Font Size: a A A

Identification Method For Expanded Kiwifruits Based On Machine Vision And Development Of Software Based On Mobile Phone

Posted on:2020-03-10Degree:MasterType:Thesis
Country:ChinaCandidate:B YanFull Text:PDF
GTID:2393330596472810Subject:Agricultural Electrification and Automation
Abstract/Summary:PDF Full Text Request
The application of expanded agent can improve the production of kiwifruits,but there are many drawbacks in using the expanded agent,for example:it causes the reduction of kiwifruits'quality,shelf life and the increasement of rotting rate during storage.Therefore,identifying the kiwifruits treated with expanded agent,named as expanded kiwifruits,is one of the significant issues to be solved after the kiwifruits are harvested.Although present studies show that the hyperspectral imaging and NIR spectroscopy technologies can be used as non-destructive identification methods for expanded kiwifruits,the instruments used are expensive,especially hyperspectral image technology.To provide a non-destructive,quick,cheap and convenient method for the identification of expanded kiwifruits,the variety‘Hayward',‘Xuxiang'and‘Huayou'kiwifruits were used as the objects of study.According to the difference in projected area and the difference in calyx shape between expanded kiwifruits and normal ones,the identification methods for expanded kiwifruits of different varieties based on digital image processing and machine vision technology were proposed in this study.A smartphone software of identification for expanded kiwifruits was developed based on Android application development technologies.The major research and conclusions of this study were as follows:(1)Analysis on the physical parameters of expanded kiwifruits and normal ones,and the identification methods for expanded kiwifruits based on maximum diameter and projected area.The methods for analyzing expanded and normal kiwifruits were proposed based on analyzing the differences of physical parameters(mass and maximum diameter)of expanded and normal kiwifruits.The kiwifruit image acquisition system,which consists of supportor,mobile phone,background and a contrast object,was constructed subsequently.The maximum diameter of kiwifruit's projective plane was obtained after a serious of processings were used to the image of kiwifruit,and the distribution of maximum diameters of three varieties of kiwifruits were analyzed.The identification methods based on maximum diameter of kiwifruit's projective plane for‘Hayward',‘Xuxiang'and‘Huayou'expanded kiwifruits were proposed.The kiwifruit was regarded as an expanded kiwifruit if the maximum diameter was higher than 6.46,6.12 and 6.46cm for three varieties of kiwifruits respectively.Otherwise,the kiwifruit was regarded as a normal one.The methods had the identification accuracy rate of 82.5%,87.9%and 89.6%for‘Hayward',‘Xuxiang'and‘Huayou'expanded kiwifruits respectively.In order to improve the identification accuracy of expanded kiwifruits,the identification methods based on projected area for‘Hayward',‘Xuxiang'and‘Huayou'expanded kiwifruits were proposed after the distribution of projected areas of expanded and normal kiwifruits of three varieties were analyzed.The kiwifruit was regarded as an expanded kiwifruit if the projected area was higher than 28.4,25.6 and 28.4cm~2 for three varieties of kiwifruits respectively.Otherwise,the kiwifruit was regarded as a normal one.The methods had the identification accuracy rate of 87.1%,88.8%and 98.3%for‘Hayward',‘Xuxiang'and‘Huayou'expanded kiwifruits respectively.(2)The identification method for variety‘Hayward'expanded kiwifruits based on calyx shape.The visual attention mechanism,nonlinear gray transformation and K-means clustering algorithm were used to extract kiwifruit's calyx respectively according to the characteristics that there were differences in the length-width ratios of the minimum enclosing rectangles of calyxes between expanded kiwifruits and normal ones.But the experimental results showed that there were many overlapping areas in the distribution of length-width ratios of expanded and normal kiwifruits which indicated that the method based on visual attention mechanism was not suitable for identifying expanded kiwifruit.The calyx could be extracted accurately using nonlinear gray transformation and K-means clustering algorithm.It was found that the length-width ratio of the minimum enclosing rectangle of expanded kiwifruit's calyx region was basically greater than 1.6,and the length-width ratio of normal one was basically less than 1.6.Therefore,whether the length-width ratio was greater than 1.6 was used as criterion to identify expanded kiwifruit.The experimental results showed that the method based on nonlinear gray transformation had the total identification accuracy rate of 90.0%for identifying‘Hayward'expanded kiwifruit,and the method based on K-means clustering algorithm had the total identification accuracy rate of91.5%.Finally,an identification method for‘Hayward'expanded kiwifruits based on kiwifruit's projected area and calyx shape was proposed,the experimental results showed that the method had the identification accuracy rate of 92.5%.(3)The identification software for three varieties of expanded kiwifruits was developed based on Android platform.The identification methods for three varieties of expanded kiwifruits were transplanted on Android platform based on OpenCV library and JAVA programming language.The functions that the identification software need to have were analyzed and the operation interface and relevant algorithms were designed.Based on the integrated development environment named Android Studio,the interface of the software were designed by writing extensible markup language files(XML files)and the operating functions were implemented using JAVA language.Finally,the performance of the identification software was tested under different Android smartphones.The experimental results showed that the time needed for identifying expanded kiwifruits were 2s,0.2s and0.2s for‘Hayward',‘Xuxiang'and‘Huayou'kiwifruits respectively.
Keywords/Search Tags:kiwifruits, expansion agent, machine vision, Android, detection
PDF Full Text Request
Related items