skip to main content

Title: Roof Shape Classification from LiDAR and Satellite Image Data Fusion Using Supervised Learning
Geographic information systems (GIS) provide accurate maps of terrain, roads, waterways, and building footprints and heights. Aircraft, particularly small unmanned aircraft systems (UAS), can exploit this and additional information such as building roof structure to improve navigation accuracy and safely perform contingency landings particularly in urban regions. However, building roof structure is not fully provided in maps. This paper proposes a method to automatically label building roof shape from publicly available GIS data. Satellite imagery and airborne LiDAR data are processed and manually labeled to create a diverse annotated roof image dataset for small to large urban cities. Multiple convolutional neural network (CNN) architectures are trained and tested, with the best performing networks providing a condensed feature set for support vector machine and decision tree classifiers. Satellite image and LiDAR data fusion is shown to provide greater classification accuracy than using either data type alone. Model confidence thresholds are adjusted leading to significant increases in models precision. Networks trained from roof data in Witten, Germany and Manhattan (New York City) are evaluated on independent data from these cities and Ann Arbor, Michigan.  more » « less
Award ID(s):
Author(s) / Creator(s):
Date Published:
Journal Name:
Page Range / eLocation ID:
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Abstract

    Information on flood inundation extent is important for understanding societal exposure, water storage volumes, flood wave attenuation, future flood hazard, and other variables. A number of organizations now provide flood inundation maps based on satellite remote sensing. These data products can efficiently and accurately provide the areal extent of a flood event, but do not provide floodwater depth, an important attribute for first responders and damage assessment. Here we present a new methodology and a GIS‐based tool, the Floodwater Depth Estimation Tool (FwDET), for estimating floodwater depth based solely on an inundation map and a digital elevation model (DEM). We compare the FwDET results against water depth maps derived from hydraulic simulation of two flood events, a large‐scale event for which we use medium resolution input layer (10 m) and a small‐scale event for which we use a high‐resolution (LiDAR; 1 m) input. Further testing is performed for two inundation maps with a number of challenging features that include a narrow valley, a large reservoir, and an urban setting. The results show FwDET can accurately calculate floodwater depth for diverse flooding scenarios but also leads to considerable bias in locations where the inundation extent does not align well with the DEM. In these locations, manual adjustment or higher spatial resolution input is required.

