Hongwei She

Feature extraction of hyperspectral images based on preserving neighborhood discriminant embedding

A novel manifold learning feature extraction approach—preserving neighborhood discriminant embedding (PNDE) of hyperspectral image is proposed in this paper. The local geometrical and discriminant structure of the data manifold can be accurately characterized by within-class neighboring graph and between-class neighboring graph. Unlike manifold learning, such as LLE, Isomap and LE, which cannot deal with new test samples and images larger than 70×70, the method here can process full scene hyperspectral images. Experiments results on hyperspectral datasets and real-word datasets show that the proposed method can efficiently reduce the dimensionality while maintaining high classification accuracy. In addition, only a small amount of training samples are needed.

Robust Registration of Remote Sensing Image Based on SURF and KCCA

The Speeded Up Robust Features (SURF) description’s success for computer vision applications makes it an attractive solution for image registration problem. For remote sensing images, SURF feature matching is often impacted by similar structures. To overcome the mentioned problem, we propose to use spatial relationship along with the SURF descriptor for remote sensing image registration. Firstly, a putative set of correspondences is obtained based on distances between SURF feature descriptors. Secondly, the spatial relationship of matched features is accomplished based on Kernel Canonical Correlation Analysis (KCCA), and then an influence function is established by the spatial relationship to figure out false matches. Introduction of spatial relationship to the SURF descriptors not only reduces the number of false matches but also help to keep the number of correct matches. Experimental results show an overall significant reduction of the mismatches while maintaining a high rate of correct matches.

Detection of the main stream of the Yellow River based on spectral feature and the dynamic transmission model

The problem of Yellow River main-stream detection with multi-spectral remote sensing images is investigated in this paper. Firstly, the flow characteristic of Yellow River was analyzed. The spectral similarity of the main-stream was discussed in succession. Then, based on the principle of spatial continuity, a main-stream dynamic transmission model was proposed. Finally, a main-stream detection approach called Main-stream Spectral Correlation Dynamic Transmission Approach (MSCDEA) was presented. The experiment indicates that the proposed algorithm is effective and can be used in practice.

Subspace divided semi-supervised SVM classification for hyperspectral images

Since the traditional SVM algorithm has high classification accuracy at the expense of huge training samples, A Subspace Divided Semi-Supervised Support Vector Machines (SDSS-SVM) classification method which need only one sample for each class is proposed. In the proposed method, a coarse classification result is obtained based on minimum distance clustering first. A General Sphere Criterion is introduced and applied to the coarse result, and the testing samples is divided into identified samples and unidentified samples. Then, the subspace division is accomplished according to the probable mixing. Samples which have the highest confidence in the subspace are selected as the training samples to subdivide the subspace of the unidentified samples to get the final classification. Classification experiment illustrates that the proposed approach can reach quite high classification accuracy.

Study on Fuzzy DTCC Method for River Main-Stream Interpretation from Remote Sensing Image

Yellow River is well known as sediment-laden river in the world with main-stream quickly change in its lower channel, so main-stream should be understood timely at flood season. Remote sensing is a primary approach to get main-stream information timely, however it has still some difficulties in interpreting it on image. DTCC (Dynamic Transmission Cross-Correlation) algorithm has been successfully described river flows with direction and continuity, as well as its surface phenomena having similarity alone flow direction. However, with water flow uncertainty similarity and the suboptimal path in flow direction, the DTCC interpretation results have some errors in some reaches. In this paper, a Fuzzy Dynamic Transmission Cross-correlation (FDTC) algorithm was proposed to interpret the river main-stream information based on fuzzy evaluation to flow similarity and fuzzy comprehensive selection in flow direction. The algorithm was applied on TM image in the lower of Yellow River and the results show that the FDTC method has an obviously improvement on main-stream interpretation.

An Image Enhancement Algorithm for River Main-Stream Based on Remote Sensing Data with Wavelet Transform

In the lower Yellow River, main-stream changes frequently for quickly change of channel sediment erosion and deposition, so it has a great significant to monitor its variation with remote sensing data during flood control decision-making. However, water spectral difference is weak and unsteady between main-stream area and other flow area in river, which leads main-streams feature unclear and difficult to be recognized on image and image enhancement becomes an essential step before doing main-stream detection. In this paper, a wavelet enchantment algorithm was proposed to prominent main-stream information through combing wavelet spatial-frequency characteristic and a spatial gradient operator. The gradient operator was mainly used to calculate wavelet coefficient gain matrix to construct the wavelet transform inverse-coefficients. The experiment results show that main-stream information was enhanced evidently as well as noise decreased, and the image preprocessing by proposed algorithm can meet the requirement for main-stream detection.

Realization of Catmull-Clark Subdivision Algorithm Based on Guadrilateral Network

Subdivision technology is becoming one of important trends of innovation in CAD/CAM modeling systems. Based on 7researches on subdivision surface theory, in this paper, a valid algorithm of generation for subdivision surface is proposed. The data structures for implementing subdivision surfaces are analyzed. The algorithm of subdivision process is explained in detail, and the matrices of generating and the rules of connecting about this subdivision algorithm are presented. Meanwhile, the continuity of subdivision is also discussed. Finally, the examples demonstrate explicit and efficiency of our method. Keywords-Catmull-Clark subdivision; subdivision rules; guadrilateral network; regular hexahedron