Realtime Pavement Roughness Measurement Via Weakly Calibrated Laser Triangulation Vision

With the rapid development of highway construction, pavement detection and maintenance management have become the important task in the field of the highway construction in our country. Pavement roughness is one of the most important parameters for evaluating pavement quality, which has a direct impact on highway service life and traffic safety. There are the existing pavement roughness detection methods, such as straight edge measuring methods, which has some disadvantages of measuring manually, measureing slowly and large errors. With the development of laser measurement technology and electronic technology, vehicle-mounted laser roughness meter based on displacement sensors or acceleration sensors has emergered. However, due to sparse measurement points,they can not reflect completely the profile of pavement wave conditions. Moreover, the related equipments are expensive. Structured light measurement is a kind of non-contact measurement technology based on optical triangulation measurement principle. which is gradually used in the pavement roughness measurement. But the current system calibration is complex and application cost is very large.To solve these problems, the paper carried out two aspects of research work. Firstly, we found the weak correlation of profile roughness and system calibration parameters under the constraint of pavement roughness national standard calculation method, solving system calibration complex problems. In this paper, a robust laser structured light automatical recognition algorithm is proposed, which is foucus on solving the critical technical issues,such as the edge maintaining of image denoising, pavement background interference suppression and laser structured light stripe extraction and fitting. Simulation analysis and experiments results demonstrate the effectiveness and robustness of the proposed method. It incudes:(1) It can achieve the accurate measurement of pavement roughness based on the pixel shift amount of laser structured light strip in the image and a system calibration parameter.Compared with the prior method, the presented algorithm has advantages of simple systemcalibration, low-cost, install and use simply under the conditions of the same measurement accuracy.(2) An anisotropic diffusion PM filter based on iterations and timestep constraints is proposed. It can effectivelly maintain the laser structured light strip minutiae charectors in the image denoising process. The iterations and timestep constraints can ensure the real-time processing performance.(3) According to the unique tubular gray distribution feature of laser structured light strip, Gaussian second derivative filter and edge suppression filter are proposed to suppress effectively complex interference factors, such as pavement vehicle shadows and landmarks that affect seriously the recognition accuracy of laser light strip. The laser light curve fitting is performed by using three-moment method to construct cubic spline function and derivative continuous and boundary conditions as constraints.