Remote Sensing: An Automated Methodology for Olive Tree Detection and Counting in Satellite Images

Abstract

Cultivation of olive trees for the past few years has been widely spread across Mediterranean countries, including Spain, Greece, Italy, France, and Turkey. Among these countries, Spain is listed as the largest olive producing country with almost 45% of olive oil production per year. Dedicating land of over 2.4 million hectares for the olive cultivation, Spain is among the leading distributors of olives throughout the world. Due to its high significance in the country’s economy, the crop yield must be recorded. Manual collection of data over such expanded fields is humanly infeasible. Remote collection of such information can be made possible through the utilization of satellite imagery. This paper presents an automated olive tree counting method based on image processing of satellite imagery. The images are pre-processed using the unsharp masking followed by improved multi-level thresholding-based segmentation. Resulting circular blobs are detected through the circular Hough transform for identification. Validation has been performed by evaluating the proposed scheme for the dataset formed by acquiring images through the “El Sistema de Informacion Geografica de Parcelas Agricolas” viewer over the region of Spain. The proposed algorithm achieves an accuracy of 96% in detection. Computation time was recorded as 24 ms for an image size of $300\times300$ pixels. The less spectral information is used in our proposed methodology resulting in a competitive accuracy with low computational cost in comparison to the state-of-the-art technique.