计算机遥感图像分类法在天然草原土地利用现状研究中的应用

本文以锡林郭勒草原国家级自然保护区为例,利用计算机遥感图像分类法对研究区2005年的TM遥感影像进行了解译,并对其进行了分类精度评估,得出(1)将该方法应用于景观单一、面积较大的草原地区土地利用分类中,具有较高的精确度,能够较好地反映土地利用类型空间分布特征(。2)利用该方法,可以在较短的时间内得到较准确的结论,具有很高的实用性。

一种简单加入空间关系的实用图像分类方法

遥感图像分类是遥感图像处理的一项基本内容 ,也是遥感应用中关键的一步。为了提高分类的精度 ,一方面是对光谱信息的合理利用 ;另一方面 ,可以加入新的信息源 ,即进行多源数据处理 ,并加入地学知识 ,尤其是对空间信息的利用是至关重要的。但是由于地学知识的复杂性及空间信息利用的难度以及数据源的限制 ,尚无公认的实用方法。该文提出了一种简单加入空间关系的分类方法 ,在没有其它数据源的情况下 ,利用空间关系特性 ,在分类中构造两个空间关系波段 ,实现空间约束 ,部分消除仅依赖光谱数据分类而引起的同物异谱和同谱异物造成的分类错误。简单实用 ,同时也验证了空间关系在分类中的重要性。

基于SPOT5遥感影像的珲春林业局森林分类研究

利用SPOT5卫星图像对研究区森林类型进行计算机分类,总体精度普遍不高,并且许多森林地类或林分类型的生产者精度和用户精度均比较低,林分类型与检验样本都存在较大偏差。在四种分类方法中,马氏分类法精度最高,总体精度达到57.13%,Kappa系数0.5208。

TWO IMPROVED GRAPH-THEORETICAL CLUSTERING ALGORITHMS

Graph-theoretical approaches have been widely used for data clustering and image segmentation recently. The goal of data clustering is to discover the underlying distribution and structural information of the given data, while image segmentation is to partition an image into several non-overlapping regions. Therefore, two popular graph-theoretical clustering methods are analyzed, including the directed tree based data clustering and the minimum spanning tree based image segmentation. There are two contributions： （1） To improve the directed tree based data clustering for image segmentation, （2） To improve the minimum spanning tree based image segmentation for data clustering. The extensive experiments using artificial and real-world data indicate that the improved directed tree based image segmentation can partition images well by preserving enough details, and the improved minimum spanning tree based data clustering can well cluster data in manifold structure.

Color image segmentation using mean shift and improved ant clustering

To improve the segmentation quality and efficiency of color image,a novel approach which combines the advantages of the mean shift(MS) segmentation and improved ant clustering method is proposed.The regions which can preserve the discontinuity characteristics of an image are segmented by MS algorithm,and then they are represented by a graph in which every region is represented by a node.In order to solve the graph partition problem,an improved ant clustering algorithm,called similarity carrying ant model(SCAM-ant),is proposed,in which a new similarity calculation method is given.Using SCAM-ant,the maximum number of items that each ant can carry will increase,the clustering time will be effectively reduced,and globally optimized clustering can also be realized.Because the graph is not based on the pixels of original image but on the segmentation result of MS algorithm,the computational complexity is greatly reduced.Experiments show that the proposed method can realize color image segmentation efficiently,and compared with the conventional methods based on the image pixels,it improves the image segmentation quality and the anti-interference ability.

Semi-supervised kernel FCM algorithm for remote sensing image classification

These problems of nonlinearity, fuzziness and few labeled data were rarely considered in traditional remote sensing image classification. A semi-supervised kernel fuzzy C-means （SSKFCM） algorithm is proposed to overcome these disadvantages of remote sensing image classification in this paper. The SSKFCM algorithm is achieved by introducing a kernel method and semi-supervised learning technique into the standard fuzzy C-means （FCM） algorithm. A set of Beijing-1 micro-satellite＇s multispectral images are adopted to be classified by several algorithms, such as FCM, kernel FCM （KFCM）, semi-supervised FCM （SSFCM） and SSKFCM. The classification results are estimated by corresponding indexes. The results indicate that the SSKFCM algorithm significantly improves the classification accuracy of remote sensing images compared with the others.

Video learning based image classification method for object recognition

Automatic image classification is the first step toward semantic understanding of an object in the computer vision area.The key challenge of problem for accurate object recognition is the ability to extract the robust features from various viewpoint images and rapidly calculate similarity between features in the image database or video stream.In order to solve these problems,an effective and rapid image classification method was presented for the object recognition based on the video learning technique.The optical-flow and RANSAC algorithm were used to acquire scene images from each video sequence.After the selection of scene images,the local maximum points on comer of object around local area were found using the Harris comer detection algorithm and the several attributes from local block around each feature point were calculated by using scale invariant feature transform （SIFT） for extracting local descriptor.Finally,the extracted local descriptor was learned to the three-dimensional pyramid match kernel.Experimental results show that our method can extract features in various multi-viewpoint images from query video and calculate a similarity between a query image and images in the database.

