Research On Maturity Detection Of Greenhouse Tomato Based On Machine Vision

Since the 16 th century,tomatoes have emerged as an essential ingredient in the public's field of vision.Not only because it contains sufficient nutrients,but also has a variety of medicinal values,for example,can stimulate thirst,clear away heat,reduce blood pressure,lower cholesterol and so on.In recent years,with the increasing demand for tomatoes,today's consumers generally improve the quality and safety requirements of fresh fruits and vegetables,and improving safety quality has become an crucial part of quality production.Therefore,it is urgent to study a more effective and practical algorithm for tomato maturity detection.Maturity testing is one of the important links in the processing of agricultural products.So far,the maturity testing of greenhouse crops has been carried out mainly by humans.This method not only has low efficiency,but more importantly,the subjective factors of people have a huge impact on the results.The use of machine vision technology to optimize maturity testing not only reduces damage to fruits and vegetables but also improves efficiency.In view of the problems in the detection of artificial maturity,this paper aims to optimize the tomato maturity detection method,and carried out the research on the detection of maturity of greenhouse tomato based on machine vision.In this thesis,Grab Cut and support vector machine are used to classify tomato maturity based on the shape and color characteristics of tomato.By comparing and analyzing multiple maturity detection algorithms,a more discriminative algorithm is selected to make the greenhouse Tomato maturity detection is better.The maturity detection is facilitated by cropping the original image and segmenting the image of the target area.This study mainly completed the following work.Firstly,the software part and hardware part of image acquisition are introduced.Then,based on the different maturity stages of tomato,the effects of greenhouse tomato images with different illumination conditions are discussed.Finally,the various research methods of image segmentation and the principle of image segmentation by Graphcuts algorithm,the difference between the traditional Grab cut algorithm and the Graphcuts algorithm and the advantages of the improved Grab cut algorithm are introduced.Images were collected under various light conditions,and three different growth stages of green ripe,semi-ripe,and mature were selected for the detection of maturity.The improved tomato image was used to segment the collected tomato image to identify and extract the tomato body of interest.Then choose a more suitable color space,analyze and judge the conversion of the tomato image from theRGB color space to the HSV color space.The feature extraction of the collected tomato images was carried out by using the above algorithm.According to the shape characteristics and color characteristics,the classification of tomato fruit was studied.The H component color feature extraction under HSV color space was determined by data comparison.The training samples were subdivided by K-means clustering operation.The classifier is designed corresponding to the subset of training samples obtained after clustering,and the SVM model is trained.Finally,the detection of greenhouse tomato maturity is achieved.The improved Grab cut algorithm has certain advantages in processing speed.From the results of image segmentation,the improved Grab cut algorithm is basically the same as the traditional algorithm,and the difference is not significant.Automatic Grab cut segmentation is achieved on the foreground and background segmentation issues.The Grab cut was first applied to the identification of greenhouse tomato images,and the maturity of greenhouse tomatoes was detected by the combination of K-means and SVM.The experimental results show that the algorithm has a good discrimination and achieves a classification accuracy of over 92%.