Journal
REMOTE SENSING
Volume 13, Issue 4, Pages -Publisher
MDPI
DOI: 10.3390/rs13040814
Keywords
urban land-use; LiDAR-aerial integration; LiDAR-aerial geo-registration; LiDAR classification; supervised machine learning; maximum likelihood; support vector machines; neural networks; bootstrap aggregation; k-fold cross-validation
Categories
Funding
- Natural Sciences and Engineering Research Council of Canada (NSERC) [RGPIN-2015-03960]
- FCE Start-up Fund of the Hong Kong Polytechnic University (BE2U)
- Research Grants Council of the Hong Kong Special Administrative Region [25213320]
Ask authors/readers for more resources
According to the World Health Organization, urban residents are expected to make up 70% of the global population by 2050. This study explores the effects of using traditional spectral signatures acquired by different sensors on the classification of LiDAR point clouds, achieving an overall classification accuracy of over 97% with the use of machine learning algorithms.
The World Health Organization has reported that the number of worldwide urban residents is expected to reach 70% of the total world population by 2050. In the face of challenges brought about by the demographic transition, there is an urgent need to improve the accuracy of urban land-use mappings to more efficiently inform about urban planning processes. Decision-makers rely on accurate urban mappings to properly assess current plans and to develop new ones. This study investigates the effects of including conventional spectral signatures acquired by different sensors on the classification of airborne LiDAR (Light Detection and Ranging) point clouds using multiple feature spaces. The proposed method applied three machine learning algorithms-ML (Maximum Likelihood), SVM (Support Vector Machines), and MLP (Multilayer Perceptron Neural Network)-to classify LiDAR point clouds of a residential urban area after being geo-registered to aerial photos. The overall classification accuracy passed 97%, with height as the only geometric feature in the classifying space. Misclassifications occurred among different classes due to independent acquisition of aerial and LiDAR data as well as shadow and orthorectification problems from aerial images. Nevertheless, the outcomes are promising as they surpassed those achieved with large geometric feature spaces and are encouraging since the approach is computationally reasonable and integrates radiometric properties from affordable sensors.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available