Research Of Basic Theory On A New Track Surveying

The speed-up of trains brings great challenge to the maintenance and surveying of tracks. The advantages of digital close-range photogrammetry technique have promoted people to collect track images by digital cameral. Using image processing technique to meet the requirements of track inspecting has become an explorable method. In order to get the track video images with enough scope, high resolution ratio, and easy to survey, the accurate and large-field two-dimensional track images would be needed. However, due to the conditional limitation, single digital image can not have the effect of high resolution ratio and large enough scope. Therefore, the photos are taken consecutively to get the sequential images, and then the track photographical mosaic is made by image mosaic technology. When the track sequential images are actually taken by non-metric digital camera from close distance, there would be large nonlinear distortion in digital image. The distorted errors would make a great impact on the precision of track surveying. In the images taken by handheld or car digital cameral, the tilt angle of photographic axis is not perpendicular. Before the mosaic of consecutive subsequent images, the tilt rack images have to be rectified into the normal case photography images, which can meet the requirements of getting the normal case photography, large scope and surveying-convenient mosaic images.In order to choose the distortion correction model for the non-metric digital camera, an indoor simulate experiment is made to compare the effects corrected between spherical model algorithm and multinomial algorithm. The experimental results show that the corrected precision of spherical model algorithm is higher than multinomial algorithm. In order to solve the problem without control point in tilt track images, the vanishing point and orientation elements are calculated by the parallel lines existed in track images during geometric correction. And then the track images may be corrected based on geometric transformational model. The statistic precision is made by residual errors of corrected images through simulated experiment. The experimental results show that the precision corrected by this method can meet the requirement of the geometric correction of track images. The feature points may be extracted from the corrected sequence track images by SIFT operator, and then affine transformation model is used to match and joint images. The experiment precision of track mosaic image can achieve basically the requirements for surveying the track smoothness.Through the simulatal experiments combined with image data processing of the actual track digital images, it may be confirmed that the precision of surveying track can basically meet the requirements of the track smoothness by close-range photogrammetry. Therefore, the technique, which the sequence track images are taken by non-metric digital cameral for surveying the track smoothness, is feasible. It has higher efficiency and lower labor intensity than the traditional technique of surveying static track, and is a potential efficient technique in the real time track maintenance and surveying for high speed railways.