Classification of hyperspectral remote sensing images based on simulated annealing genetic algorithm and multiple instance learning

A hybrid feature selection and classification strategy was proposed based on the simulated annealing genetic algonthrn and multiple instance learning （MIL）. The band selection method was proposed from subspace decomposition, which combines the simulated annealing algorithm with the genetic algorithm in choosing different cross-over and mutation probabilities, as well as mutation individuals. Then MIL was combined with image segmentation, clustering and support vector machine algorithms to classify hyperspectral image. The experimental results show that this proposed method can get high classification accuracy of 93.13% at small training samples and the weaknesses of the conventional methods are overcome.

Classification Method Research to Remote Sensing Images

With rapid development of remote sensing technology, the resolution of remote sensing images is increasingly improved; then people can extract more useful data and information from these images. Thus, an important information extraction method from remote sensing images - image classification, becomes more and more important. Based on phenopthase and band composition characteristics, this paper firstly discusses the important role of background parameters in remote sensing images classification; then based on geographical infomation system technology, the computerized automatic classification to high-medium-low-yield croplands in Dingxiang County of Shanxi Province in rotate sensing images has been carried out by using eompound layers classification method of multi-thematic information; compared the classification result to the visual interpretation results, the accuracy increases from 70% to above 90%.

Research on the Natural Image Classification and Segmentation Algorithm based on GPU and Neural Network

In this paper, we conduct research on the natural image classification and segmentation algorithm based on GPU and neural network. The application of image segmentation is very broad, almost appeared in all areas related to image processing, and involved in various types. With the fast development of computing technology and integrated circuit technology, the renewal speed of graphics hardware. Our method combines the GPU with network to optimize the traditional image segmentation and classification methods which will be meaningful. In the future, we will focus our attention on the hardware deployment of the GPU to modify the current approach.

Discriminative Structured Dictionary Learning for Image Classification

In this paper, a discriminative structured dictionary learning algorithm is presented. To enhance the dictionary's discriminative power, the reconstruction error, classification error and inhomogeneous representation error are integrated into the objective function. The proposed approach learns a single structured dictionary and a linear classifier jointly. The learned dictionary encourages the samples from the same class to have similar sparse codes, and the samples from different classes to have dissimilar sparse codes. The solution to the objective function is achieved by employing a feature-sign search algorithm and Lagrange dual method. Experimental results on three public databases demonstrate that the proposed approach outperforms several recently proposed dictionary learning techniques for classification.

An Automated Approach to Passive Sonar Classification Using Binary Image Features

This paper proposes a new method for ship recognition and classification using sound produced and radiated underwater. To do so, a three-step procedure is proposed. First, the preprocessing operations are utilized to reduce noise effects and provide signal for feature extraction. Second, a binary image, made from frequency spectrum of signal segmentation, is formed to extract effective features. Third, a neural classifier is designed to classify the signals. Two approaches, the proposed method and the fractal-based method are compared and tested on real data. The comparative results indicated better recognition ability and more robust performance of the proposed method than the fractal-based method. Therefore, the proposed method could improve the recognition accuracy of underwater acoustic targets.

Research on High Resolution Satellite Image Classification Algorithm based on Convolution Neural Network

Artifi cial neural network is a kind of artificial intelligence method to simulate the function of human brain, and deep learning technology can establish a depth network model with hierarchical structure on the basis of artificial neural network. Deep learning brings new development direction to artificial neural network. Convolution neural network is a new artificial neural network method, which combines artificial neural network and deep learning technology, and this new neural network is widely used in many fields of computer vision. Modern image recognition algorithm requires classifi cation system to adapt to different types of tasks, and deep network and convolution neural network is a hot research topic in neural networks. According to the characteristics of satellite digital image, we use the convolution neural network to classify the image, which combines texture features with spectral features. The experimental results show that the convolution neural network algorithm can effectively classify the image.

A Novel Approach for Unsupervised Segmentation of Homogeneous Regions in Gray-scale Images

An improved approach for JSEG is presented for unsupervised segmentation of homogeneous regions in gray-scale images. Instead of intensity quantization, an automatic classification method based on scale space-based clustering is used for nonparametric clustering of image data set. Then EM algorithm with classification achieved by space-based classification scheme as initial data used to achieve Gaussian mixture modelling of image data set that is utilized for the calculation of soft J value. Original region growing algorithm is then used to segment the image based on the multiscale soft J-images. Experiments show that the new method can overcome the limitations of JSEG successfully.

