Study And Application Of Image Segmentation Based On Snake And The Level Set

With the development of artificial intelligence and computer technology, human beings began to study how to make computers simulate human thought processes, and give some work about anglicizing and understanding the image to computer, resulting in the computer vision. Image segmentation is a very important and difficult technology in image processing, and is an important key step in research field of computer vision. It is a technology and process which makes an image into each region with characteristics by the gray level, color, texture and other characteristics of the image and extract interested objectives. With the increasing requirements of the target object extraction, color images segmentation are paid more and more attention. Color image contains a wealth of color space information, texture information and other image information for more accurate image segmentation, color image segmentation is a future trend. Image segmentation has a wide range of applications, which can be used in industrial image processing, biomedical image processing, image transmission, identification, image retrieval, image coding, motion tracking and other fields, and therefore, the research on image segmentation has important practical significance.Although image segmentation has very extensive applications, but there is not an image segmentation algorithm to adapt to all applications. Over the years, many image segmentation methods are made. Whether it is grayscale or color image segmentation, each segmentation method has its own strengths and weaknesses and it's range is not wide, or it can not be effectively addressed in some cases and is unable to get satisfactory segmentation results. For this, the paper research on the gray image segmentation and color image segmentation in-depth, and strive to find a better and more efficient image segmentation algorithm to lay a good foundation for moving target detection.Moving target detection widely used in daily life and brought a huge market value as another important issue in the field of computer vision. Moving target detection's principle is that makes the moving object and background in the video separate from, and extracts the characteristics of moving objects and related sports information. As the basis of target tracking and classification. target detection has an important role. It can provide the target object's initial motion information to target tracking and target classification, the accuracy of target detection affects on the accuracy of tracking and classification.This ultimate goal is to get moving object by object segmentation contours, then detect the target object and extract features of the target object, make application for subsequent tracking. First, the paper introduces the basics of image segmentation, including the definition of image segmentation and classification of image segmentation algorithm. and then introduces the basic theory of active contour models, solution process of energy minimization, and improvement and development of the active contour model. We proposed a segmentation algorithm for objects that exhibit relatively homogeneous photometric characteristics, embedded in complex background clutter. And we introduce the global minimization of the active contour model into our proposed method. Our method is that a simple binary classifier can be arrived at by summing the log-probability of error, for both missed detections and false alarms. And we propose an adaptive lookout region, whose size depends on the statistics of the data. It can improve the convergence rate by calculating global minimization methods through the use of active contour model, and makes the energy functional with a global optimality. and because there is no use to calculate and initialize the distance function repeatability, computational algorithms can be reduced to some extent.Secondly, the color image segmentation is researched. For one thing, the paper introduces three color spaces simply; the analysis shows that selecting the appropriate color space plays an important role to image segmentation. Each color space play a role in application areas to the color image segmentation of the research in their respective, people also try to apply the appropriate color space on these occasions. Secondly, there is an analysis to some of the current color image segmentations. And then the basic theory of level set method and typical of the MS model and the CV model are briefly introduced, on this basis, color image segmentation method based on the level set method is proposed, which make full use of the target image region and edge information, and replace a weighted gray value with Euclidean distance so as to obtain a more accurate color edge. At the same time, considering the various relationships between the color channels, each channel minimizes the distortion because of the difference in the shape information to improve the accuracy of image segmentation.Finally, moving target detection methods are introduced and their advantages and disadvantages are analyzed. Then the results of image segmentation apply to moving target detection. The principle of moving target detection is that make the current frame image segmentation contour information as the primary contour of the next frame to detect moving target objects. After detecting moving targets, we need to calculate area, centroid and other features of in the contour, set an appropriate threshold value, remove the larger or smaller contour, so we can achieve the target object and target object area, external rectangular box. centroid and other relevant characteristics information.Experimental results show that the proposed image segmentation algorithm and motion detection are able to obtain good results, and they have smaller computational complexity and the advantages of fast convergence speed. Finally, this thesis summarizes the work and put forward to next research.