This study is a pilot project for the “San Francisco Flood Plain Project” (SFFPP), meant to delimit flood plain areas owned by the Brazilian federal government. The objective is to determine the attainable accuracy in river water surface delineation using satellite imagery from Landsat, Sentinel-1 and -2. We prioritize the evaluation of Landsat data due to its long systematical time series, allowing hydrological analysis requiring observations of at least 40 to 60 years and data from the Sentinel missions with their high frequency of revisit, improved spatial resolution (compared with Landsat) and possibility of observation in wet season (S-1). In our approach, we evaluated the accuracy by spectral bands individually and in combination, as well as polarization. We also tested a number of thematic information extraction techniques unsupervised (K-means and EM Cluster Analysis) and supervisioned (Random Forest - RF, k-nearest neighbors - KNN, Maximum Likelihood Classification – ML, Support Vector Machine – SVM, Mahalanobis). To validate our results, we used a PlanetScope mosaic (3 m). Results indicate that shortwave infrared bands have a higher capacity to separate water surface from other classes. For SAR data, the best separation was obtained by VV polarization (compared with VH). Techniques all reached agreement values >94% for the Sentinel-2 image, >93% for the Sentinel-1 image and >;86% for the Landsat-8. We consider both methodologies effectives to extract the water surface and appropriate for the real estate issues of the SFFPP project.
This article describes the “Sao Francisco Floodplain Project” (SFFPP) aiming at defining a Mean Ordinary Flood Line (MOFL) on the banks of the Sao Francisco River, in East Central Brazil. Land inserted within the MOFL of large rivers in Brazil are characterized as government-owned and are ruled by a special legislation. The lack of consensus for an effective method for this delimitation has raised much conflicts between dwellers, land owners and the federal government. To solve this, the SFFPP first aims to determine the mean flood plain level using historical water level data and then find all the dates since the launch of the first satellite sensor with a 30 m resolution (Landsat-4, 1982) corresponding to these particular water levels with a small margin of error. All Landsat images corresponding to these dates were acquired to produce a delineation of the MOFL. In a thorough series of tests to extract the water surface, the K-means segmentation using the shortwave infrared band of Landsat yielded the best results. A first refining of the MOFL was performed by interpolating the Landsat bands to improve the smoothness of the waterline. This refinement reduced the average distance error between the Landsat water edge and the true water edge from 9 to 7.5 meters. Then, some sections of the MOFL was further completed or refined using high-resolution multi-source satellite images where available. These first results were very encouraging and we were able to acquire Landsat images for each section of the river corresponding to the mean flood water level. Because Brazil has been suffering a significant reduction in rainfall since 2013, no recent SAR or optical images such as Sentinel-1 and -2 could be used. Analysis of the water level time series confirmed an alarming decreasing trend in the water discharge of the Sao Francisco River.
A number of methods have been developed for the automatic identification and delineation of individual tree crowns from high spatial resolution satellite image to provide support for the management and maintenance of forests both in natural and urban environments. In this paper we present a method that integrates a Marked Point Processes (MPP) model and Template Matching (TM) to extract individual tree crowns in two tropical environments. The MPP is an extension of Markov random fields in which objects are defined by their position within a space of possible positions and their marks (e.g. shape). The MPP has been increasingly used for the recognition of objects but most implementation use an oversimplified model as mark. We argue that the MPP could take better advantage of the geometry of trees by incorporating a three-dimensional model as a mark. Conversely, TM is an approach to pattern recognition that takes the characteristics of the objects into account. Our method uses cross-correlation for determining which objects have been correctly targeted by the MPP. The correlation between the illuminated 3D crown model and the image is an inheritance from TM. The methodology was applied in synthetic images and sub-images of the WorldView satellite in two different contexts in Brazil. The results are validated by counting the correctly identified trees and by comparing their size with our interpreted version. Results are encouraging with 65 to 90% of correctly identified trees. The most difficult cases are mostly related to the existence of clustered tree crowns.
With the availability of high-resolution satellite data, much research has been focused on the automatic detection and classi cation of individual tree crowns. Most of these studies were applied to temperate climates of the northern hemisphere, especially for forests of coniferous. Very few studies have been applied to the detection of trees in the tropical regions, least of all in the urban environment. Urban trees play a major role in maintaining or even improving the quality of life in cities by their contribution to the quality of the air, by absorbing rain water, by refreshing the air through transpiration and providing shadow. In this study we explored the potential of high-resolution WorldView-2 satellite data for the identi cation of urban individual tree crowns in the city of Belo Horizonte, Minas Gerais, Brazil, through an object-oriented approach. Irrelevant areas were masked (e.g. buildings, asphalt, shadows, exposed soil) using a threshold of NDVI. Three di erent approaches were tested to isolate and delineate individual tree crowns: region growing, watershed and template matching. For the rst two approaches several parameters were tested to nd the best result for the isolation of the individual tree crowns. An in-house program has been developed for template matching using a set of seven di erent templates of di erent species. A set of 300 individual tree crowns were visually interpreted in the WorldView-2 image to serve as validation and to compare the performance of the three di erent approaches. Then, the comparison was performed between the visual interpretation and the results of each approach by calculating the di erence between the areas as a ratio of the validated area. Our results show that the region growing approach provided the best results, with an accuracy of over 80%.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.