Leaf image retrieval using combined shape feature sets with fuzzy integral

In this paper, leaf image retrieval using combined three shape feature sets is presented. The shape feature sets adopted include Centroid-contour distance (CCD) curve, Moment invariants (MIs), and Angle code histogram (ACH). At first, a thinning-based method is proposed to locate possible starting points of a leaf image contour so that the approach is more computationally efficient in image matching. This starting point location method can also benefit other shape representation schemes that are sensitive to starting points. After the similarity measures of individual feature sets are computed, a fuzzy integral is employed to combine them. The fuzzy integral approach has a distinct advantage in releasing the user's burden from tuning the combination parameters required in the weighted summation approach that is time-consuming and cannot guarantee the best combination performance. Experimental results on our 830 leaf images from 83 plants (10 samples from each plant) show that our approach compared favorably with other two methods tested, the Curvature scale space (CSS) method and the Modified fourier descriptor (MFD) method.