We just recently completed a cool project for an airport client who was having issues with their concrete surface and “pop-outs” caused by extended freeze/thaw weather events. Pop-outs are caused when the surface of the concrete sheds pieces that are about an inch wide and can be anywhere from <1cm to 3cm deep. The following graphic shows what the pop-outs look like in the field.
The client was looking for a way to quantify the number of pop-outs per slab using an automated process to avoid having to survey every pop-out which would prove to be cost-prohibitive based on the overall size of the project.
Earth Eye deployed 2 teams of data collection vehicles to compare the imagery that could be obtained from our right-of-way cameras as well as from our pavement camera.
The pavement camera has a resolution of 1mm and gives us the ability to resolve the pop-outs from a nadir view, making it easier to automate the extraction of these features from the imagery. Also, the nadir view gives us more spatial accuracy, so the locations of the pop-outs can be accurately mapped and then compared with future imagery to help quantify the amount of new pop-outs that have arisen since the last inventory. Furthermore, the gray-scale image provided by the pavement camera provided more contrast between the concrete surface and the pop-out which is much lighter in color. It was determined that the nadir-view pavement camera provided the best starting point, from which to test the automated pop-out extraction process. The following image illustrates a sample pavement image – note the pop-outs are very visible without having to zoom into the image.
The next image shows the results of our automated classification routine without any manual augmentation of missed pop-outs. We are realizing a consistent yield of greater than 95% of pop-outs identified as compared to control slabs that were collected manually in the field. Being able to efficiently map the pop-outs with a very high-yielding and automated algorithm allows us to efficiently map the pop-outs to support maintenance operations for this airport facility.
All of the pop-outs are geospatially referenced, so we can export all of the pop-outs as polygons with an area measurement associated with them. This area can then be converted to a severity and used to prescribe a specific maintenance activity based on the size and depth of the pop-out. The goal of the project was to create a quantified measurement (count) of the pop-outs for this entire project and we successfully completed this task with high-yielding, geospatial results.