Standardization of the Shape of Ground Control Point (GCP) and the Methodology for Its Detection in Images for UAV-Based Mapping Applications

The challenge of georeferencing aerial images for an accurate object to image correspondence has gained significance over the past couple of years. There is an ever-increasing need to establish accurate georeferencing techniques for Unmanned Aerial Vehicles (UAVs) for tasks like aerial surveyance of mines/construction sites, change detection along national highways, inspection of major pipelines, intelligent farming, among others. With this paper, we aim to establish a standard method of georeferencing by proposing the design of a simple, white colored, L-shaped marker along with the pipeline for its detection. In a first, the less common DRGB color space is used along with the RGB color space to segment the characteristic white color of the marker. To carry out recognition, a scale and rotation invariant modification of the edge oriented histogram is used. To allow for accurate histograms, improvements are made on canny edge detection using adaptive approaches and exploiting contour properties. The histogram obtained displayed a characteristic distribution of peaks for GCP-markers. Thus, a new peak-detection and verification methodology is also proposed based on the normalized cross-correlation. Finally, a CNN model is trained on the Regions of Interest around the GCP-markers that are received after the filtering. The results from EOH and CNN were then used for classification. Regions with a diverse range of locality, terrain, soil quality were chosen to test the pipeline developed. The results of the design and the pipeline combined were quite impressive, with regards to the accuracy of detection as well as its reproducibility in diverse geographical locations.