Abstract
This study proposes a light convolutional neural network (LCNN) well-fitted for medium-resolution (30-m) land-cover classification. The LCNN attains high accuracy without overfitting, even with a small number of training samples, and has lower computational costs due to its much lighter design compared to typical convolutional neural networks for high-resolution or hyperspectral image classification tasks. The performance of the LCNN was compared to that of a deep convolutional neural network, support vector machine (SVM), k-nearest neighbors (KNN), and random forest (RF). SVM, KNN, and RF were tested with both patch-based and pixel-based systems. Three 30 km × 30 km test sites of the Level II National Land Cover Database were used for reference maps to embrace a wide range of land-cover types, and a single-date Landsat-8 image was used for each test site. To evaluate the performance of the LCNN according to the sample sizes, we varied the sample size to include 20, 40, 80, 160, and 320 samples per class. The proposed LCNN achieved the highest accuracy in 13 out of 15 cases (i.e., at three test sites with five different sample sizes), and the LCNN with a patch size of three produced the highest overall accuracy of 61.94% from 10 repetitions, followed by SVM (61.51%) and RF (61.15%) with a patch size of three. Also, the statistical significance of the differences between LCNN and the other classifiers was reported. Moreover, by introducing the heterogeneity value (from 0 to 8) representing the complexity of the map, we demonstrated the advantage of patch-based LCNN over pixel-based classifiers, particularly at moderately heterogeneous pixels (from 1 to 4), with respect to accuracy (LCNN is 5.5% and 6.3% more accurate for a training sample size of 20 and 320 samples per class, respectively). Finally, the computation times of the classifiers were calculated, and the LCNN was confirmed to have an advantage in large-area mapping.
Highlights
Land-cover mapping is one of the most essential applications of remote sensing, and its importance has been affirmed by a growing number of social and natural scientists who are utilizing land-cover maps to acquire critical information about issues such as urban planning [1,2], carbon-cycle monitoring [3,4], forest monitoring [5,6,7], and various multidisciplinary studies [8,9,10]
An light convolutional neural network (LCNN) with a patch size of five, an support vector machine (SVM) with a patch size of three, and an SVM with a pixel-based system had the highest accuracies. This result indicates that the LCNN-3 was successfully designed for land-cover mapping with sufficient model capacity to attain high accuracy without overfitting with small sample sizes due to its light design
The average overall accuracies of the three classifiers were calculated according to sample sizes, and pairwise one-sided p-values were obtained by conducting paired t-tests to test significant differences with LCNN-3 for each training sample size
Summary
Land-cover mapping is one of the most essential applications of remote sensing, and its importance has been affirmed by a growing number of social and natural scientists who are utilizing land-cover maps to acquire critical information about issues such as urban planning [1,2], carbon-cycle monitoring [3,4], forest monitoring [5,6,7], and various multidisciplinary studies [8,9,10]. The various studies that have undertaken this issue have not been consistent in determining the best classification algorithms [21] This is because the image types and test sites differ from study to study, and the results vary depending on whether the ancillary data is used and site-specific factors that determine the land-cover classes and legend criteria [19,21,28]. Since the sensitivity of an algorithm to the training sample size varies, the superiority of the classifiers is not determined [22,24] Because these confounding factors impede an integrated conclusion, the remote-sensing community has been suffering from the inability to select the most suitable classifier
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
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
