Applied Mechanics and Materials Vols. 411-414

Abstract: In order to improve the degree and real-time of the vehicle image detection, a background extraction method based on the probability mean value method and the background update based on the weighted coefficient method through divided area are proposed through the acquisition of real-time traffic information and processing of video images for intelligent transportation systems. Finally a prototype of background extraction and background update is got, and it achieves the detection of moving vehicles. The experimental results show that this method is simple, small amount of calculation and it has a good robustness; it can extract a good background image quickly and detect a complete shadow of vehicles. So this method can meet the requirements of real-time detection of multiple moving targets.

Abstract: This paper presents a fast local stereo algorithm that suitable to real time applications. Thanks to the techniques like Box-filtering, fixed-window-based stereo matching algorithms can be really fast, but perform not well in some areas, i.e. the repetitive pattern and low texture areas. In order to improve the reliability of fixed-window algorithm and keep the algorithms speed, the proposed approach can deal with the repetitive pattern and low texture areas at a small computational cost. Experimental results show that the proposed approach provides a big improvement in accuracy compare to fixed-window algorithm, and the speed of the algorithm is still fast.

Abstract: In order to achieve better image segmentation and evaluate the segmentation algorithm, a segmentation method based on 2-D maximum entropy and improved genetic algorithm is proposed in this paper, and the ultimate measurement accuracy criterion is adopted to evaluate the performance of the algorithm. The experimental results and the evaluation results show that segmentation results and performance of the proposed algorithm are both better than the segmentation method based on 2-D maximum entropy method and the standard genetic algorithm. The segmentation of the proposed algorithm is complete and spends less time; it is an effective method for image segmentation.

Abstract: Rician noise pollutes Magnetic Resonance Imaging (MRI) image and makes later work worse. This paper proposed an filter algorithm which comprehensive utilize Genetic Algorithm (GA) and Shearlet transform. Firstly, it performs a wavelet multi-scale decomposition of image; then, it builds target function in GA; thirdly, it uses the GA to optimal coefficients of Shearlet wavelet threshold value in different scale and different orientation; finally, we obtain the composite image by using inverse lifting wavelet transform. Experimental results show that, the new algorithm presented here is much effective in removing Rician noise.

Abstract: This paper selects the target tracking algorithm suitable for specific target environment: using Mean Shift algorithm based on space edge direction histogram at initialization, selecting tracking algorithm based on block when there is a shelter. On the basis of algorithm analysis and software experiment and studying of TI Company's TMS320DM642 DSP chip internal structure and development process, these two algorithms researched in this paper were transplanted to DSP platform and a series of optimization were been made to the algorithms codes after transplanted ,implementing target tracking and identified via DSP development board instead of PC.

Abstract: The paper takes license location method based on the low-freq image vertical boundary and region searching. doing the vertical first-order differential operation on the graph, taking the vertical boundary information out, Use the line scan method into license region searching; detect region information of the suitable license texture status-hopping, the algorithm located accurately and lose less operation time than traditionally which has been achieved at the goal for license location fast and accurately.

Abstract: In order to guide the robots for harvesting citrus fruits, method based on normalized RGB model and chromatic aberration map was developed to detect citrus fruits with shadow within tree canopy. Several images of natural citrus-grove scene were photoed, and the color properties of target objects were analyzed. A rule for segmenting citrus fruits from background was put forward by fusing the segmented results of the normalized R channel map and the chromatic aberration map of R and B channels. The results show that the fruits with shadow can be detected integrally using the proposed method.

Abstract: AUVs are usually equipped with video cameras to obtain the environment underwater information. Underwater images often suffer from effects such as diffusion, scatter and caustics. In order to improve the image quality and contrast, image restoration is need to be carried out before other image process. In this paper, a novel adaptive de-noising algorithm based on multi-wavelet transform was proposed in order to remove the Gaussian noise from the blurred underwater image. Firstly, the Gaussian noise deterioration of the image model was given. Secondly, the wavelet transform algorithm using Biorthogonal as a basic wavelet for underwater image decomposition and reconstruction was presented. Finally, Haar and Biorthogonal basic wavelet were chosen separately for adaptive de-noising algorithm for the blurred image restoration. By contrast with other filter methods, the experiment results verified its useful behaviors, and demonstrate that the raised de-noising approach can achieve fairly desired de-noising effectiveness for underwater image.

Abstract: A direct, efficient method for the problem of epipolar rectification in the uncalibrated casewas proposed. The method was based on minimizing a measure ofdistortion, by introducing anepipolar distance transform. The transform converted image intensity values to a relative locationinside a planar segment along the epipolar line, so it was robust to noises. The ratio of the distancesbetween two matching points in the epipolar lines was theoretically proved invariant to an affinetransformation for planar surfaces. To calculate the relative rotation between both cameras, thealgorithm was decomposed into three-steps to limit the distortion. Results show that the new measureis more appropriate for image rectification, and the three-step algorithm has obtained an accuracycomparable result both in estimation error and visual effect, especially when the initial epipolar linesare far from horizontal.

Abstract: Gaussian filtering algorithm has the defect that it will cause a blur at the image edges, therefore, an optional Gauss filter denoising method based on difference image fast fuzzy clustering is proposed. In this method, Gauss filtered image is firstly calculated, the difference image between the original image and the Gauss filtered image is acquired hereafter; and then fast FCM clustering of the Gauss filter image is carried out, the image histogram frequencies are taken as weighting coefficients of objective function when clustering, therefore the noise points of the original image are gotten; finally, optional Gauss filtering algorithm is applied to these noise points of the original image. Experiment results show that this method is fast and effective, its anti-disturbance performance is well, and it can effectively prevent edges from being blurred.