Building Classification of VHR Airborne Stereo Images Using Fully Convolutional Networks and Free Training Samples

Abstract

<p><strong>Abstract.</strong> Semantic segmentation, especially for buildings, from the very high resolution (VHR) airborne images is an important task in urban mapping applications. Nowadays, the deep learning has significantly improved and applied in computer vision applications. Fully Convolutional Networks (FCN) is one of the tops voted method due to their good performance and high computational efficiency. However, the state-of-art results of deep nets depend on the training on large-scale benchmark datasets. Unfortunately, the benchmarks of VHR images are limited and have less generalization capability to another area of interest. As existing high precision base maps are easily available and objects are not changed dramatically in an urban area, the map information can be used to label images for training samples. Apart from object changes between maps and images due to time differences, the maps often cannot perfectly match with images. In this study, the main mislabeling sources are considered and addressed by utilizing stereo images, such as relief displacement, different representation between the base map and the image, and occlusion areas in the image. These free training samples are then fed to a pre-trained FCN. To find the better result, we applied fine-tuning with different learning rates and freezing different layers. We further improved the results by introducing atrous convolution. By using free training samples, we achieve a promising building classification with 85.6<span class="thinspace"></span>% overall accuracy and 83.77<span class="thinspace"></span>% F1 score, while the result from ISPRS benchmark by using manual labels has 92.02<span class="thinspace"></span>% overall accuracy and 84.06<span class="thinspace"></span>% F1 score, due to the building complexities in our study area.</p>

Building Segmentation From Airborne VHR Images Using Mask R-CNN

“…All images were also rescaled to 1024 × 1024 with a batch size of 2 for 100 epoches in FCN. The learning rate was set to 1e-5 according to (Chen et al, 2018). The same weight decay and momentum were defined as in the case of the Mask R-CNN.…”

“…Topographic maps have been used to generate training samples automatically for VHR images (Maggiori et al, 2017, Kaiser et al, 2017, Chen et al, 2018. However, the topographic maps are often mismatched with images due to time differences.…”

Methods for the Prediction and Specification of Functionally Graded Multi-Grain Responsive Timber Composites

“…Since the appearance of a normal virtual model is similar to but not the same as a photograph, it is difficult to obtain highly accurate detection results in the real world even if the image is used for deep learning training. A method is also proposed for generating label images by using aerial photographs and map information (Chen et al 2018). In this research, we aim to improve the representation of the VR models by using textured aerial photographs on 3D models.…”