Traditional Village Building Extraction Based on Improved Mask R-CNN: A Case Study of Beijing, China

Abstract

As an essential material carrier of cultural heritage, the accurate identification and effective monitoring of buildings in traditional Chinese villages are of great significance to the sustainable development of villages. However, along with rapid urbanization in recent years, many towns have experienced problems such as private construction, hollowing out, and land abuse, destroying the traditional appearance of villages. This study combines deep learning technology and UAV remote sensing to propose a high-precision extraction method for conventional village architecture. Firstly, this study constructs the first sample database of traditional village architecture based on UAV remote sensing orthophotos of eight representative villages in Beijing, combined with fine classification; secondly, in the face of the diversity and complexity of the built environment in traditional villages, we use the Mask R-CNN instance segmentation model as the basis and Path Aggregate Feature Pyramid Network (PAFPN) and Atlas Space Pyramid Pool (ASPP) as the main strategies to enhance the backbone model for multi-scale feature extraction and fusion, using data increment and migration learning as auxiliary means to overcome the shortage of labeled data. The results showed that some categories could achieve more than 91% accuracy, with average precision, recall, F1-score, and Intersection over Union (IoU) values reaching 71.3% (+7.8%), 81.9% (+4.6%), 75.7% (+6.0%), and 69.4% (+8.5%), respectively. The application practice in Hexi village shows that the method has good generalization ability and robustness, and has good application prospects for future traditional village conservation.

Building Extraction on SegFormer Model and Post Processing from Remote Sensing Images

“…the Images folder is used to save the image slices; the sample labels are saved in the form of 0 and 1 binary rasters to the Masks folder, where 1 is the feature (3) Other configurations In deep learning models, the parameters that need to be configured according to the situation before training are called hyperparameters. Hyperparameter tuning is a commonly used method of model tuning, such as adjusting the number of training times (Epoch) [38], the learning rate [10], the single-step operation volume (Batch Size), etc. The Learning Rate is usually in the range of 0.0 and 1.0, and commonly used ones are 0.01, 0.001 and 0.0001, etc.…”

“…Semantic segmentation is an important and challenging task in computer vision, where the goal is to segment an image into multiple regions, each of which corresponds to a specific category or object [8]. The design of SegFormer is inspired by previous successful semantic segmentation methods, such as DeepLab [9] and Mask R-CNN [10]. Before SegFormer was proposed, there existed many other neural network architectures for image segmentation such as FCN [11], U-Net [5], Mask R-CNN, FPN [11], DeepLab, BiSeNet [12,13] and OCRNet [6,13].…”

“…The x-axis in the training log is usually the Epoch or Iterator counts; the y-axis is often different evaluation metrics to indicate the accuracy of the model, commonly used are Loss[10], Accuracy, IoU and F1[33]. Model convergence with a small generalisation gap is the goal of training, and the generalisation gap is the difference between the model'sperformance on the training and validation sets, which is due to the difference in data between the training and validation sets.…”

RSI-YOLO: Object Detection Method for Remote Sensing Images Based on Improved YOLO

“…This method uses HRNet feature extraction network to detect small targets better and proposes a new overlapping segmentation method to solve the problem of target misfit and missing detection. Wenke Wang et al [ 17 ] proposed an extraction method of traditional village buildings. This method improves Mask R-CNN to use the Path Aggregate Feature Pyramid Network (PAFPN) and Atlas Space Pyramid Pool (ASPP) to enhance the backbone model for multi-scale feature extraction and fusion.…”

Towards Sustainable Rural Revitalization: A Multidimensional Evaluation of Rural Vitality in China's Traditional Villages

“…However, rapid urbanisation and industrialisation have significantly impacted rural development, resulting in traditional villages facing various challenges, such as population exodus [9]. The collapse of numerous ancient buildings [10], the interruption of non-heritage cultural inheritance [11], and the rigid integration of old and new buildings [12] have led to the decline of rural settlements [13]. Despite the plethora of research on traditional Chinese villages, the majority remains fixated solely on heritage preservation, overlooking the broader spectrum of rural development initiatives.…”