Applied Mechanics and Materials Vols. 263-266

Abstract: The objective of this paper is to develop methodology that can recognize the Lanna handwritten character on historical documents by using character feature extraction technique. Historical documents are national treasures. Insignificant effort has been made to preserve Lanna historical documents. Other nations such as Egypt, China and Greece are investing a large effort in restoring and preserving their national historical documents. As a starting point, the focus is on using one Lanna historical document for performing experiments and verifying recognition methods available in this research area. The proposed system consists of three modules, which are image preprocessing module, feature extraction module and character recognition module. The details of each module are following: first, the input image is transformed into a suitable image for feature extraction module. Second, the proposed system extracts character features from the image. Finally, the extracted character information, which is kept in form of bit string, is calculated a similarity value for recognition result. The experiment was conducted on more than 4,000 Lanna handwritten characters by using 10-fold cross-validation classification method which is using 3,600 for training characters and 400 for testing character. The cross-validation process is repeated 10 times, with each of the 10 subsets used exactly once as the validation data. The precision of the proposed system is around 89.73 percent.

Abstract: This paper presents a new method for classification of remote sensing image based on multiple classifiers combination. In this method, three supervised classifications such as Mahalanobis Distance, Maximum Likelihood and SVM are selected to sever as the sub-classifications. The simple vote classification, maximum probability category method and fuzzy integral method are combined together according to certain rules. And adopted color infrared aerial images of Huairen country as the experimental object. The results show that the overall classification accuracy was improved by 12% and Kappa coefficient was increased by 0.12 compared with SVM classification which has the highest accuracy in single sub-classifications.

Abstract: In order to improve the efficiency for phased array radar's ESM, an ACO and SVM conjoint method is used in this paper to solve the problem of phased array radar signal recognition. By introducing ACO to supervise SVM parametric selection, the method is able to quickly discover seemly parameter value and improve SVM separation efficiency. Experimental results show that textual algorithm possess upper exactness rate to phased array radar that the whole pulse signals sorting can be identified. With normal-SVM and RST-SVM means to compare, the algorithm SVM parameter access time is short, thereby shorten the monolithic hour.

Abstract: We propose a new design to detect a target degraded by non-uniform illumination function and additive noise placed in non-overlapping background noise. The method is based on estimated illumination function and Wiener filtering theory, which provides robustness to non-uniform illumination and noisy conditions, especially for non-overlapping background noise. Computer simulation results are presented to verify the performance of the method.

Abstract: Due to multipath noise pollution, SAS image consists of two parts : target and noise .They can be described by K + K mixture distribution . How to separate noise data which obey K distribution from the target which also obeys K is a hot topic in SAS image field. This paper used the minimum error probability Bayes classifier to solve this problem, and achieved good results. At the same time, this paper also studied the factors that affect the classification results, such as the absolute value difference of training sample parameters and K distribution parameters.

Abstract: A statistical method of particle size distribution is researched in order to measure the mineral blasting effect. Aimed at resolving the problems of sensitivity to noise and over-segmentation existing in the traditional watershed algorithm(TWA), a improved watershed segmentation scheme based on morphological reconstruction and mark extraction is presented. Firstly, two-stage grayscale morphological reconstruction (TGMR)is employed to filter the image. Secondly, the watershed seed area and the watershed line area are pre-calibrated in the gradient image. Finally, the watershed transformation(WT) is conducted so as to further eliminate over-segmentation. Simulations show that the method can effectively remove the noise jamming and restrain the over-segmentation. It is also able to accurately locate the ore region contour in order to improve the segmentation results.

Abstract: Bridge recognition algorithm based on straight-line characteristic is proposed in order to automatically recognize bridge from aerial images, which includes the steps of edge detection, straight-line extraction, coarse location for bridge, accurate location for bridge. Meanwhile, realize the fast accurate location for bridge area by modified 8-neighborhood connectivity processing. The experiment result shows the reliability and efficiency of the method proposed in this article.

Abstract: Crowd abnormal behavior recognition is essential for intelligent visual surveillance in public places to ensure the safety of the public. This is a challenging work because crowd behaviors are complex which are influenced by various factors. This paper divided these factors into three categories: physical factors, social factors and psychological factors. Then an overview about crowd behavior modeling approaches was given. After that, the paper described and analyzed some influential existing algorithms in crowd abnormal behavior recognition from the view point of behavioral factors they used. Finally, the paper discussed the future research directions in this area and some research proposals were given.

Abstract: To efficiently find hidden clusters in datasets with complex distributed data，inspired by complementary strategies, a hybrid genetic clustering algorithm was developed, which is on the basis of the geodesic distance metric, and combined with the Fuzzy C-Means clustering (FCM) algorithm. First, instead of using Euclidean distance，the new approach employs geodesic distance based dissimilarity metric during all fitness evaluation. And then, with the help of FCM clustering, some sub-clusters with spherical distribution are partitioned effectively. Next, a genetic algorithm based clustering using geodesic distance metric, named GCGD, is adopted to cluster the clustering centers obtained from FCM clustering. Finally, the final results are acquired based on above two clustering results. Experimental results on eight benchmark datasets clustering questions show the effectiveness of the algorithm as a clustering technique. Compared with conventional GCGD, the hybrid clustering can decrease the computational time obviously, while retaining high clustering correct ratio.

Abstract: We present a novel dynamic fuzzy sets (DFS) method, which is the generalization of fuzzy sets (FS) and the dynamization of interval-valued intuitionistic fuzzy sets (IVIFS). First, by analyzing the degree of hesitancy, we propose a DFS model from IVIFS. Second, we introduce the distance measure of DFS. Finally, a pattern recognition example is given to demonstrate the application of DFS, and the experimental results show that the DFS method is more effective than some IVIFS methods.