Applied Mechanics and Materials Vols. 635-637

Abstract: This paper presents a machine vision inspection method for winding high frequency inductors, which affects the reliability and quality of the electronic products. This paper proposes how to quickly and correctly improve the quality of component detection, an important issue for surface mounted device (SMD) inductors manufacturers. SMD components easily damage the phenomenon of the electrode, and the brightness of the brightness of the damaged area of the electrode close to normal, not easy to be precise defect area separated from the electrode area.

Abstract: We present a new adaptive denosing method using compressive sensing (CS) and genetic algorithm (GA). We use Regularized Orthogonal Matching Pursuit (ROMP) to remove the noise of image. ROMP algorithm has the advantage of correct performance, stability and fast speed. In order to obtain the optimal denoising effect, we determine the values of the parameters of ROMP by GA. Experimental results show that the proposed method can remove the noise of image effectively. Compared with other traditional methods, the new method retains the most abundant edge information and important details of the image. Therefore, our method has optimal image quality and a good performance on PSNR.

Abstract: In order to classify the objects in nature images, a model with color constancy and principle component analysis network (PCANet) is proposed. The new color constancy model imitates the functional properties of the HVS from the retina to the double-opponent cells in V1. PCANet can be designed and learned extremely, which comprises only the very basic data processing components: cascaded principal component analysis (PCA), binary hashing, and block-wise histograms. At last, a SVM is trained to classify the object in the image. The results of experiments demonstrate the potential of the model for object classification in wild color images.

Abstract: This paper is about the object rotational motion estimation based on optical flow equation. Being a non-touch measuring technique, it is of important value in some special occasions. It is to set up rigid motion equations by optical flow character, and then using two-step iterative method to estimate motion equation and calculate rotation speed for each coordinate axis. We present a simplified calculation method for some object which specific structure parameters are known. The experiment results show the calculation is accurate.

Abstract: This study claims an algorithm of calibration which is executed on the basis of projection matrix. This algorithm directly estimates intrinsic parameter on the basis of rotation matrix’s unitary orthogonality combined with Cholesky decomposition from the obtained projection matrix. Then, false is excluded by rotation matrix’s determinant constraints, and ultimately, camera location and orientation matrix are obtained and estimated parameters are optimized with the minimum error of reprojection residual being cost function. This algorithm is taken under a pinhole camera model and can calibrate the camera from single view with variable focal length. Both simulation data and true image experiments have proved the feasibility and robustness of this algorithm.

Abstract: Texture Information is widely used as one of the main low-layer features in the content-based image retrieval. In general, when the retrieval is carried out in texture image space, the same description method is adopted to regular and irregular texture images. As a large amount of regular and irregular texture images existed in the image database, it is very difficult to describe every texture with the same description method. In this paper, a retrieval strategy for texture image is proposed. The proposed strategy is divided into steps: First, classify texture images by Wold decomposition into regular and irregular texture images, then describe and retrieve them by regular and irregular texture description separately. Experimental results have showed that proposed strategy can improve classification and retrieval precision.

Abstract: In this paper, a frame concealment scheme is proposed to combat the channel errors by using block-level and pixel-level motion vector extrapolation (MVE). Firstly, the proposed algorithm classifys the lost blocks into different types and then determines their corresponding MVs. Secondly, by using the pixel-level MV extrapolation, the candidate MV set of different lost pixels is obtained combined with the MVs of the blocks. Simulation results show this method is highly effective in sense of PSNR.

Abstract: The paper currency image recognition method based on Gabor filter set is discussed in this paper. According to the paper currency image features, the suitable parameters of Gabor filter set are selected for the extraction of paper currency characteristics, the multi-scale and multi-directional texture characteristics of paper currency image are gotten; then the texture images are meshed, and the row and column projection sum of grid pixels' average grey are calculated, finally, the template match method based on grid projection characteristics is used for paper currency recognition. Experiments show that, this method has strong anti-interference ability, it can raise the recognition rate of old or dirty paper currency greatly, and it costs little time.

Abstract: This paper presents fuzzy color histogram feature-based image retrieval method and texture spectrum fuzzy histogram feature analyzes the image database indexing techniques and the introduction of the experimental system for an improved method of fuzzy indexes. Algorithm reflects the underlying characteristics of high-level concepts and integration, relevance feedback and machine learning mechanism combining ideas. In this paper, the algorithm full processing power of computer systems, has a certain reference value and practical significance.

Abstract: In general, it is difficult to segment accurately image regions or boundary contours and represent them by feature vectors for shape-based image query. Therefore, the object similarity is often computed by their boundaries. Hausdorff distance is nonlinear for computing distance, it can be used to measure the similarity between two patterns of points of edge images. Classical Hausdorff measure need to express image as a feature matrix firstly, then calculate feature values or feature vectors, so it is time-consuming. Otherwise, it is difficult for part pattern matching when shadow and noise existed. In this paper, an algorithm that use Hausdorff distance on the image boundaries to measure similarity is proposed. Experimental result has showed that the proposed algorithm is robust.