The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Publications Copernicus
Articles | Volume XLII-2/W13
04 Jun 2019
 | 04 Jun 2019


M. Hassanein, M. Khedr, and N. El-Sheimy

Keywords: Precision Agriculture, UAV, Crop Row Detection, HSV, PCA

Abstract. Precision Agriculture (PA) management systems are considered among the top ten revolutions in the agriculture industry during the last couple decades. Generally, the PA is a management system that aims to integrate different technologies as navigation and imagery systems to control the use of the agriculture industry inputs aiming to enhance the quality and quantity of its output, while preserving the surrounding environment from any harm that might be caused due to the use of these inputs. On the other hand, during the last decade, Unmanned Aerial Vehicles (UAVs) showed great potential to enhance the use of remote sensing and imagery sensors for different PA applications such as weed management, crop health monitoring, and crop row detection. UAV imagery systems are capable to fill the gap between aerial and terrestrial imagery systems and enhance the use of imagery systems and remote sensing for PA applications. One of the important PA applications that uses UAV imagery systems, and which drew lots of interest is the crop row detection, especially that such application is important for other applications such as weed detection and crop yield predication. This paper introduces a new crop row detection methodology using low-cost UAV RGB imagery system. The methodology has three main steps. First, the RGB images are converted into HSV color space and the Hue image are extracted. Then, different sections are generated with different orientation angles in the Hue images. For each section, using the PCA of the Hue values in the section, an analysis can be performed to evaluate the variances of the Hue values in the section. The crop row orientation angle is detected as the same orientation angle of the section that provides the minimum variances of Hue values. Finally, a scan line is generated over the Hue image with the same orientation angle of the crop rows. The scan line computes the average of the Hue values for each line in the Hue image similar to the detected crop row orientation. The generated values provide a graph full of peaks and valleys which represent the crop and soil rows. The proposed methodology was evaluated using different RGB images acquired by low-cost UAV for a Canola field. The images were taken at different flight heights and different dates. The achieved results proved the ability of the proposed methodology to detect the crop rows at different cases.