Demand
BeeXact features so-called mobile mapping cars. These cars take camera images of environments, such as houses and streets, and transmit this information. BeeXact wanted to automate the transformation of camera images to 2D. This allows the coverage of the ground (street, sidewalk, grass), to be mapped more quickly. With this information, contracts, among others, are supported in the construction of a fiber optic network.
Previously, camera images were viewed by humans and the ground coverage was entered manually. Datacadabra thought it could be done more efficiently by using AI.
The process
Datacadabra set to work to investigate the accuracy with which camera images can be automatically transformed into 2D BGT (Basisregistratie Grootschalige Topografie, in English: basic registration of large-scale topography) polygons. Freely translated, this means that the camera images are converted to a 2D map showing what is grass and what is the street.
Data
We obtained the camera images from BeeXact's own "mobile mapping cars. These images were already created, but the processing was not automatic.
Services
Datacadabra built a tool based on computer vision. The application of this type of artificial intelligence allowed an algorithm to process the camera images and the tool indicated what type of cover is present at which spot. This was then displayed on a 2D image.
Feedback
We designed a user interface/application that allowed BeeXact to load and process the data from the algorithm.
The result
The result is a strong labor (cost) savings for BeeXact. Video images from the mobile mapping cars no longer need to be analyzed and entered by hand. Because of Datacadabra's algorithm and tool, only a manual correction is needed. In addition, functionalities have been added; including the "blurring" of faces and license plates. This was previously not performed automatically, so here too an efficiency improvement has been made.
