| Rapeseed is the main oil crop,feed crop and ornamental crop in China.However,the intelligent level of machinery used for rapeseed planting and harvesting is still relatively low.The harvesting efficiency of rapeseed combine harvester is closely related to the feeding amount,and the harvesting quality of rapeseed is determined by the quality of its harvest,mainly by the impurity content,etc.Aiming at the problems of low intelligence level,low harvest efficiency and difficult evaluation and online detection of harvest quality of rapeseed combine harvester in the harvest process,this paper proposes a rapeseed harvest feeding quantity and quality monitoring system based on machine vision.This article first analyzes the work process of rape combine harvester,on this basis,according to the working principle of the harvester,rape growth characteristics and the characteristics of the granary of harvesting rape,build the rape harvest feeding amount detection system and rape harvest impurity content detection system,and according to different object recognition,obtained the corresponding image recognition algorithm,to complete the online monitoring for the feed rate and amount of impurity,improve the intelligent level of rape combine harvester.The main work and specific research results of this paper are as follows:(1)According to the growth characteristics of rape,convolutional neural networks is used in this paper to design and build a set of image classification models that can accurately predict the density of rapeseed.By comparing traditional image classification models with image classification models based on deep learning and Field experiments,we have proved that the model designed in this paper is more in line with the characteristics of rape.In addition,machine vision technology is used in this paper to study the height measurement of rape in the field,which can detect the feeding amount more accurately.(2)For quality monitoring,the impurity content is used as an evaluation index.According to the characteristics of harvested rape,the difference in color between rapeseed and straw is used to detect the impurity content.An image brightness equalization algorithm is proposed,which effectively solves the problem of uneven brightness in the granary;Based on the comparison and analysis of commonly used threshold segmentation algorithms,the maximum entropy threshold segmentation method is improved,and the rape harvest impurity detection is constructed in combination with image processing technology system.(3)Through the laboratory simulation experiment and experimental bench simulation experiment of the system,the scientificity and rationality of the feed quantity detection system and the impurity content detection system used in this paper are verified. |
