基于计算机视觉与神经网络的芒果等级分类研究-检测技术与自动化装置专业论文.docxVIP

优秀毕业论文 精品参考文献资料 桂林工学院硕士学位论文Abstract 桂林工学院硕士学位论文 Abstract The fast non—destructive inspection on fruit is very important to improve the production efficiency and reduce the labor intensity．Taking mango as the study object， the paper described the method of mango classification and the flaw characteristic on the surface of mango based on the computer vision technology and the neural network technology．Realized the mango classification according to the size，the color,the flaw and the wavelet characteristic of mango using the BP neural network technology．The main content are as follows： The surface of mango is quite smooth and has good reflection performance，SO the paper selected the black extinction image background for its advantages to the following process．Based on this，the paper described many kinds of pre．processing method such as the image processing window determination，the image filter,the image intensification，the picture edge examination，the image division and SO On． Taking both speed and effect into account，the paper used fast median filtering algorithm to eliminate noise disturbance merged in image gathering process，used the gray linear substitution to carry on the intensification，used the wavelet transformation to carry on the mango picture edge examination．Because the brightness of flaw area was between black background and the smooth surface of mango，the paper divided the mango image and the background using the Maximal Variance method． Based on the study to the shape，color,flaw and wavelet characteristic of mango， the paper chose transverse diameter(W)，the mango color(H component value)(C)， fruit shape index(E)，the flaw area percentage(P)，mango’S wavelet characteristic parameter(Cal，Chl，Cvl，Cd 1)as its shape characteristic，color characteristic，flaw characteristic，wavelet characteristic，and described their influence to classification through the qualitative analysis to these characteristics． Introduced the merit of using the neural network to t