Nielsen, M. R.2; Stigaard Laursen, Morten5; Jonassen, M. S.2; Nielsen, K.6; Jørgensen, R. N.4
John V. Stafford
1 The Maersk Mc-Kinney Moller Institute, Faculty of Engineering, SDU2 Danish Technological Institute3 Mærsk Mc-Kinney Møller Instituttet for Produktionsteknologi, Faculty of Engineering, SDU4 Syddansk Universitet5 The Maersk Mc-Kinney Moller Institute, Faculty of Engineering, SDU6 Mærsk Mc-Kinney Møller Instituttet for Produktionsteknologi, Faculty of Engineering, SDU
This paper presents a novel machine vision-based approach detecting and mapping gall mite infection in dormant buds on black currant bushes. A vehicle was fitted with four cameras and RTK-GPS. Results compared automatic detection to human decisions based on the images, and by mapping the results into the same GPS map as the original ground truth recorded in the field. Human inspection found 52% of the infected bushes and the algorithm found the same 52%. The algorithm marked 7% of the healthy bushes, due to healthy buds looking larger closer to the camera. Consequently, the remaining challenge lies in optimising the viewing perspective.
Precision Agriculture 2013: Papers Presented at the 9th European Conference on Precision Agriculture, 2013, p. 517-524
Gall mite Machine vision Mapping Sensor Stress
Main Research Area:
European Conference on Precision Agriculture, 2013