Drone aerial mapping involves creating a satellite / aerial in-scaled image that can be utilized by a forensic engineer, e.g. transportation engineer, or accident reconstructionist during the accident reconstruction step. The images taken by any UAV are later processed in various software packages to build two-dimensional (2D) and three-dimensional (3D) models. This is accomplished by a modern type of photogrammetry. Information from the camera (image size, focal length, sensor width), and from the drone (GPS coordinates, flying altitude), is needed when building 2D and 3D models with the modern photogrammetry method. A model is built by matching common points between images and then combining images into one large project. Model building requires significant processing power (especially if images contain geotags) and may take twelve or more hours, or even days.
There are two types of accuracy to consider when developing UAS models. Relative Accuracy is how well a physical object or location is represented by a model based on UAS images. This accuracy is defined by Ground Sampling Distance (GSD). GSD is reported in inches per pixel and is dependent on camera specifications and flight altitude. For example, a UAS equipped with a 16 megapixel (Zenmuse X5) camera flying at an altitude of approximately 130 feet has a GSD of approximately 0.4 inches/pixel. Typically, between one to three times the GSD deviation is expected within a model. This is an acceptable degree of accuracy when balanced against the scale of a project (the area covered by the drone) and the time spent collecting the data. The second type of accuracy to be considered is Absolute Accuracy. This is defined by how closely locations in the model reflect true positions on the Earth. Absolute Accuracy error is introduced by consumer grade GPS systems utilized by cheaper UASs. To improve this accuracy, a Real Time Kinematic GPS UAS can be deployed or Ground Control Points (points with RTK GPS accuracy) need to be established before collecting data. Alternatively, Virtual Ground Control Points can be utilized.