Image Correction And Image Edge Detection Technology In Detection Of Aluminum Flat Tube

Computer vision is a special simulation technology to study how to use the computer and related imaging equipment to constitute a system like human visual system.It is mainly through the image or video acquisition equipment such as digital cameras,industrial CCD camera, video camera, Collecting image information, then dealing with the particular computer program we can use the needed content information of image or the target information in the video to imitate the function of the human visual system.With the continuous research of computer vision technology and in-depth development,its application scope is more and more widely, like in the monitoring of industrial production, precision testing instruments, medical image analysis, military monitoring and it also has excellent performance in other fields. With the development of designing and manufacturing technology of image acquisition equipment, the collected images have a higher definition and rich color information. The improvement of intelligent algorithms and machine learning methods further promote the development of computer vision technology, These factors contribute to the development of computer vision technology. Then, Even though the research of computer vision theory and technology application has been developing since its origination. Compared with the human visual system, there is still a big gap between computer vision system and human visual system.This is mainly because of in the process of acquisition and subsequent processing of the image, the light source in the image will affect the processing result. In order to reduce the adverse effects of the scene illumination on the image acquisition, the true color information of the object surface is correctly identified to enhance the robustness of computer vision in the face of color features. In this paper, we mainly study the fusion of color constancy algorithm under the condition of a single illumination and image edge detection technology and its image texture features used in industrial aluminum flat tube production testing application.Because the present color constancy algorithm under the condition of single illumination are based on a particular assumption, This makes the estimation result of the algorithm is larger when processing the image with the assumption that the condition is not satisfied. So in this paper, the existing color constancy algorithm through new fusion framework(support vector regression, least squares support vector regression)getting Regression fitting of algorithm estimation results and the final results meet the requirements of the global optimal solution, improve the estimation performance and accuracy. The error indicators of the simulation result also further proved that the fusionestimation not only reduces the error using the algorithm, but also improve the robustness of its response to various types of scene images.In this paper, the least squares support vector regression(LS-SVR) algorithm is used as the fusion framework of the five candidate algorithms. In the experimental simulation of the fusion algorithm, compared to the other five candidate algorithms, the algorithm has smaller estimation error and higher robustness. But compared to the traditional support vector regression(SVR) algorithm, the algorithm can train all the samples as the support vectors, so the sparse feature is lost. Even under the conditions of the 1143 real images, it needs a lot of time to train the model, which is not conducive to the practical application in the future. In order to solve this problem, the model is improved and optimized, and the samples which are closely related to the color space and the surrounding images are selected as support vectors, But those who deviate from the sample group is be eliminated. The experimental results show that the computation efficiency can be greatly improved at the expense of the minimum estimation error.Image illumination estimation is an important part of computer vision, In the same,image detection technology is another important component of computer vision, It has a very high research value in the industrial production, especially in the precision instrument manufacturing industry. For example, in this paper, the crack detection of aluminum flat tube, which is difficult to detect completely even through the human visual system, Actual production through to heat pipe internal pressure to flow through flume, by observing the heat conduction pipe in the tank is effervescing to determine whether the segment cracks, this is a very expensive resource. Through the CCD camera lens to capture the heat conduction pipe in the area as the target area, The computer through the collected image processing, compared to give if there is a crack tip. In this paper, we use the image texture feature matching method and the edge detection method to detect the contour. The final experimental results show that the method can accurately detect the cracks in this section of aluminum flat tube by using the method of edge detection.Using the method of image texture feature, the need to spend a lot of time to statistics and training image features, and the accuracy is lower than the edge detection method.