracy rate reached 98%,and the Res Net18 as the backbone network achieved faster recognition results.At the same time,compared with other different classification models,experiments show that this method can quickly and accurately classify the maturity of fresh cut flowers.(2)For the morphological characteristics of fresh cut flowers,a method for detecting morphological indicators of fresh cut flowers buds and stems based on target detection algorithm is proposed.A YOLOV5 network model based on attention mechanism was constructed to conduct object detection experiments on the height of buds and lengths of flower stems in the horizontal image of fresh cut flowers.And to evaluate the results of the model,the improved network model has better target detection capabilities.The region of fresh cut flowers can be identified more accurately.Then,according to the detection region in each picture,the morphological index of the fresh cut flowers is calculated.Then,compare the calculated results of the model with the actual measured results.The error of feature calculation is small.The proposed method can calculate and extract the morphological information of fresh cut flowers more accurately.(3)Integrate the maturity characteristics of fresh cut flowers,the morphological characteristics of flower buds and stems,and the color characteristics of flower buds,and analyze the multi-characteristic index of fresh cut flower quality.Comprehensive classification of the comprehensive information of fresh cut flowers.A multi-index comprehensive classification model using extreme learning machines is proposed,and network optimization algorithms are used to train and optimize the parameters of the model.Finally,the comprehensive classification effect of fresh cut flowers based on the Particle Swarm Optimization algorithm and the extreme learning machine model is the best,reaching a classification accuracy of 97.9%.The comprehensive analysis of fresh cut flowers with different quality characteristics can be carried out more accurately,and the corresponding comprehensive classification of fresh cut flowers can be obtained.Compared with the traditional machine learning classifier model,our method has a better classification effect.At the same time,in order to meet the needs of real-time display,the interface of the host computer for comprehensive grading of fresh cut flowers is designed.The comprehensive grading method for fresh cut flowers studied in this article can comprehensively evaluate fresh cut flowers with different characteristics and qualities.It can classify fresh cut flowers more accurately and quickly.It can provide a feasible scheme for quality grading in deep processing of fresh cut flowers.