The Mechanism Research On Cottonseed’s Internal Quality Testing Based On Image Processing Technology

The seed is the most basic means of production in agricultural production, the quality of cottonseed isone of the important factors which is affecting the yield and quality of cotton. One of the important indexesof seed quality was seed germination rate. The standard germination experiment was a conventionalmethod to check the quality of cotton seed, which was carried out under the condition of laboratory. But theproduction practice showed that germination was different and had a big gap between ideal conditions inthe laboratory and the natural environment. And it had long experiment period and the intensity of laborwas big, the speed, accuracy and the result analysis and so on of data processing can’t meet the needs of thepractical production. Therefore, it was necessary to explore a sorting technology of cotton internal qualitybased on a kind of image processing technology in view of the problems.The mechanism research oncottonseed’s internal quality testing based on image processing technology. Firstlly, the picture ofcottonseed were acquired and the color feature extraction were extracted by the software. The testing modelof cotton seed’s internal quality was established based on neural network to check the seed vigor of cottonseed. The main research content was as follows:(1) The different ways of light source and lighting were analyzed. The white light of LED and theforward lighting were selected according to the characteristics of cottonseed’s color. The cottonseed imageswere analyzed under the RGB color model. The bimodal characteristics of image were obvious, meanwhilethe background and target were distinguished clearly. The value distribution of image grey was analyzed. Itconcluded that the pixel distribution of R, G, B color component of cottonseed’s image.(2) The different methods of image filtering and image segmentation were analyzed. The gaussianfiltering method was selected for image filtering, and7x7template for processing of the filter template waschoosed. Image segmentation The k-means clustering algorithm method was choosed, and the backgroundand target were segmented effectively. The collection platform of cotton seed’s color was designed basedon MATLAB GUI platform. The color characteristic information of cotton seed collected by the samplingplatform which was rapid extraction by color histogram method.(3) The relationship between seed vigor, germination rate and simulated field seedling emergence ofcottonseed lot were researched. The correlation between color features and seed vigor of cottonseed werediscussed. The experiment results showed that cotton color characteristic parameters and the soaking liquidconductivity value presented highly correlation. The study offered a scientific basis for rapid detection ofcotton-seed’s vigour based on digital image processing technology. It laid a theoretical foundation to thefast nondestructive testing of the cotton-seed’s internal quality.(4) The internal quality testing model of cotton was established based on the BP neural network. Theneural network and network parameters were designed and set. New Land Early45and Ding Feng10wasgot by300each to training the model. The two cotton varieties of100grains were carried out respectively to the experiment. The results showed that the method using BP neural network detection rate canrespectively reached82.7%and86.1%.