Research Of Image Processing Technology Applied In Mechanical Engineering

With the rapid development of economic society and scientific technology, computer digital iamge processing and analysis technology obtained rapid ascension, and has been widely used in mechanical engineering.The infrastructure construction in our country is developing by leaps and bounds, thus, it is very important to ensure the reliability and safety of the engineering contruction projects. Therefore, in geotechnical engineering, the deformation measurement and analysis of partical materials such as soil and rock, has become a popular research topic at present. How to improve the deformation precision and accuracy is the research focus. At the same time, natural disasters, such as earthquake and debris flow, took place frequently. The simulation of its formation has important research value and practical significance to prevent disasters.Digital image based soil deformation measurement is a kind of full-field and non-contact analysis method. For pictures caputured by digital camera, the measurement mainly has three steps:camera calibraion, image processing and analysis, data visualization. The unique texture structure of soil makes it fit to use cross-correlation method to analyze images. To solve the problem of large computation of basic particle image velocimetry, cross-correlation method based on the fast Fourier transform was put forward, which is used to acquire pixel data of soil deformation, including displacement, velocity and strain. Then, the data accuracy is improved to sub-pixel by Gaussian fit and spurious data are filtered. Camera calibration aims at determining the correlation between the image and object coordinates system, which is used in the later real deformation measurement. The calibration methods include linear, second-order polynomial, and the new-proposed three-layer feed-forward neural network based calibration algorithms. This paper gives contrastive analysis of the three algorithms in performance and proves their effectiveness. A useful soil deformation measurement system was developed in this paper. First, acquire pixel deformation data by cross-correlation method based on fast Fourier transform, then, according to the network calibration model obtain real deformation data, finally, visualize the deformation process according to the data result, including graphs of vector field, contour and velocity vector quantilization. Use these data and graphs to analyze the soil deformation characteristics. Besides, this system allows user to select region of interest (ROI) to decrease computation, and batch processing is added to make video for the whole deformation. Experiments show that this system could acquire deformation data from soil images accurately, efficiently and continuously, which provides a strong proof that image processing technology has practical significance and application value in the research field of geotechnical engineering.To simulate the formation of the natural disasters is essentially to simulate the dynamic process of particles and walls. Since people require higher vision, OPENGL library, as the main three-dimensional visualization and graphics rendering tools, it has been effectively applied in the simulation of dynamic model. The three-dimentional interactive demonstration by OPENGL makes the motor behavior of particles more realistic.