A preliminary evaluation of vision and laser sensing for tree trunk detection and orchard mapping

Mapping is a significant issue in mobile robot applications. Mobile robots can build a map or a model of the environment using different sensors. An orchard is a suitable agricultural environment for mobile robot applications since it is a semi-structured environment, where trees are planted in nominally straight rows. This paper presents a new method to extract features from the orchard environment using a camera and laser range scanner to create a map of the orchard. The map of the orchard is based on the detection of tree trunks. In this study, image segmentation and data fusion methods are used for feature extraction, tree detection and orchard map construction. Integration of both machine vision and laser sensor provides more robust information for tree trunk detection and orchard mapping. The resulting map composes of the coordinates of individual trees in each row as well as the coordinates of other non-tree objects detected by the sensors.