Image Enhancement And Segmentation Algorithm Based On Dynamical System And Its Application In Forest Fire Remote Sensing

Two key issues in image processing are image enhancement and image segmentation.Good effects of image enhancement and segmentation are the foundation for the follow-up work proceeded smoothly.With the development of information technology and electronic equipment,the increasing amount of data,the more diversified types of image,the particular and the unpredictable complexity of image,the concern about the speed and quality of image processing is growing because of the real-time requirement.Therefore,effectively realizing real-time,automatic,efficient and high quality image enhancement and segmentation are still extremely important and urgent issues.The dynamic system has significant advantages in the study of bifurcation problems with unusual complexity and chaos in nature.Some complex problems of image processing which cannot be solved by using the previous image processing methods can be better solved by using the theory of dynamical system.Dynamic system can be used in image enhancement and segmentation to get better processing effect and get wide attention.This paper will launch the research from following several aspects:the above problems of image enhancement and image segmentation;the application of dynamical properties of the dynamical system in image enhancement and image segmentation.During this process,simulation experiments have been described in this paper using wildfire remote sensing images in accordance with current development of imagery.The main works and innovations of this thesis are listed below:(1)Image enhancement:In order to solve the problem of de-noising of images with more image data,rich texture and difficult eliminating interference information,this paper proposed the filtering method based on the improved anisotropic diffusion dynamics model and improved anisotropic diffusion PM model.The traditional PM model is not adaptive.Through the improvement of the model diffusion function and the threshold value,it can not only achieve the effect of de-noising and edge preserving,but also realize the adaptive threshold selection.In order to solve the problem of the edge weakening caused by the filtering noise and the problem that the image is not clear and the contrast is low,six dimensional feed forward neural network model is constructed in this paper.Based on the analysis of the dynamic properties of the model,an image enhancement algorithm based on the six dimension feed forward neural network model is proposed.These two algorithms are both suitable for gray image and color image and be more effective compared with other algorithms.(2)Image segmentation:This paper proposes an edge detection algorithm based on coupled cellular neural network dynamic model concerning the issues of remote sensing images' kind edge detection with large data,high complexity and real time requirement.In this algorithm,the traditional neural network is improved by using the hyperbolic tangent function and the model coupling,which reduces the number of iterations and makes it more real-time.In order to solve the problem of high precision and difficult to detect the edge and threshold adaptively,a color image edge detection algorithm based on the reaction-diffusion equation is proposed.The algorithm improves the cellular neural network based on the reaction-diffusion model,which makes it more suitable for the color image processing on the threshold adaptive selection.For the complex image with more interference information,it is difficult to achieve the segmentation accurately,and an image segmentation method based on dynamic gradient adaptive geodesic active contour model is proposed.Through the improvement of three important parameters of the GGAC model,this method can not only achieve adaptive selection on the parameters,but also be more robust.The experimental results show that the segmentation quality is high.Because the color of the color image is complex,it is difficult to accurately achieve the specific region segmentation problem.In this method,the color clustering algorithm is introduced into the fire area segmentation of color forest fire remote sensing image,and the validity of the algorithm is verified by experiments.Finally,the results of image enhancement and segmentation are used to calculate the area,perimeter,length and width of the forest fire area,and determine the shape of the forest fire.These data provide the data support for the identification of forest fires,and play an important role in forest fire assessment.At the same time,this paper has developed the corresponding remote sensing image processing system of forest fire,which has a far-reaching significance to the forestry automation management.The above research results analyzed and studied from the application perspective of dynamic model on image enhancement and image segmentation.Some breakthroughs have been made in theoretical analysis.It provides a new idea for image processing based on dynamical system,which has some theoretical significance and practical value.