Scholarly record
USABILITY EVALUATION OF CONVOLUTIONAL NEURAL NETWORKS IN ROADS EXTRACTION FROM AERIAL IMAGES
Abstract
The main source of information needed to build, update and maintain spatial databases are aerial images. The biggest advance of aerial images is their capability to efficiently cover large areas. With the continuous development of technology, aerial imaging has become cheaper and more affordable than any time before. With this advancement of technology, large amounts of data need to be processed. In order to extract information from aerial images, they need to be processed into ortho-rectified images and then vectorized. The process of vectorization (extracting information from images and creating spatial structures) is mainly done by human operator. With the amount of the data and demand for short processing period in order to guarantee information recency, this process needs to be automated. Neural networks have proven their great usability in wide range of applications varying from stock estimation to autonomously driven vehicles. Convolutional neural networks create set of neural networks with specialization on computer vision. Their ability to correctly distinguish between multiple object presented in the images have been evaluated. We designed architecture of convolutional neural network that is suitable for road extraction from aerial images. The proposed architecture was tested in the area of city PieпїЅ?any, Slovakia. The task designed for convolutional neural network is to recognize roads in the aerial images and label pixels that are parts of the road segments. The accuracy over 85% achieved during the experiment indicates great potential in using convolutional neural networks in objects extraction from aerial images.
Publication Impact Profile
Publication details
References10
Kluckner S., Bischof H., Semantic classification by covariance descriptors within a randomized forest, Computer Vision Workshops (ICCV), IEEE, 2009, pp 665-672.
Kluckner S., Mauthner T., Roth P.M., Bischof H., Semantic classification in aerial imagery by integrating appearance and height information, Lecture Notes in Computer Science, vol. 5995, 2009, pp 477-488.
Mnih V., Machine Learning for Aerial Image Labeling, PhD. Thesis, Canada, Toronto, 2013.
Wei H., Yangyu H., Li W., Fan Z., Hengchao Li., Deep Convolutional Neural Networks for Hyperspectral Image Classification, Journal of Sensors, Hindawi Publishing Corporation, China, 2015.
Sermanet P., LeCun Y., Traffic sign recognition with multi-scale convolutional networks, Proceedings of the International Joint Conference on Neural Network (IJCNN`11), IEEE, 2011, pp 2809-2813.
Sainath T.N., Mohamed A.-R., Kingsbury B., Ramabhadran B., Deep convolutional neural networks for LVCSR, Proceedings of the 38th IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP`13), pp 8614-8618, Canada 2013.
Lecun Y., Bengio Y., Hinton G., Deep learning, Nature, Volume 521, Issue 7553, 2015, pp. 436-444.
Srivastava N., Hinton G., Krizhevsky A., Sutskever I., Salakhutdinov R., Dropout: A Simple Way to Prevent Neural Networks from Overfitting, Journal of Machine Learning Research 15, 2014, pp 1929-1958.
Kingma D.P., Ba J., Adam: A Method for Stochastic Optimization, 3rd International Conference for Learning Representations, USA, 2015.
Brown C. D., Davis H. T., Receiver operating characteristics curves and related decision measures: A tutorial, Chemometrics and Intelligent Laboratory Systems, Volume 80, Issue 1, 2006, pp 24 38.
View or Download full articleAccess options
SWS access login
Login as SWS Scientific CommitteeLogin as SWS Scientific PartnerLogin as SWS AuthorAuthors and approved SWS contributors will read and export their own linked papers after identity matching by SWS profile, email and SGEM GlobalID.
For librarian assistance: [email protected]
Purchase Instant Access
- Article can be downloaded after successful payment.
- Article may be used according to SWS library access terms.
- Article cannot be redistributed.
