Abstract
In sparse representation based classification (SRC) and weighted SRC (WSRC), it is time-consuming to solve the global sparse representation problem. A discriminant WSRC (DWSRC) is proposed for large-scale plant species recognition, including two stages. Firstly, several subdictionaries are constructed by dividing the dataset into several similar classes, and a subdictionary is chosen by the maximum similarity between the test sample and the typical sample of each similar class. Secondly, the weighted sparse representation of the test image is calculated with respect to the chosen subdictionary, and then the leaf category is assigned through the minimum reconstruction error. Different from the traditional SRC and its improved approaches, we sparsely represent the test sample on a subdictionary whose base elements are the training samples of the selected similar class, instead of using the generic overcomplete dictionary on the entire training samples. Thus, the complexity to solving the sparse representation problem is reduced. Moreover, DWSRC is adapted to newly added leaf species without rebuilding the dictionary. Experimental results on the ICL plant leaf database show that the method has low computational complexity and high recognition rate and can be clearly interpreted.
Highlights
Plant species automatic identification via computer vision and image processing techniques is very useful and important for biodiversity conservation
The sparse representation (SR) coefficients obtained by discriminant WSRC (DWSRC) might not be as sparse as the ones achieved by the classical sparse representation based classification (SRC) and weighted SRC (WSRC), but the SR coefficients of the relevant training samples still own greater magnitudes, because all training samples of a candidate similar class generally play an important role in sparsely representing the test sample
We investigate the performance of the proposed DWSRC method for plant species recognition and compare it with four state-of-the-art approaches, including plant species recognition using WSRC [21], Leaf Margin Sequences (LMS) [28], Manifold–Manifold Distance (MMD) [29], and Centroid Contour Distance (CCD) [30]
Summary
Plant species automatic identification via computer vision and image processing techniques is very useful and important for biodiversity conservation. Hsiao et al [15] proposed two SRC based leaf image recognition frameworks for plant species identification and experimentally compared them. The existing SRC and WSRC would fail in leaf based large-scale plant species recognition, because it is time-consuming to solve the l1norm minimization problem in a dictionary composed of all training data across all classes. In classical SRC, WSRC, and their modified approaches, to achieve good sparse performance, all training samples are employed to build the overcomplete dictionary [13, 14]. It is time-consuming to solve the SR problem through the overcomplete dictionary, especially in large-scale image recognition task.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
