Precision evaluation of a GPS based auto-guidance system in an agricultural vehicle by computational vision methods

• Autor(a) principal: Castro, Rigoberto Castro
• Data de Publicação: 2017
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Biblioteca Digital de Teses e Dissertações da USP
• Texto Completo: https://www.teses.usp.br/teses/disponiveis/18/18149/tde-15052024-161448/
• topic: agricultura de precisão; computational vision; GPS navigation; image processing; navegação GPS; precísion agriculture; processamento de imagem; smart vehicles; veículos inteligentes; visão computacional

Technological advances have been successfully achieved in precision agriculture using auto-guidance systems in agricultural vehicles. Among these advances, the increase of efficiency and the productivity in field operations can be highlighted. Some auto-guidance driving systems are implemented using the GPS RTK system, which allows operations to centimeter accuracy. However, the geographic positioning errors, the vehicle dynamics, the agricultural devices and the field environment (slopes, soil condition, etc.) may influence the performance of GPS based autonomous agricultural vehicles. In this way, the evaluation of the auto-guidance driving systems becomes essential to the achievement of high precision levels in field operations. This evaluation can be performed by measuring the displacements using precise sensors installed in the vehicle, such as: cameras, lasers, odometer, and ultrasonic sensors, among others. Among the local sensing options, it is well-know that computational vision methods allow the location of any system in the space, becoming it a technical alternative for this evaluation. In this way, the objective of this research is to propose a methodology to assess the accuracy of auto-guidance systems under real field conditions by means of computer vision methods. The vehicle under study is a tractor equipped with an auto-guidance system, which is composed of a GPS RTK unit and an inertial measurement unit (IMU). The instrumentation consisted of two Canon Rebel T5 cameras with focal lens of 50 and 18 millimeters respectively. The pinhole camera method was used to map vehicle location in the field using computational vision techniques. In the study, multiple field tests were performed, proving that the use of the computer vision method is accurate to evaluate auto-guidance systems if devices, procedures, and parameters are properly selected