NEURAL NETWORK RECOGNITION OF BUILDING OBJECTS IN AERIAL IMAGES
DOI:
https://doi.org/10.32782/IT/2023-1-5Keywords:
convolutional neural networks, image segmentation, recognition, aerial image, computer visionAbstract
Automated recognition of building objects based on aerial image data is a complex problem in computer vision. It is related to variations in the appearance of buildings and the semantic characteristics of scenes in urban areas. This paper proposes a method of automated recognition of building objects on digital aerial photographs based on 2D-CNN. The first step is to download and process high spatial resolution images from unmanned aerial vehicles. Then classification and segmentation of the image are performed based on 2D-CNN neural network architecture with softmax function for the output layer and rectified linear block (ReLu) for the remaining layers. Convolutional layers apply filters to all pixels of the input image to obtain a set of high-level abstract features. Data segmentation was performed to classify each pixel from the UAV images, where the field of view (fov) for each image is considered a sliding window of size 3×3 of the input data. It made it possible to determine whether the object of the researched scene belongs to a certain class. The last step is the binary mask creation of the building object recognition based on the cross-entropy loss function. The neural network was trained at the pixel level, which allowed to increase the accuracy of identification of building objects and reduce the number of misclassified zones. Experimental results showed a significant improvement in the accuracy of building recognition in a publicly available dataset. Specifically, the OA, AA, and K metrics improved by 2.6%, 5.6%, and 3.2%, respectively, for the training dataset and by 1.2%, 1.8%, and 1.5% for the test dataset.
References
Kalantar B., Halin A.A., Al-Najjar H.A.H., Mansor S., van Genderen J.L., Shafri H.Z.M., Zand, M. A Framework for Multiple Moving Objects Detection in Aerial Videos. In Spatial Modeling in GIS and R for Earth and Environmental Sciences. Elsevier: Amsterdam, The Netherlands. 2019. P. 573–588.
Kalantar B., Mansor S.B., Halin A.A., Shafri H.Z.M., Zand, M. Multiple moving object detection from UAV videos using trajectories of matched regional adjacency graphs. IEEE Trans. Geosci. Remote Sens. 2017. Vol. 55. P.5198–5213.
Kalantar B., Mansor S.B., Sameen M.I., Pradhan B., Shafri H.Z.M. Drone-based land-cover mapping using a fuzzy unordered rule induction algorithm integrated into object-based image analysis. Int. J. Remote Sens. 2017. Vol.38. P. 2535–2556.
Waleed A., Turky A., Basil D. Detecting Buildings and Nonbuildings from Satellite Images Using U-Net. Computational Intelligence and Neuroscience, 2018. P.1687-5265. DOI: 10.1155/2022/4831223.
Krówczyńska M., Raczko E., Staniszewska N., Wilk, E. Asbestos–cement roofing identification using remote sensing and convolutional neural networks (CNNs). Remote Sens. 2020. Vol. 12, P.2-16.
Cheng G., Han J., Lu X. Remote sensing image scene classifcation: Benchmark and state of the art. Proceedings of the IEEE, 2017. Vol.105 (10). P. 1865-188.
Avudaiamma R., Dayana S., Prabhu R. and Swarnalatha A. Automatic Building Extraction from VHR Satellite Image. International Conference on Current Trends towards Converging Technologies (ICCTCT). 2018. P. 1-6, DOI: 10.1109/ICCTCT.2018.8551119.
Long J., Shelhamer E., Darrell T. Fully convolutional networks for semantic segmentation. In Proceedings of the 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Boston, MA, USA. 2015. P. 3431–3440.
Liu L., Ouyang W., Wang X., Fieguth W.P., Chen J., Liu X., Pietikinen M. Deep learning for generic object detection. A survey. I International Journal of Computer Vision. 2020. Vol. 128. P. 261–31.
Yuan J. Learning building extraction in aerial scenes with convolutional networks. IEEE Trans. Pattern Anal. Mach. Intell. 2017. P. 2793–2798.
Xu Y., Wu L., Xie Z., Chen Z. Building extraction in very high resolution remote sensing imagery using deep learning and guided filters. Remote Sens. 2018. Vol. 10. P. 144.
Huang J., Zhang X., Xin Q., Sun Y., Zhang P. Automatic building extraction from high-resolution aerial images and LiDAR data using gated residual refinement network. ISPRS J. Photogramm. Remote Sens. 2019. P. 91–105.
Kim E.K., Lee H., Kim J.Y., Kim S. Data Augmentation Method by Applying Color Perturbation of Inverse PSNR and Geometric Transformations for Object Recognition Based on Deep Learning. Appl. Sci. 2020. Vol. 10. P. 3755. DOI: 10.3390/app10113755
Yang J., Guo J., Yue H., Liu Z., Hu H., Li K. CDnet: CNN-Based Cloud Detection for Remote Sensing Imagery. IEEE Trans. Geosci. Remote Sens. 2019. Vol. 57. P. 6195–6211.
Rahman T., Chowdhury M.E., Khandakar A., Islam K.R., Islam K.F., Mahbub Z.B., Kadir M.A., Kashem S. Transfer Learning with Deep Convolutional Neural Network (CNN) for Pneumonia Detection using Chest X-ray. Appl. Sci. 2020. Vol. 10. P. 32-33.
Mäyrä J., Keski-Saari S., Kivinen S., Tanhuanpää T., Hurskainen P., Kullberg P., Poikolainen L., Viinikka A., Tuominen S., Kumpula T., et al. Tree species classification from airborne hyperspectral and LiDAR data using 3D convolutional neural networks. Remote Sens. Environ. 2021. Vol.256. P. 112-322.
Каштан В.Ю., Гнатушенко В.В. Контурна сегментація цифрових супутникових знімків з використанням особливих точок вейвлет-перетворення. Системні технології. Регіональний міжвузівський збірник наукових робіт. 2019. Випуск 1 (120). С.3 - 11.
Yılmaz I., Imamo˘glu M., Özbulak G., Kahraman F., Aptoula E. Large Scale Crop Classification from Multi-temporal and ˙Multi-spectral Satellite Images. In Proceedings of the 2020 28th Signal Processing and Communications Applications Conference (SIU). 2020. P. 1–4.
Cheng H., Lian D., Gao S., Geng Y. Evaluating Capability of Deep Neural Networks for Image Classification via Information Plane. In Proceedings of the European Conference on Computer Vision (ECCV). 2018. P. 168–182.
