Abstract:
Survey of roads for cracks, holes and any other distress with computerized or robotic imagers are completed on even basis. It is more obvious that moving plate form that capture images produces fluctuations in the image stabilization. Thus, measurements will in fact leads to inaccuracy. This work introduces a complete new approach towards the stabilization of 3D asphalt images. In Kalman filter affine transformation is used in the state space model. Here shaking in the image platform is demonstrated as virtual simulator to estimate the significant translation in the image and correct it via Kalman filter. Later a more robust and dynamic approach with practical implementation has been used with the Extended Kalman filter. A Simulink model is designed for Kalman filter and MATLAB code is written for Extended Kalman filter to demonstrate and to actually implement the technique for the stabilization of images. This work backs to previous effort on the metrology of asphalt images by means of Kinect. Vibration effects and displacement in different potholes with variable ranges has been studied. The Kalman filter is use for pure translation while EKF is use for translation as well as rotation purpose. A substantial difference between Kalman filter and EKF estimation can be seen through error.