    more » « less
  2. Abstract The Local Climate Zone (LCZ) classification is already widely used in urban heat island and other climate studies. The current classification method does not incorporate crucial urban auxiliary GIS data on building height and imperviousness that could significantly improve urban-type LCZ classification utility as well as accuracy. This study utilized a hybrid GIS- and remote sensing imagery-based framework to systematically compare and evaluate different machine and deep learning methods. The Convolution Neural Network (CNN) classifier outperforms in terms of accuracy, but it requires multi-pixel input, which reduces the output’s spatial resolution and creates a tradeoff between accuracy and spatial resolution. The Random Forest (RF) classifier performs best among the single-pixel classifiers. This study also shows that incorporating building height dataset improves the accuracy of the high- and mid-rise classes in the RF classifiers, whereas an imperviousness dataset improves the low-rise classes. The single-pass forward permutation test reveals that both auxiliary datasets dominate the classification accuracy in the RF classifier, while near-infrared and thermal infrared are the dominating features in the CNN classifier. These findings show that the conventional LCZ classification framework used in the World Urban Database and Access Portal Tools (WUDAPT) can be improved by adopting building height and imperviousness information. This framework can be easily applied to different cities to generate LCZ maps for urban models. 
    more » « less
  3. null (Ed.)
    Recent advances in big spatial data acquisition and deep learning allow novel algorithms that were not possible several years ago. We introduce a novel inverse procedural modeling algorithm for urban areas that addresses the problem of spatial data quality and uncertainty. Our method is fully automatic and produces a 3D approximation of an urban area given satellite imagery and global-scale data, including road network, population, and elevation data. By analyzing the values and the distribution of urban data, e.g., parcels, buildings, population, and elevation, we construct a procedural approximation of a city at a large-scale. Our approach has three main components: (1) procedural model generation to create parcel and building geometries, (2) parcel area estimation that trains neural networks to provide initial parcel sizes for a segmented satellite image of a city block, and (3) an optional optimization that can use partial knowledge of overall average building footprint area and building counts to improve results. We demonstrate and evaluate our approach on cities around the globe with widely different structures and automatically yield procedural models with up to 91,000 buildings, and spanning up to 150 km 2 . We obtain both a spatial arrangement of parcels and buildings similar to ground truth and a distribution of building sizes similar to ground truth, hence yielding a statistically similar synthetic urban space. We produce procedural models at multiple scales, and with less than 1% error in parcel and building areas in the best case as compared to ground truth and 5.8% error on average for tested cities. 
    more » « less
  4. Over the past several decades, urban planning has considered a variety of advanced analysis methods with greater and lesser degrees of adoption. Geographic Information Systems (GIS) is probably the most notable, with others such as database management systems (DBMS), decision support systems (DSS), planning support systems (PSS), and expert systems (ES), having mixed levels of recognition and acceptance (Kontokosta, C. E. (2021). Urban informatics in the science and practice of planning. Journal of Planning Education and Research, 41(4), 382–395. doi:10.1177/0739456X18793716; Yigitcanlar, T., Desouza, K. C., Butler, L., & Roozkhosh, F. (2020). Contributions and risks of artificial intelligence (AI) in building smarter cities: Insights from a systematic review of the literature. Energies, 13(6), 1473). Advances in information technologies have moved very slowly in the field of urban planning, more recently concerning ‘smart city’ technologies while revolutionizing other domains, such as consumer goods and services. Baidu, Amazon, Netflix, Google, and many others are using these technologies to gain insights into consumer behaviour and characteristics and improve supply chains and logistics. This is an opportune time for urban planners to consider the application of AI-related techniques given vast increases in data availability, increased processing speeds, and increased popularity and development of planning related applications. Research on these topics by urban planning scholars has increased over the past few years, but there is little evidence to suggest that the results are making it into the hands of professional planners (Batty, M. (2018). Artificial intelligence and smart cities. Environment and Planning B: Urban Analytics and City Science, 45(1), 3–6; Batty, M. (2021). Planning education in the digital age. Environment and Planning B: Urban Analytics and City Science, 48(2), 207–211). Others encourage planners to leverage the ubiquity of data and advances in computing to enhance redistributive justice in information resources and procedural justice in decision-making among marginalized communities (Boeing, G., Besbris, M., Schachter, A., & Kuk, J. (2020). Housing search in the Age of Big data: Smarter cities or the same Old blind spots? Housing Policy Debate, 31(1), 112–126; Goodspeed, R. (2015). Smart cities: Moving beyond urban cybernetics to tackle wicked problems. Cambridge journal of regions, Economy and Society, 8(1), 79–92). This article highlights findings from a recent literature review on AI in planning and discusses the results of a national survey of urban planners about their perspectives on AI adoption and concerns they have expressed about its broader use in the profession. Currently, the outlook is mixed, matching how urban planners initially viewed the early stages of computer adoption within the profession. And yet today, personal computers are essential to any job. 
    more » « less
  5. Abstract

    Marine megafauna are difficult to observe and count because many species travel widely and spend large amounts of time submerged. As such, management programmes seeking to conserve these species are often hampered by limited information about population levels.

    Unoccupied aircraft systems (UAS, aka drones) provide a potentially useful technique for assessing marine animal populations, but a central challenge lies in analysing the vast amounts of data generated in the images or video acquired during each flight. Neural networks are emerging as a powerful tool for automating object detection across data domains and can be applied to UAS imagery to generate new population‐level insights. To explore the utility of these emerging technologies in a challenging field setting, we used neural networks to enumerate olive ridley turtlesLepidochelys olivaceain drone images acquired during a mass‐nesting event on the coast of Ostional, Costa Rica.

    Results revealed substantial promise for this approach; specifically, our model detected 8% more turtles than manual counts while effectively reducing the manual validation burden from 2,971,554 to 44,822 image windows. Our detection pipeline was trained on a relatively small set of turtle examples (N = 944), implying that this method can be easily bootstrapped for other applications, and is practical with real‐world UAS datasets.

    Our findings highlight the feasibility of combining UAS and neural networks to estimate population levels of diverse marine animals and suggest that the automation inherent in these techniques will soon permit monitoring over spatial and temporal scales that would previously have been impractical.

    more » « less