Dissertation
Dissertation > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device

Study of Classification of Fruit Flies Insects Based on Digital Image Processing Technology

Author HaoZhongHua
Tutor NiYuanPing
School Kunming University of Science and Technology
Course Control Theory and Control Engineering
Keywords Digital Image Processing 27 kinds of fruit fly image Adaptive threshold algorithm Mathematical Morphology Support Vector Machine
CLC TP391.41
Type Master's thesis
Year 2010
Downloads 97
Quotes 0
Download Dissertation

Papers using digital image processing and pattern recognition techniques belonging to 3 subfamilies, Group 4, 8 are 27 kinds of Yunnan region's major fruit fly pests automatic identification Classification. The realization of the Yunnan region mainly 27 kinds of fruit fly pests of automatic identification. The main research contents and results are as follows: (1) 27 kinds of fruit fly pests image preprocessing. In this paper, a color image binarization, image denoising and other aspects of the study. The results showed that the blue color component of the fruit fly grayscale image having a large contrast image details are clearer. Using a median filtering method of the blue component of image denoising, reducing noise. (2) 27 kinds of fruit fly pests image segmentation. This paper uses the idea of ??adaptive threshold algorithm will be a fruit fly image is divided into head, thorax, abdomen, left wing, right wing five parts. The characteristics of the different parts, using a different segmentation algorithms. For image segmentation traditional mathematical morphology algorithm is improved deburring and for transparent wings segmentation. The application results show that the proposed algorithm is divided in strong noise background with good results. (3) 27 kinds of fruit fly pests image feature extraction and optimization. On the analysis of the characteristics of each part and taking into account the actual situation of noisy images, extracted on the wing area, markings, focus and other features; extracted in the chest color, texture and other features. The results show that the practical application of these characteristics are typical and effective, they apply under strong noise image feature extraction, not with the displacement of the image, rotate, stretch and change. (4) 27 kinds of fruit fly pests image classification. Taking into account the limited sample of the actual situation in accordance with lower error rate and improve the real-time processing capabilities and the principle of minimum misclassification, the paper selected from the binary tree and support vector machine classification method of combining. The characteristic difference between binary large sample classification with fast and intuitive features; support vector machine compared with the traditional statistical theory, is a specialized case of small training sample machine learning theory of law, it solves the neural network exist over learning and local extreme defects. Experiments show that binary tree and support vector machine classification method of combining is an effective method.

Related Dissertations
More Dissertations