Research On Rapid Measurement Technology For Grain Storage Volume Based On Three-dimensional Laser Scanning

Inventory checking is one of the most significant sectors for grain reserves, and it plays an important role on the macro-control of food, food security, etc. Today the annual inventory checking of grain storages spends a great deal of manpower and material resources, but the accuracy of measurement results is low and the working efficiency is not high. Considering the needs of the national science and technology support project, this paper focuses on the research of grain storage rapid modeling and volume measurement based on three-dimensional laser scanning technology.The achievement and innovation of this thesis are listed bellow:1. In order to solve the local data redundancy problem brought about by the low scan position in the limit interior grain storage, this paper gives a targeted adaptive re-sampling method. The method combines with the MLS down-sampling method, according to the local sampling rate, the local curvature characteristics, chooses a reasonable down-sampling rate to reduce data redundancy and to remove some noise.2. Against the Segmentation problem of the grain surface and the construction, this paper designs and develops an automatic segmentation method based on two-dimensional grid. This method projects the range data into two-dimensional uniform grid, and then according to the characteristics that grain surface has small fluctuation, building has large height difference and mixed region will appear discontinuous, segments the grain surface and non-grain surface region.3. Considering the requirements of multi-view fast registration and rapid modeling, this paper proposed a model-based fast automatic multi-sites registration algorithm by introducing prior information about the scanned scene into the process of registration and modeling. This method establishes the priori scene model firstly, and then aligns every range data to the model. So the method avoids the rigid requirements of overlapped area for pair-wise registration and the cumulative effect of errors, at the same time, the method solves the high memory, large computation problems in the multi-view registration, multi-view fast registration and rapid modeling is done.4. Non-uniform sampling and hole filling problem is one of the key issues involved modeling accuracy, this paper proposes a data homogenization method based on the competition mechanism. This method selects the best description points in dealing with the integration of overlapped regions through competitive mechanisms; for the false hole region, according to data quality of the complement data and the surrounding region data, chooses the best description; for the real hole region, uses the processed surrounding region data with high quality to repair the hole region; at the same time the uniform re-sampling is done.5. The design and development about the related software has been done. The demonstration work was applied in Qingyundian and Zhuozhou grain storage. The modeling quality and the volume estimation accuracy of the grain storage were evaluated.