Applied Mechanics and Materials Vols. 220-223

Abstract: In this paper, we present a new approach by local gray level difference based competitive fuzzy edge detection. In the light of human visual perception, a preprocessing step is proposed to simplify original images and further enhance the performance of edge extraction. Then we define the feature vector of each pixel in four directions and six edge prototype. Finally, BP neural network is used to classify the type of edge, and the competitive rule is adopted to thin the thick edge image. From the experimental result, it can be seen that the edge detection method proposed in this paper is superior to Canny method and Log method under the noisy condition.

Abstract: A novel transition region extraction and thresholding method based on both frequency and degree of gray level changes is proposed by analyzing properties of transition region. Frequent gray level based transition region extraction methods are greatly affected by noise. To eliminate the algorithm limitation, a modified descriptor taking both degree and frequency of gray level changes into account is developed. The proposed algorithm can accurately extract transition region of an image and get ideal segmentation result. The experimental results show its superiority and feasibility.

Abstract: In recent years, the gray-scale thresholding segmentation has emerged as a primary tool for image segmentation. However, the application of segmentation algorithms to an image is often disappointing. Based on the characteristics analysis of infrared image, this paper develops several gray-scale thresholding segmentation methods capable of automatic segmentation in regions of pedestrians of infrared image. The approaches of gray-scale thresholding segmentation method are described. Then the experimental system is established by using the infrared CCD device for pedestrian image detection. The image segmentation results generated by the algorithm in the experiment demonstrate that the Otsu thresholding segmentation method has achieved a kind of algorithm on automatic detection and segmentation of infrared image information in regions of interest of image.

Abstract: This paper presents a method of using Genetic Algorithm (GA) to optimize template and image searching process, using template matching to recognize target. An initial matching template is set manually according to 2D shape and the optimizing template is obtained by GA optimizing to meet the requirement of real-time and effective performance. Then the pixel position is encoded into genes, template correlation degree function works as fitness function to do GA search to recognize the target. The relating image process experiments show that this method has good real-time and robustness performance.

Abstract: In order to improve the measurement accuracy of machine vision, this paper focuses on the effect of calibration grid size and position on machine vision measurement accuracy by analysing the measurement error of the points inside and outside of the grid. The experimental results show that the measurement accuracy of the internal points is higher than that of the external points. The measurement errors increase firstly, then decrease, increase, and finally decrease which measuring point from the edge to the opposite edge in the calibration grid. While measurement errors of outside points increases with the increasing distance to the corner point. If the center of calibration grid coincides with the center of calibration board, measurement accuracy is high. However, if the center of calibration grid doesn't coincide with the center of calibration board, measurement accuracy is low. This results may provide direct means for the application of machine vision system in engineering.

Abstract: The rule of dark channel prior has made significant effect in outdoor image dehazing. The camera on air duct cleaning robots will capture foggy pictures when they are working because of raised dust, these foggy pictures have serious impact on the robot cleaning work. According to the characteristics of pictures in air ducts, we remove haze of these images based on dark channel prior, experimental results show that this method has good effect.

Abstract: As the rapid progress and development of DSP technology, a new way to solve the video signal processing was accepted. At present, binocular measurement instrument are mostly based on image capture card and PC on the market, expensive and bulky. For this problem, binocular vision measurement system based on DSP has been used. The design of elementary program structure about video processing which based on DSP/BIOS was finished and verified for its real-time. The results of the experiment show that the driver can be well used in the system.

Abstract: An adaptive underwater dam image crack edge detection algorithm, which was based on multi-structures and multi-scale elements, was presented for the deficiencies of regular and single structure elements edge detection. Firstly, six representative structure elements were constructed and the structure elements were expanded by multi-scale analysis method. Then, multi-structure and multi-scale elements were used to detect crack edge separately. The adaptive weighted coefficients were determined by the edge information entropy. Finally, the image edges can be obtained by the synthetic weighted existing edges method. The experimental results demonstrate that the proposed algorithm not only can remove the noise effectively but also maintain the image edge detail well.

Abstract: The calculation model of binocular vision sensor is constructed based on the detection principle of laser triangle geometrical relationship in the circumstances of the measured object non-static when measuring the vision sensor, the relationship between the measured object displacement and the CCD image point displacement. Measured point is researched and two-dimensional space coordinates of the measured point is calculated. Analyzed the relationship between displacement and structure parameters and simulated by matlab, optimized the optical parameters, in order to complete the measurement task better.

Abstract: Appearance features are important for tracking persons in stationary scenes. The proposed algorithm was based on appearance mold built by attributed relational graph (ARG). The ARG was used for modeling human appearance features containing color and spatial information. The matching degree of the ARGs was utilized for analyzing tracking situations in current frame. For tracking persons under occlusion, multiple feature patches were generated and the genetic algorithm was used for finding optimal labels of patches. Experiments showed the utility and performance of the proposed approach.