CREC (SRP): Automated Location and Identification of SRP Canals Infrastructure Components with Advanced Image Analytics

CREC (SRP): Automated Location and Identification of SRP Canals Infrastructure Components with Advanced Image Analytics Automated Location and Identification of SRP Canals Infrastructure Components with Advanced Image Analytics In response to the LOI As-Built Modeling of SRP Canals Infrastructure with Terrestrial Laser Scanning Technologies, Mr. Todd E. Rakstad (SRPs Land-Survey manager, Land/Survey Division), Mr. Stan Dickey (Field Supervisor), and Mr. Chad Woolgar (Survey Technician Supervisor) met with Dr. David Grau (PI) in order to discuss how to better respond to SRPs needs. Indeed, SRP experts want to pursue an automated or at least semi-automated (with minimal human intervention) identification and location of SRPs key infrastructure components and assessment of geographical features. Even though the LOI revolved around scanning technologies at SRP request, initial tests by SRP have proven laser scanning technologies unfeasible from economic, storage, and computing perspectives. For instance, storage costs alone would account for more than $2 million annually. These costs do not account for the highly timeintensive (frequently manual) data collection and 3D object-recognition post-processing efforts (Sternberg et al. 2004, Jaselskis et al. 2005, Zhu and Brilakis 2007, Kashani and Grau 2012). During the discussion, image analytic processing techniques from aerial and/or satellite images emerged as a potential option to automatically identify and locate assets. Finally, during the meeting the land-surveying experts also expressed the need to document the infrastructure condition specific to the metropolitan water canals. The specific scope of this project is to automate the detection, classification, and localization of SRPs assets that are relevant, distinguishable, and frequent along the canals. The proposed method (see Figure 1) uses as input high-resolution vertical images of the existing canal infrastructure, produced with above-the-ground photogrammetric sensors (e.g. satellites, or drone-mounted cameras). The surface is first traversed for the detection of visual patterns that characterize common SRPs infrastructure components. The resulting features are used both for the initial training and subsequent classification of key component features into/from a component classification database. The end result is a raster image segmented into georeferenced clusters of information, most of which are recognized as common components of a certain type (e.g. power towers, stations, adjacent-property walls, water, etc.), and, from those, corresponding geographical features (center lines, boundary lines, electrical lines, right-of-way divisions, etc.). The end-user can modify the automated result, which is used as additional learning feedback to improve the methods classification capabilities. The five objectives of this study are defined as follows: 1. Formalize the numerical representation strategy for the identification of SRP infrastructure components. 2. Develop image analytic algorithms to identify and match components on raster images with their counterparts on the 2D surface. 3. Geo-reference the identified components. 4. Engage undergraduate and graduate students in this study. 5. Produce a final and intermediate research reports. Due to the short 1-year duration of this project, this study represents a first step to automate the location, identification, and assess changes in the canals infrastructure and also geographic condition using advanced image analytics from vertical photogrammetric images. During the meeting with SRP experts, potential follow-up studies were discussed to extend the location and 2 identification of components to other distinguishable urban areas (e.g. streets, power generation facilities), or to integrate the geo-located components and geographical references in georeferenced databases (e.g. Geographic Information Systems (GIS)). Reduction of Chemical Use and Water Consumption in Cooling Towers Using Soft Wave Water Technology