FROM BIG DATA TO ESSENTIAL DATA
SMALL DRONES, OFF THE SHELF CAMERAS, STRUCTURE FROM MOTION (SfM) AND VIRTUAL SURVEYING FOR EFFICIENT, ACCURATE AND SAFE ROAD SURVEYING:
Affordable tools (drones, off the shelf high resolution cameras and structure from motion (SfM) software) are now lined up for efficient production of high quality BIG data – aerial images, point clouds and surface models can be delivered in very short times. But they constitute big, suffocating data for most standard engineering design applications such as CAD. The challenge is to efficiently extract the essentials from big data so that the engineers can get on with their work.
Figure 1 below shows an ortho photo of approximately 1700 foot of roadway, on average about 60 foot wide. This ortho photo was produced with Agisoft Photoscan Professional (an affordable Structure from Motion software package) from 382 aerial images collectively taking up about 3.7GB of disk space. Point cloud, surface geometry and texture together make up an additional 2.8GB. The Ortho Photo with GSD 0f 0.04’ in geotiff format occupies 127MB while the digital surface model, also in geotiff format, with 0.08’ grid size takes up 265MB. All of the 6.5GB of data was rolled out in a matter of less than 24 hours. But in this form the data is too big for practical use in most standard CAD packages.
Figure 2 below represents an oblique big data view of a section of the road.
Figure 2: Oblique view of big data, ready for analysis
Figure 3 shows the same view with features of interest vectorized and tightly draped over the 3D terrain.
Figure 3: Common road geometry features vectorized and superimposed “snugly” on the big data surface model.
Figure 4, shows the ESSENTIAL data that was extracted from BIG data using Virtual Survey Tools.
Figure 4: The essentials of road geometry, easily imported to CAD for further design tasks by engineers
Figure 5 below shows typical geometry features needed for road design work. Conventionally these features are surveyed either by terrestrial LIDAR or by real time kinematic (RTK) global navigation satellite system (GNSS) methods, both of which take substantially longer periods of time in field data acquisition and both of which require boots on the ground in the right of way for extended periods of time. Using small drones and off the shelf cameras the data acquisition effort, including ground control survey, is drastically reduced to about an hour on site, thus not only saving significant costs but also improving the safety of the survey crew. Using Virtual Survey Tools the features shown in Figure 5 were vectorized and exported to shp and dxf formats in approximately 8hrs – all in the comfort and safety of the office environment.
Figure 5: Typical road geometry features superimposed on high resolution ortho-photo.
The features depicted in Figure 5, when delivered in both dwg and shp formats together make up about 45MB of data.
The table below shows the relative storage requirements for the products resulting from a typical SfM work flow.
Table 1: Storage Space required for geo-spatial products generated in SfM workflow and subsequent feature extraction with Virtual Survey Methods
Figure 6: Relative Storage Space required for geo-spatial products generated in SfM workflow and subsequent feature extraction with Virtual Survey Methods.
And what about accuracy? Table 2 below shows a comparison of the coordinates derived from the SfM survey of 31 targeted check points with geodetic class GPS results.
Table 2: Comparison between SfM and GPS results
Utilizing appropriate visualization and virtual survey tools, the high resolution of the original imagery facilitates precise capturing of geometric features such as shown in Figure 7 below.
Figure 7: High resolution of ortho-photo and digital surface model ensure precise capturing of features in the comfort and safety of the virtual environment
Small drones, off the shelf cameras, structure from motion (SfM) software and virtual surveying are the new tools for accurate and safe road surveying.