Study On Strawberry Recognition Classification And Spatial Positioning Model Based On Machine Vision

In recent years,the strawberry planting area in China continues to expand,and the yield has ranked the first in the world.Strawberry fruit contains a variety of nutrients,are favored by consumers,however,strawberry main by artificial cultivation and harvest,the harvest is short,the shortcomings of picking time consuming,not resistant to storage limits its commercial promotion progress,with strawberry cultivation mode of high promotion,implementation is of great significance to the mechanization of strawberry picking.Due to the individual differences of fruits,the maturity and weight of fruits in the same period are different,so it is necessary to classify the fruits in different grades during mechanized picking.In order to prevent the fruits from secondary mechanical damage during picking,it is necessary to accurately locate the fruits.In view of this,this paper studies the fruit identification,classification and positioning of strawberries based on the elevated strawberries.The specific research is as follows:(1)The hardware and software system of strawberry image processing is built.The hardware part includes camera,adjustable lens,bracket,annular shadow-less light source and computer.The software part includes image acquisition control module,image preprocessing module,target segmentation and extraction module.(2)The color space conversion method is used to transform the image into YCb Cr space,and Cr channel is selected for target subsequent processing.Gaussian filtering is used to eliminate the noise in the image.Compared with FCM segmentation and OTUS double threshold method,FCM segmentation method is selected according to the needs,with an average detection rate of 94.45% and an average consumption time of 0.956 s.The segmented image is processed by binarization,corrosion and expansion in turn.Canny edge detection operator is used to detect the target area,and then the minimum rectangle and image fusion method are used to extract the target strawberry area,which is used to make the sample library,and the image geometric transformation method is used to enhance the sample library.(3)According to the theory of Goog Le Net convolution neural network,the strawberry recognition model is built,and the model is trained by the established sample database.The recognition experiment shows that the average recognition rate of the model is 98.65%,and the average detection time of a single image is 0.0468 s.Then,according to the CR color component,area perimeter and area characteristics of strawberry image,the prediction models of strawberry fruit maturity and weight classification were established respectively.The accuracy of the model classification experiment is 92.10% and 87.57% respectively.(4)The binocular vision positioning system is built,and the conversion relationship from two-dimensional pixel coordinates to three-dimensional spatial coordinates is established.Zhang’s calibration method is used to calibrate the binocular equipment.The internal and external parameters of binocular equipment are obtained,and the left and right views are corrected.Based on the model analysis of strawberry particle extraction,the location prediction of picking point was carried out.When the distance is 30 cm,the positioning error is±5 mm and the error rate is±3%The experimental results show that the established model has a good test effect and can accurately identify,classify and locate strawberry.