Fast Registration and Homogenization for Multi-sites Laser Scanning Data of Grain Storage

摘要

In order to get accurate measurement results in the annual inventory checking of grain storages, it is needed to use 3D measuring technology to obtain the volume information of the grain storages. However, the multi-sites and low position requirement for sitting the laser scanner inside of the grain storage led to serious data redundancy in the local area of the scanner sites and uneven distribution of the range data. Considering the above redundancy and the uneven distribution problems, a range data redundancy reduction method based on adaptive re-sampling is presented at first. Then the range data should be segmented to detect the internal structure information and to extract the grain surface. As the basic structure of the grain storage is fixed, a fast automatic registration method based on the priori model is proposed to register the multi-sites range data. In view of the uneven distribution of range data and the occlusion holes, a multi-sites range data homogenization algorithm based on the competition mechanism is also proposed to merge the multi-sites range data and to fill the occlusion holes. Finally, the grain surface can be reconstructed by surface fitting method and the grain volume is estimated by the numerical integration method. Examples of tow common grain storage types, Flat Storage and Squat Silos, are given and discussed to show its effectiveness and accuracy.