Plant Leaf Image Recognition using Multiple-grid Based Local Descriptor and Dimensionality Reduction Approach

Thipwimon Chompookham, Sarayuth Gonwirat, Siriwiwat Lata, Sirawan Phiphiphatphaisit, Olarik Surinta

Abstract

The identification process of plant species is one of the significant and challenging problems. In this research area, many researchers have focused on identifying the plant leaf images because the leaves of a plant are found almost all year round. The achieve method of the plant leaf image recognition is based on unique extraction features from the plant leaf and using the well-known machine learnings as a classification method. As a result, recognition accuracy was often not very high. In order to improve recognition accuracy, we proposed a multiple grids technique based on the local descriptors and dimensionality reduction. Firstly, we divided the plant leaf image according to grid size and calculated the local descriptors from each grid. Secondly, the dimensionality reduction is proposed to transform and decrease the correlated variables of the feature vector. Finally, the feature vector with a relatively low-dimensional is transferred to the machine learning techniques, which are the support vector machine and multi-layer perceptron algorithms. We have evaluated and compared the proposed algorithm with the bag of visual words method and the deep convolutional neural network (including AlexNet and GoogLeNet architectures) on the Folio leaf image dataset. The experiments show that the proposed algorithm has improved and obtained very high accuracy on plant leaf image recognition.

Keywords — Plant leaf recognition; Multiple grids approach; Local descriptor; Dimensionality reduction; Support vector machine; Multi-layer perceptron.

Read article — https://dl.acm.org/doi/abs/10.1145/3388176.3388180

Conclusion

In this paper, we have investigated many different plant leaf recognition techniques on a Folio dataset. From the experimental results, we conclude that the performance of multiple grids and dimensionality reduction based descriptors, which is our proposed method, is much better than the histogram of oriented gradients combined with bag-of-words technique and fine-tuned deep CNN architectures which are AlexNet and GoogleNet architectures as well. We also have shown that the principal component analysis (PCA), which is the dimensionality reduction technique, increased the accuracy performance and decreased the number of the feature vector of the plant leaf recognition system. Nevertheless, the data augmentation technique can increase the accuracy performance of the plant leaf recognition system. This technique added more than 4,000 illumination images to the training set. Subsequently, we used only 510 images to train the plant leaf recognition system. As a result, the accuracy result of our proposed method is slightly decreased than the fine-tuned deep CNNs with the combined data augmentation technique.

According to the high accuracy of the deep CNNs, in future work, we would like to study the effect of parallel CNN architecture and use this architecture to train the plant leaf images. This technique maybe necessary to improve training times and accuracy performance.