Research on Image Segmentation Algorithm based on Fuzzy C-mean Clustering

This paper presents a fuzzy C- means clustering image segmentation algorithm based on particle swarm optimization, the method utilizes the strong search ability of particle swarm clustering search center. Because the search clustering center has small amount of calculation according to density, so it can greatly improve the calculation speed of fuzzy C- means algorithm. The experimental results show that, this method can make the fuzzy clustering to obviously improve the speed, so it can achieve fast image segmentation.

Research on Novel Natural Image Reconstruction and Representation Algorithm based on Clustering and Modified Neural Network

In this paper, we conduct research on the novel natural image reconstruction and representation algorithm based on clustenng and modified neural network. Image resolution enhancement is one of the earliest researches of single image interpolation. Although the traditional interpolation and method for single image amplification is effect, but did not provide more useful information. Our method combines the neural network and the clustering approach. The experiment shows that our method performs well and satisfactory.

High-resolution remote sensing mapping of global land water

Land water, one of the important components of land cover, is the indispensable and important basic information for climate change studies, ecological environment assessment, macro-control analysis, etc. This article describes the overall study on land water in the program of global land cover remote sensing mapping. Through collection and processing of Landsat TM/ETM+, China's HJ-1 satellite image, etc., the program achieves an effective overlay of global multi-spectral image of 30 m resolution for two base years, namely, 2000 and 2010, with the image rectification accuracy meeting the requirements of 1:200000 mapping and the error in registration of images for the two periods being controlled within 1 pixel. The indexes were designed and selected reasonably based on spectral features and geometric shapes of water on the scale of 30 m resolution, the water information was extracted in an elaborate way by combining a simple and easy operation through pixel-based classification method with a comprehensive utilization of various rules and knowledge through the object-oriented classification method, and finally the classification results were further optimized and improved by the human-computer interaction, thus realizing high-resolution remote sensing mapping of global water. The completed global land water data results, including Global Land 30-water 2000 and Global Land 30-water 2010, are the classification results featuring the highest resolution on a global scale, and the overall accuracy of self-assessment is 96%. These data are the important basic data for developing relevant studies, such as analyzing spatial distribution pattern of global land water, revealing regional difference, studying space-time fluctuation law, and diagnosing health of ecological environment.

Classification of hyperspectral remote sensing images using frequency spectrum similarity

An algorithm of hyperspectral remote sensing images classification is proposed based on the frequency spectrum of spectral signature.The spectral signature of each pixel in the hyperspectral image is taken as a discrete signal,and the frequency spectrum is obtained using discrete Fourier transform.The discrepancy of frequency spectrum between ground objects' spectral signatures is visible,thus the difference between frequency spectra of reference and target spectral signature is used to measure the spectral similarity.Canberra distance is introduced to increase the contribution from higher frequency components.Then,the number of harmonics involved in the proposed algorithm is determined after analyzing the frequency spectrum energy cumulative distribution function of ground object.In order to evaluate the performance of the proposed algorithm,two hyperspectral remote sensing images are adopted as experimental data.The proposed algorithm is compared with spectral angle mapper (SAM),spectral information divergence (SID) and Euclidean distance (ED) using the product accuracy,user accuracy,overall accuracy,average accuracy and Kappa coefficient.The results show that the proposed algorithm can be applied to hyperspectral image classification effectively.

Research on Auto-Classification Method of Remote Sensing Images in Mountainous Areas——An Application in Southwest of China

In mountainous areas, it is the undulant terrain, various types of geomorphic and land use that make the remote sensing images great metamorphism. Moreover, due to the elevation, there are many areas covered with shadow, clouds and snow that make the images more inaccurate. As a result, it would be very difficult to carry out auto-classification of RS images in these areas. The study took Southwest China as the case study area and the TM images, SPOT images as the basic information sources assisted by the auxiliary data of DEM, NDVl, topographical maps and soil maps to preprocess the images. After preprocessing by topographic correction and wiping off clouds, snow and shadows, all the image data were stacked together to form the images to be classified. Then, the research used segmentation technology and hierarchical method to extract the main types of land use in the area automatically. The results indicated that the qualitative accuracies of all types of land use extracted in Southwest China were above 90 percent, and the quantitative accuracies was above 86 percent. The goal of reducing workloads had been realized.

Efficient page layout analysis on small devices

Previously we have designed and implemented new image browsing facilities to support effective offiine image contents on mobile devices with limited capabilities： low bandwidth, small display, and slow processing. In this letter, we fulfill the automatic production of cartoon contents fitting small-screen display, and introduce a clustering method useful for various types of cartoon images as a prerequisite stage for preserving semantic meaning. The usage of neural networks is to properly cut the various forms of pages. Texture information that is useful for grayscale image segmentation gives us a good clue for page layout analysis using the multilayer perceptron （MLP） based x-y recursive algorithm. We also automatically frame the segment MLP using agglomerative segmentation. Our experimental results show that the combined approaches yield good results of segmentation for several cartoons.