A novel image processing framework to detect geometrical features of horticultural crops: case study of Anthurium flowers

An image processing framework was developed to automatically detect the geometrical features of horticultural crops via identifying theirs proximal and distal ends. The developed algorithm detects boundary of the crops and fits a suitable B-spline curve on it. Subsequently, the algorithm calculates the first and second derivatives of the curve and finds its geometrical key points as mathematical critical points. Thus, it is possible to calculate geometrical features of the crops such as perimeter, area, curvature, and especially length and width, at any desired angle that the crops are placed under the camera. The algorithm was tested on three cultivars of Anthurium flower and results showed that the reconstructed shapes of the flowers considered high adaptation to the original shapes. Geometrical key points of the flower shape were correctly detected in all experiments. In addition, geometrical features of rotated flowers were calculated with a relative error of less than 2%. The presented algorithm could be used to detect geometrical attributes of horticultural crops with irregular shapes from two dimensional (2D) images. It would be useful to develop real-time quality inspection and grading systems, as well as to collect accurate and detailed morphological data from 2D objects.