چكيده به لاتين
In order to decrease the effects of noise and vibration of railway vehicles and the environments during the operation of railway tracks, it is necessary to inspect the status of them frequently. One of the main tasks in inspecting railroad tracks is to measure rail corrugation. The rail corrugation occurs during operation due to wear as a quasi-sinusoidal irregularity that appears on the running surface of a rail. The vertical component of the wheel-rail contact dynamic force, as induced by rail corrugation is an important source to vibration. Also, according to the corrugation wavelength and railway vehicle speed, a dynamic excitation is generated which is an important source to noise. It is essential to utilize a system to measure and monitor the rail corrugation as one of the most important sources of noise and vibration in railway, permanently. In this thesis, a new measurement system based on image processing proposed to perform the corrugation measurement in a reliable manner, while traditional measurement systems are based on acceleration measurement or chord measurement and they are not reliable due to physical contact or need for knowing the characteristics of the track. The proposed laser triangulation system measures the rail corrugation based on image processing, precisely, and quickly. This approach is an efficient technique for high-speed and high-accurate rail corrugation measurement. Experimental results show the capability of the system for practical usages.
The accuracy of measurement for corrugation domain and wavelength depends on the measurement accuracy of the proposed method. The accuracy of a measurement system based on triangulation principle is mainly affected by extraction of laser stripe peak in the image captured by the camera. The distribution of light intensity of a laser stripe in the image follows the Gaussian pattern, but is accompanied by noise. While many noise-filtering algorithms and peak detection algorithms have been developed, the Wavelet transform has demonstrated that it is capable of performing both of these tasks efficiently and robustly, but at pixel-level. To achieve high accuracy and high precision needed for rail corrugation measurement, the sub-pixel accuracy is needed. Here, a method based on Wavelet transform has been presented for laser peak detection with sub-pixel accuracy through utilizing the capability of wavelet transform for pattern matching and using the wavelet transform coefficients for sub-pixel accuracy detection by computing the mean weight of coefficients. The experiments show that the proposed laser peak detection method is able to achieve two times better results than the present methods, especially in low signal-to-noise ratios. The prototype system for rail corrugation measurement has been designed and manufactured based on the proposed peak detection algorithm. The results of rail corrugation measurements show the ability of the proposed method for industrial use.
The proposed system for high accurate rail corrugation measurement should be accurately calibrated. In this thesis, the proposed rail corrugation measurement system utilized only a single 2D plane for calibration. This method is based on Zhang, which was used only to calibrate the camera previously. Now by developing the mentioned method, it has been utilized to calibrate the whole camera-laser system. In the proposed method, it is assumed that the camera is calibrated at first, and its intrinsic parameters are determined. But, at the same time that the image captured for camera calibration, the laser beams have been projected onto the calibration plane and the image of the checker-board plane with laser lines projected onto the plane is captured together. The calibration board can be placed in any situation and orientation of the space that is within the field of view of the camera and laser. By knowing the exact position of the points located on the corners of the squares on the checkerboard and finding their corresponding points in the image taken from the calibration plane, the location of the calibration board relative to the camera center is obtained. The intersection of laser planes with a planar checker-board is modeled as a line equation. In order to obtain the coordinates of the intersection points, the detected laser line in the image that has been extracted through the above laser peak detection method and a line detection algorithm, is projected to the calibration plane. Thus, the extracted points lay on both of the calibration board and laser plane. By placing the calibration board at different positions and repeating the above process, a set of lines is extracted which spans the laser plane. Therefore, by fitting a plane to this set of lines, the parameters of the laser plane are determined. So, the main objective of the accurate system calibration which is to find an efficient way to establish an accurate relation between 2D laser points on the image, and their 3D corresponding coordinates in the scene by use of the proposed algorithm is obtained.
