18m 16s | 172 view | 1 rating

Presenting Author: Ehsan Kiani Oshtorjani - Akdeniz University, Agricultural Machinery & Technologies Engineering , Turkey Link to Conference: http://conferences.au.dk/cigr-2016/sessions-and-participants/#/programme?_k=621wji Title:Design of an Embedded System based on Machine Vision for Autonomous Weed Control Applications Description Precision agriculture seeks modern technologies to lower the cost of automatic farming operations. This paper proposes a real-time crop detection system designed for an embedded autonomous weeding machine. The system uses a modern pattern recognition methodology implemented on a cost-effective portable hardware platform. Maize seedlings do not have certain geometric pattern. Therefore, the proposed methodology employs advanced machine vision and learning techniques for irregular pattern recognition and object classification aligned for rapid field crop inspection. The algorithm is mainly based on Viola-Jones framework and optimizes the computational efficiency and response time over the parameters such as cultivator travel speed, weed emergence, lighting conditions, plant morphological and growth variation, camera view angle and height, and etc. All analysis process is performed on a single-board computer with an on-board camera. Since maize breeders in Antalya prefer a commercially affordable within-row weeding machine, the proposed system is tested over a maize land. The weed/crop discrimination results of classifier training on a typical local crop row under a common lighting condition indicate a satisfactory performance for regular cultivator travel speeds. Contributors EHSAN KIANI OSHTORJANI (Agricultural Machinery and Technology Engineering Departmnet, Akdeniz University, Turkey), MEHMET TOPAKCI (Agricultural Machinery and Technology Engineering Departmnet, Akdeniz University, Turkey), ILKER UNAL (Technical Vocational School, Akdeniz University, Turkey)