Applied Mechanics and Materials Vols. 284-287

Abstract: Most patterns in continuous video sequences are similar. Temporal distortion, e.g. frames dropping, insertions, transposition, is a challenging issue for video reconstruction to find the actual missing positions in video sequences. The aim of this paper is to raise the detection accuracy and synchronize video frames back to original positions following temporal synchronization distortions. The successive video frames have similar statistics but the statistics in some local regions may differ from one another. Therefore, the block partition is partitioned into non-overlapping blocks by each frame, and then the local variance is calculated and taken as the block feature in each block. For most of the video frames, the pixels within the frame blocks are correlated and the maximum eigenvalue will be far from other eigenvalues. In this case, the maximum eigenvalue is set as the dominated block feature. For less correlated blocks, the values of the eigenvalues will be a little closer. In this case, the mean value of the eigenvalues represents the dominated block feature. Then, the sum of variance is regarded as the frame feature to calculate from these selective dominated blocks. Simulation results show the proposed methods are robust in evaluating the missing positions against temporal distortions.

Abstract: This paper introduces a kind of image process, moving-window method, for enhancing the consistency in searching for corresponding structured lines on images captured from different views. Effectiveness of the approach is demonstrated by a comparison of correctness with three of the most popular algorithms, the Multiple-line, the Scale Invariant Feature Transform (SIFT), and the Phase-Only Correlation method (POC).

Abstract: Reversible data hiding has drawn lots of interest in the last a few years. With reversibility, original media can be recovered without any distortion from the marked media after the embedded data has been extracted. In this paper, we present a new scheme which utilizes the wavelet transform and better exploited large variance of wavelet coefficient differences to achieve high capacity and imperceptibility. With the particularity of minor changes in the wavelet coefficients after embedding data, low visual distortion can therefore be obtained in the marked image. Furthermore, an extraordinary attribute of our scheme is that the use of embedding level differs greatly from previous schemes. Experimental results showed that the performance our scheme outperforms the state-of-the-art reversible data hiding schemes.

Abstract: It is the research hotspot for evolutionary algorithms to solve the contradiction between exploration and exploitation. Cellular artificial bee colony (CABC) algorithm is proposed by combining cellular automata with artificial bee colony algorithm from the perspective of the neighborhood in this paper. Each bee in the population structure defined in CABC has a fixed position and can only interact with bees in its neighborhood. The overlap between neighborhoods of different bees may make a bee an employed bee in one neighborhood and an onlooker bee in another neighborhood and vice versa, which increases the diversity of the population. The neighborhood and evolutionary rule help to control the selection pressure effectively, and the improved search mechanism in artificial bee colony algorithm is proposed to enhance the local search ability. The experimental results tested on four benchmark functions show that CABC can further balance the relationship between exploration and exploitation when compared with three ABC-based algorithms.

Abstract: In this paper, we present a study of evolving artificial neural network controllers for autonomously playing maze-based video game. A system using multi-objective evolutionary algorithm is developed, which is called as Pareto Archived Evolution Strategy Neural Network (PAESNet), with the attempt to find a set of Pareto optimal solutions by simultaneously optimizing two conflicting objectives. The experiments are designed to address two research aims investigating: (1) evolving weights (including biases) of the connections between the neurons and structure of the network through multi-objective evolutionary algorithm in order to reduce its runtime operation and complexity, (2) improving the generalization ability of the networks by using neural network ensemble model. A comparative analysis between the single network model as the baseline system and the model built based on the neural ensemble are presented. The evidence from this study suggests that Pareto multi-objective paradigm and neural network ensembles can be effective for creating and controlling the behaviors of video game characters.

Abstract: This paper presents a reliable and robust palmprint verification approach using palmprint feature point number (FPN). The feature verified by support vector machine (SVM). It has the advantages of capturing palm images in peg-less scenarios and by a low cost and low-resolution (100dpi) digital scanner. The low-resolution images lead a less database size. There are 4800 palmprint images were collected from 160 persons to verify the validity of the proposed approach and the results are satisfactory with 98.30% classification correct rate (CCR). Experimental results demonstrate that the proposed approach is feasible and effective in palmprint verification. Our findings will help to extend palmprint verification technologies to security access control systems.

Abstract: This paper presents a method for detecting feature points from an image and locating their matching correspondence points across images. The proposed method leverages a novel rapid LBP feature point detection to filter out texture-less SURF feature points. The detected feature points, also known as Non-Uniform SURF feature points, are used to match corresponding feature points from other frame images to reliably locate positions of moving objects. The proposed method consists of two processing modules: Feature Point Extraction (FPE) and Feature Point Mapping (FPM). First, FPE extracts salient feature points with Feature Transform and Feature Point Detection. FPM is then applied to generate motion vectors of each feature point with Feature Descriptor and Feature Point Matching. Experiments are conducted on both artificial template patterns and real scenes captured from moving camera at different speed settings. Experimental results show that the proposed method outperforms the commonly-used SURF feature point detection and matching approach.

Abstract: Regensburger and Scherzer described a symbolic computation method for moments and filter coefficients of scaling functions and obtained parametrizing compactly supported orthonormal wavelets. Following the idea, we are devoted to a study moments and parameterization construction for 3-band biorthogonal scaling coefficients with several vanishing moments. Firstly, we investigate the relations between filter lengths and symmetry. Then, we prove the relationship between dual continuous moments of 3-band biorthogonal scaling functions in theorem 2. This theorem reveals that the sum of continuous moments of dual scaling functions and is completely determined by the lower discrete moments. And we show the fact that the odd-indexed discrete moments are determined by the smaller even-indexed discrete moments. Finally, a family 3-band biorthogonal scaling coefficients with discrete moments as parameters are explicitly expressed based on computer algebra.

Abstract: This study addresses a non-supervised approach to extract TV programs via repetition based detection of the Inter-Programs (IPs) and audio based segmentation and classification algorithm to analyze the massive raw TV stream. Acoustic and visual information are both adopted for IPs detection so as to avoid missing true-positive. Novel audio fingerprints scheme and shot based indexing algorithm are introduced to guarantee the efficient and superior detection performance. After the TV programs are further segmented into clips, Gaussian Mixture Models (GMMs) are used to classify the clips into three types, namely, pure speech, non-pure speech, and non-speech. Experiments on a test dataset composed of more than 500 hours content-unknown TV streams show that the F-measure of the programs extraction and content analysis achieve 0.986 and 0.887 respectively. The experiments also demonstrate that the proposed algorithm for detecting repeated IPs outperforms the state-of-art approach.

Abstract: Lane detection is the key technology of the intelligent vehicle based on machine vision. In order to improve the detection of real-time, a lane detection and prediction algorithm based on Randomized Hough Transform is developed in this paper. The algorithm includes lane detection algorithm and prediction algorithm. First of all at identification stages, scan the pretreated image in order to search lanes candidate points, and combine with the lanes angle range of constraints, fit the candidate boundary points by Randomized Hough Transform for the improvement of real-time and robustness. The driveway line prediction algorithm is also proposed. With the dynamic searching window, weights of prediction are adaptive and they can make the line prediction more accurate. Test results show that algorithm has good real-time and robust performance.