Research On Corn Seed Varieties Intelligent Identification System

With the rapid development of computer and information technology, it is a necessary trend to fulfill agriculture product test automation and intelligential. Based on the foundation of computer vision and pattern recognition theory, took nongda108,ludan981,zhengdan958,zhengdan6028,liaoyu2hao as recognition object, got the corn seeds images and characteristic parameters, Reduced-order optimization, established the best classification model, the paper realized intelligent identification to corn seeds and improved the accuracy. The correct recognition rate in an average of five corn varieties is up to 97%.1. In this paper, we mainly improved the image obtain system and quality of the corn seeds image. Additional, the lightening room is also improved, so as to improve the lighten source, as a result, the quality of corn seeds image is improved. It is beneficial to promote the precision of corn seeds characteristic identification and the accurate of intelligence identification.2. With the advance image preprocess method, promote the image preprocess effect. During the division of background, with the minimum error probability, we obtain the threshold value in the red section of corn seeds gray image to binarize. In order to improve the image quality and characteristic obtaining precision, we processed the median filter and morphology, and obtained accurate corn seeds profile image.3. With Principal component analysis and support vector machine, we processed the corn seeds characteristic parameter degradedly. We chose the first 5 principal components from 19 characteristic parameters which can reflect the corn seeds morphology, as the final classification indicate. And with the Principal component analysis, we simplify the classification model to a large degree.4. With support vector machine to identify the corn seeds, we improve the accurate of identification. Put the principal characteristic as the final classification indicate into the model which is found by the support vector machine, we tested Zhengdan958, Nongda108, Ludan981, Zhengdan6028, Liaoyu2, and the accurate was 95%, 97%, 95%, 100% and 98%, respectively. 5. On the basis of VC++6.0 and MATLAB7.0, we explored the corn seeds intelligent identification software system, through test this software, we found the outlook was friendly, and it can identify the corn seeds with higher precision, as well as the flexibility of expansion.