While a number of different methods have been developed to build drive-by thermal imaging systems, Essess’ patent-pending technology is unique in the manner in which it combines expertise from multiple fields – including thermal physics, structural physics, computer vision, and machine learning – specifically to the problem of drive-by imaging using repurposed vehicles. This differentiated approach enables a level of accuracy and reliability necessary to commercialize drive-by thermal imaging at scale.
For example, the custom-designed imaging system captures images of the same structure in multiple batches at different wavelengths. The first set of images for any given structure are captured in an initial range of wavelengths (for example, 350 nm to 1.2 µm), while the second set of images are captured in a different range of wavelengths (for example, 8 µm to 12 µm) – all while the vehicle is moving along the road. A single imaging system may capture images in both wavelength ranges. The combination of near infrared and long wave infrared image capture enables much greater accuracy when matching buildings to GPS coordinates. This example is just one among many of the unique ways that Essess has enabled the conversion of drive-by thermal imaging technology from the experimental labs of academia to meet the rigor of the commercial market.
The patent-pending technology covers not only the hardware and related software capabilities but also data processing after scans have been completed. The scan data is imported from an electronic data storage location into the patent-pending data processing pipeline. The pipeline unpacks any videos into images using a suite of proprietary algorithms, converting gray scale images to temperature images, grouping images for vertical stitching, and performing the vertical stitching leveraging computer vision and SLAM techniques. The geo-location data is also imported into the processing pipeline and is used to geo-tag vertical panoramas and match vertical panoramas to buildings. At this stage, the constructed images are processed by a set of algorithms that determine the average surface temperature and infer the internal temperature of the building. This serves as an input into the thermal analytical engine that calculates building surface heat flow and energy scores.