Regresar

Parallel CPU-Based Processing for Automatic Crop Row Detection in Corn Fields

Abstract:

With technological advancement, current computer applications generate large volumes of heterogeneous data (text, images, video, audio), causing a considerable increase in processing execution times. The use of heterogeneous architectures in the systematization of processes is an alternative to obtain better processing times, optimizing computational resources, through the interaction between CPU (Central Processing Unit) and GPU (Graphics Processing Unit) architectures for parallel processing. Specifically, in precision agriculture (PA) Image analysis for the automatic detection of crop lines is essential for the use of autonomous systems applied to different agricultural processes in real-time. In the present work, the use of parallel programming is proposed for the implementation of an existing algorithm for the detection of curved and straight lines of crops, based on micro-ROI (Small regions of interest), with images captured in corn-fields during the early stages of growth. Matlab (Parallel Computing Toolbox) libraries were used for parallel and sequential implementation in a multicore CPU to contrast execution times in a set of 300 images. The results obtained were statistically validated using the T-Student test in the R programming language. The evaluated times indicate accelerations of 20% on average in recognition of crop lines, demonstrating a notable improvement in the performance of the parallel algorithm over the sequential one.