Columnar Objects Change Detection Based On Point Cloud Data

The three-dimensional laser scanning technology has the advantages of faster data acquisition and accuracy in satisfying map construction than the traditional single-station data acquisition technology such as a total station.It is very suitable for the collection of three-dimensional data in the urban street environment to serve the Incremental update of the city map and fine urban management,etc.The existing construction of three-dimensional maps of cities mainly depends on the combination of on-board mapping systems and on-board mobile mapping systems.However,the high complexity of the urban street environment has also brought about many inconveniences in the data collection and processing of on-board mobile mapping systems.The quality of data collected by the vehicle-based mobile mapping system will be influenced by many factors in the street environment,seasonal changes in green vegetation near the ground,ground-based stationary objects(walkway trees,newsagents,temporarily parked vehicles),and liquidity objects(factors such as pedestrians,vehicles driving on motorways),and highly transmissive surfaces(such as glass)can cause data loss,which is a great obstacle to map construction.In addition,the variability of the width of urban roads has directly led to the general mobile scanning vehicles can not enter the narrow streets for data collection,which is one of the reasons for the lack of data.Moreover,the rapid development of the city requires more frequent updates of the city's map data,and simply relying on the mobile mapping system for data collection is relatively costly.The three-dimensional reconstruction of two-dimensional images belongs to the category of photogrammetry.It uses consumer digital cameras as data acquisition equipment to acquire several images of regions of interest with sufficient overlap,and uses computer vision technology to acquire three-dimensional data.It is compact and easy to operate.It also has the advantage of low cost.The thesis mainly did the following tasks:(1)Taking the cylindrical target in front of the college building as an experimental object,two kinds of change detection methods are proposed--the overall change detection based on the point cloud and the change detection based on GIS objects and image point cloud data.Both methods use image point cloud data.Using a consumer digital camera to capture image data of the experimental site,image point cloud data is obtained using SFM(Moving Structure)+ MVS(Multi-View Stereo Geometry)image 3D reconstruction technology.This process does not require camera calibration preprocessing and is fully automated.The dense 3D point cloud data with a local coordinate system is output.Using the acquired feature point coordinate data,the dense point cloud is also transformed into a unified coordinate frame to prepare for change detection processing.(2)Aiming at the problems that existing data acquisition technologies are prone to data vulnerabilities and high costs,an octree-based point cloud overall change detection method is proposed.The scanner acquires high-precision point cloud data as reference data,and the image point cloud data is used as data to be tested.The method constructs an octree structure for the reference data and the data to be detected,compares the homologous units in the octree structure of the two-stage point cloud data with each other,and denoises and divides the obtained initial detection results,and finally Successfully detected the true columnar target change.Although this method has poor real-time performance and is cumbersome to handle,it has high accuracy and can be used in the incremental update application of three-dimensional maps in cities.(3)In order to improve the efficiency of change detection,a method based on GIS objects and image point cloud data change detection was proposed.The GIS object in the complete GIS database was used as the benchmark data,and the image point cloud data was used as the data to be tested.In the first method,the amount of data that needs to be processed is almost halved.The obtained image point cloud data is evenly sliced according to the height,and then the orthographic data of different heights of each GIS object is orthographically projected,and then transformed into the coordinate system of the GIS object as the data to be detected.Based on the KD-tree algorithm,the neighboring points of the GIS object on each slice are searched to determine whether the object has changed.The change detection method based on GIS objects reduces dimensional data to two-dimensional space,reduces computational costs,and achieves rapid and efficient change detection.However,the dimensionality reduction operation results in the loss of information and cannot analyze the characteristics of the detection results.The fast and efficient features are more suitable for the application of urban facilities management.