The Potential of Automated Extraction of Objects from Airborne Laser Profiling Applied in Urban Cartographic Updating

Introduction: The possibilities of construction of extractions in urban areas, from data derived from the profiling of the airborne LASER is a reality, and has been the focus not only among scientific communities, but also in public bodies, as an alternative tool to support in the mapping update and control of land use. From the point cloud it is possible to generate images, where the results can help in the extraction of the edges of the buildings. The objective of this study is to develop an evaluation methodology in the automated extraction of buildings from image processing techniques originated in the LIDAR point cloud. To do this, we considered evaluating the tolerances according to the PEC-PCD and its class table. For image processing, IHS transformation and K-means classification were used, thus presenting possibilities for advances in the automated extraction of the edges of buildings with a higher degree of confidence. The evaluation of the accuracy obtained in the methodology, using data from vectorization and data obtained by topography, could serve as reliability parameters for new analyses of automated extraction of these objects. Objective: The objective is to develop an evaluation methodology in the automated extraction of 2D building vectors from images from the LiDAR point cloud, presenting low-cost techniques in the extraction and delimitation of its edges, with the use of image processing. Then, to evaluate the accuracy obtained by statistical analysis according to PEC-PCD, with analysis of possible systematic errors by Student's t-test. Method: In 2013, the City of Recife, PE, performed an aerial flight with LASER profiling of the area corresponding to the entire extension of the municipality, a total of 220 km2, obtaining images of intensity of the cloud of LiDAR points. These images visually discretize the objectives, such as: buildings, vegetations, etc. This made possible a study of the behavior of the point cloud at the edges of these objectives, thus making possible, through digital image processing techniques, a greater understanding of the reflectance behavior at these edges, with vector delimitations through polygons, lines and points. Vector acquisitions were made on the scale 1:1000 and the coordinates of the vertices of the polygons obtained were analyzed, comparing them with the coordinates obtained by topographic survey, thus allowing statistical analysis according to the PEC-PCD. The raster product chosen for the extraction of the characteristics was the composition image, originated from the fusion of the images of intensity and hypometric images. During the process of extracting the features, the influence of the edges of the objects on the interpretation of their faces was perceived, a factor that had to be investigated more cautiously during the work. It was observed that some edges of the raised object have thicker edges due to the effect of the edges and that this phenomenon is accentuated when large scales are used, which hinders the interpretation process in the vectorization. It was observed that these edges are the result of the projection in the cloud of three-dimensional points in the image plane. In the specific case of the buildings, which had a raised top, the LASER scanner obtained three-dimensional dots not only from their roof, but also from their vertical faces. The vectorized area corresponds to two buildings located within the Campus Joaquim Amazonas, UFPE, both belonging to the nanostructures laboratory of the Technological Center of Pernambuco (CTG). Results and Discussion: The polygonal data described in the determination of the topography coordinates were processed in the planimetric mode of the AstGeoTop software and obtained as angular closure error 8.8", linear closure error 0.0232 meters and relative linear error 1/8452.7184. The statistical analyses carried out to evaluate systematic errors between the coordinates obtained by topography and vectorization, through the Student's t tests and its bilateral distribution table, allowed to obtain a parameter value with constant t=1.895, with values of degree of freedom n-1=7 and α/2=0,5, parameters that are 90% reliable. From the results, it was observed that there is no trend in any direction, but there is a systematic error in the coordinate E (this one). By applying the Chi-square test, a quality compatible with Class C of the PEC-PCD was achieved, with a standard error of 0.50 meters, for the scale of 1:1000. Finally, the method used to observe the edges is general, since an unsupervised classification was used. It has also been observed that the direction of flight has an impact on the thickness of the generated edge shadow, which is not the case in other regions where the flight has taken place in other directions. However, this should be better evaluated, including in-depth behavioral studies with other models of optical-mechanical scanning mirrors. The results were obtained for unconjugated buildings located on flat ground. It is recommended to use this methodology for other types of land and with different occupation characteristics. Implications of the research: The research presents important results through products originating from the LiDAR System and in the generation of cartographic products in the extraction of the edges of buildings. Together with the precision and speed with which the acquisition of this data takes place, it becomes a technology with a potential tool in terms of territorial management. An evaluation of the extraction of contours of objects on the ground from data from an airborne LASER profile was presented as a reliable alternative in the mapping update of the occupation of urban areas. Originality/Value: Innovative and low cost for the vector mapping update of urban areas of municipalities.