Research On Sugarcane Internodes And Sugarcane Buds Identification Based On Machina Vision

In order to cut out effective sugarcane fragments with sugarcane buds through the machine intelligence.Using machine vision to identify sugarcane internodes. For the Characteristic difference of sugarcane internodes in form and color, combining pattern recognition and machine vision technology, in-depth study on the recognition algorithm. The algorithm basically solves the key issues of identification of sugarcane internodes It has a high value for theoretical basis to develop automatic equipment of sugarcane pretreatment. A lot of trial and test about internodes recognition have been made in the paper, and made some of test research about sugarcane buds, including image acquisition, image segmentation and sugarcane buds quality inspection.The main content on this dissertation are as follows:(1) The image acquisition hardware system was established, and prepared of the procedures of image acquisition, processing and testing.buds detection,such sugarcane edge detection, sugarcane image enhancement, filtering noise removal, segmentation; Proposed the evaluation standard of internodes detection methods and recognition effect for sugarcane internodes recognition(2) To address the tilt problem when sugarcane into the image acquisition system, proposed the method that taking the slope of fitting line, witch linear fit based on LS for center point set of sugarcane boundary. Using the border mutation and gray characteristics of sugarcane internodes, proposed a method by combing with linear fit and gray gradient to distinguish the direction of sugarcane.(3) Using the inflection point properties of sugarcane edge that appearing in internodes; Proposed a method of internodes recognition based on linear fit of center point set of the upper and lower edge. For the discontinuous statistical distribution of gray in Sugarcane internodes, Combined with the contribution of each color component, raised a method of internodes recognition based on median and fuzzy decision(4) For the complementary nature of the information base on the two continuous image and two images with the same location and different perspectives, raised an information fusion algorithm based on moving images and rotating images in different shooting angle. The experiment of Information fusion decision by extraction of the R component of tow images that with rotation angle access to single internodes identification rate was 95%.And the test of information fusion decision of multi-color component, and then by fuzzy clustering, the recognition rate of single internodes was 98.33%.(5) Use of the image characteristics of sugarcane internodes, extracting the H component of HSV color space, after binary image, division of the regional, extraction Center of mass ratio, Crude ratio and the white point ratio etc 7 characteristic indexes。Take the advantages of Support Vector Machine (SVM) in dealing with small samples and high dimensional data, the introduction of SVM to categories sugarcane internodes region. The recognition rate of classes of sugarcane internodes and internodes were 93.36%; The recognition rate of the number of internodes after minimum distance clustering were 94.12%.(6) Extraction sugarcane buds、internodes and sugarcane regional through artificial, and then taking a test analysis of there RGB and HSV color space information.Algorithm on sugarcane buds image segmentation:marquee area image segmentation with sugarcane buds based on OTSU, maximum entropy, the maximum variance FCM and KFCM(based Gaussian kernel).For the over-segmentation image is proposed compensation method based on symmetric.The less-segmented image is proposed a separation method using roundness and regional location prior knowledge. To reduce the running time, improved 8-point s equivalent inscribed circle was used to sugarcane bud separation and location.(7) Wavelet transform could decompose the image into approximate low-level image and the horizontal, vertical, inclined high-frequency detail components. For the vertical details component effectively reflect the position of sugarcane internodes, raised a method of internodes recognition based on Wavelet transform and fuzzy clustering. It is overcome the defects that method by fitting and gray statistics could only identify one sugarcane internodes in an image. Using this algorithm, single internodes recognition rate and overall recognition rate were 100% and 91.95%.