Abstract
High spatial resolution images obtained by cameras onboard on unmanned aerial vehicles (i.e drones) have significant potential in modern agriculture, as they allow the accuracy and efficiency of some field operations to be improved from their analysis and interpretation. Georeferenced crop rows are used as an input for guiding precision agricultural machinery. Agricultural machinery is supported by highly accurate, real-time kinematic global satellite navigation systems. Georeferenced crop rows generation is an expensive and time-consuming task. This project addressed a crop rows generation problem in sugarcane crops through image processing and computer vision techniques. Using high-resolution aerial imagery from unmanned aerial vehicles an automated method was designed and built to generate georeferenced crop rows quickly and accurately. The computational tool "Crop Rows Generator CRG v1.0" was developed. CRG involves computer vision techniques and a high-performance computing approach which are capable of processing high-resolution large images and on these images detect, generate and mapping crop rows in sugarcane fields, with a few clicks. CRG is open source and consists of a client interface (CRG-QGIS Plugin) integrated as a plugin into the geographic information system software (QGIS Desktop), a processing core accessed by a command line interface (CRG-CLI) and resources exposed through on HTTP methods by REST API (CRG-API). Generate crop rows using the developed tool improves processing above to 200% compared to manual method and beyond 650% compared to the mechanized method. Crop Rows generation performance is above to 83% and overall most of the horizontal errors are in range between (2.5 to 10) cm.
Demo Videos
Video 1
Working from high-resolution images, CRG generates georeferenced crop rows in a couple of simple steps.
Video 2
Complex boundaries are supported as an area of interest for crop rows generation.
