Detect Building project was born from the customer’s need to be able to detect the existence or otherwise of buildings in certain points assigned via geographical coordinates. This tool must allow the identification and classification of the building through visualization on a map.
The geospatial analysis of the area examined will lead to an optimization of the evaluation process for sustainable interventions aimed at installing fiber optics.
Starting from the sentinel-2 images with a spatial resolution of 10m, proprietary super-resolution algorithms were applied which allow images to be obtained at 1 m resolution. This aspect is fundamental to obtain good results otherwise it would be much more difficult to identify small buildings and constructions.
Using super-resolution images, machine learning models, and specifically object detection, have been developed to determine whether or not the building is present. After having identified the element it is also possible to detect the distance from the coordinate point taken into consideration.
To identify constructions, a deep learning algorithm based on convolutional neural networks (CNN) was developed, capable of classifying images as “Building” (“TRUE”) and “Building” (“FALSE”).
A subsequent phase was to create an ad hoc link for each point to which the exact address viewable on the Google Street View platform is connected, in this way the company representative is able not only to view the building in exam, but to be able to navigate via the Google platform to find other information necessary for the installation of optical fiber, such as the conditions of the road surface, sidewalks, parapets, small bridges, bumps, etc.
There are various qualitative benefits identified in the design phase to highlight the potential of the solution proposed by Latitudo 40. Thanks to the data already in possession and the integration of the data provided by Latitudo 40, the fiber optic installation company will be able to access little-known information or in any case consultable with a large expenditure of resources and time.
The customer will then be able to:
- reduce the timescales for evaluating interventions in the areas under consideration;
- simplify the analysis by reducing the number of inspections to plan;
- provide already digitized and QGIS ready data, allowing savings in man hours for the digitization of inspections;
- standardize the analysis which is always performed in the same way, reducing the risk of having non-aligned data;
- limit the amount of inspections, especially in rural areas far from operational headquarters;
- have a historian (since 2017) to study urban evolution;
- reduce emissions of polluting substances caused by travel for inspections.