A strawberry weight estimation method and device based on point cloud reconstruction and a medium

Abstract

The application discloses a strawberry weight estimation method and device based on point cloud reconstruction and a medium, and belongs to the field of target detection and classification. The method comprises the following steps: obtaining original point cloud of strawberries placed on a plane; performing plane segmentation on the original point cloud to obtain strawberry point cloud and plane point cloud, and obtaining a normal vector of the plane point cloud; performing outlier filtering on the strawberry point cloud, and calculating the center and main direction of the strawberry point cloud; constructing a strawberry coordinate system according to the normal vector, the center and the main direction, and generating an aabb bounding box with the constructed coordinate system as a reference coordinate system; based on the aabb bounding box, the strawberry point cloud is iteratively reconstructed using a dichotomy method to generate reconstructed point cloud; obtaining a convex hull of the reconstructed point cloud, and calculating the volume of the convex hull; and substituting the volume of the convex hull into a preset weight prediction formula to obtain a predicted weight. The application considers the geometric constraint of the placement plane, reconstructs the actual three-dimensional shape of the strawberry, and predicts the weight through the volume, so that the prediction effect is more accurate.

Images