Instrumentation and method for high accuracy geo-referencing of sugar beet plants

Efficient use of automatic field operations will allow care and management of crops in very different systems from what is known today and may lead to the possibility of individual plant care systems. Automatic field operation systems have the potential to reduce environmental impact while preserving economics in crop production. These systems require accurate and reliable information about the position of individual crop plants and, if possible, additional information about crop growth status. The aim of the presented research was to generate a geo-spatial map of individual crop plants derived from geo-referenced data recorded during seeding operation. A standard sugar beet precision seeder was retrofitted with optical sensors that could detect seeds as they were released into the furrow. Furthermore, a real time kinematic global positioning system (RTK-GPS) and a dual axis tilt sensor provided the global position and attitude angles of the seeder. A data acquisition system was configured for recording and storage of global positions, seeder attitude data, and seed drop detections during seeding. This paper outlines the methodology of processing the recorded data into a geo-spatial seed map. The developed instrumentation was used in field experiments under typical conditions and operation velocities up to 5.3 km h(-1). The validation showed that 95% of the sugar beet seedlings emerged within 37.3 mm from the seed drop positions contained in the geo-spatial seed map. An error analysis associated with the estimation of geo-referenced plant positions consisted of positioning sensor errors, seed displacement in the furrow and deviation between location of plant emergence and the corresponding seed location. Furthermore, the error contribution from individual sources was of the same magnitude, except for the error due to deviation between location of plant emergence and the corresponding seed location. As this fully random error will always occur, it was considered meaningless trying to further minimize the sensor errors. Inclusion of seeder attitude data in the data processing significantly improved the accuracy of the estimation of geo-referenced plant positions and therefore it was concluded that a dual axis tilt sensor should be a required part of the instrumentation. Furthermore, it was shown that high accuracy of the estimation of georeferenced plant positions required a zero horizontal velocity of the seed released from the seeding mechanism. In general, the overall accuracy of the estimation of geo-referenced plant positions was satisfactory to allow subsequent individual plant scale operations. (c) 2007 Elsevier B.V. All rights reserved.