DETECTION OF PEOPLE AND VEHICLES USING VERY HIGH-RESOLUTION SATELLITE IMAGES
The result's identifiers
Result code in IS VaVaI
<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F61989100%3A27350%2F21%3A10249985" target="_blank" >RIV/61989100:27350/21:10249985 - isvavai.cz</a>
Result on the web
<a href="https://www.scopus.com/record/display.uri?eid=2-s2.0-85131676985&origin=resultslist&sort=plf-f&featureToggles=FEATURE_NEW_DOC_DETAILS_EXPORT:1" target="_blank" >https://www.scopus.com/record/display.uri?eid=2-s2.0-85131676985&origin=resultslist&sort=plf-f&featureToggles=FEATURE_NEW_DOC_DETAILS_EXPORT:1</a>
DOI - Digital Object Identifier
<a href="http://dx.doi.org/10.5593/sgem2021/2.1/s10.65" target="_blank" >10.5593/sgem2021/2.1/s10.65</a>
Alternative languages
Result language
angličtina
Original language name
DETECTION OF PEOPLE AND VEHICLES USING VERY HIGH-RESOLUTION SATELLITE IMAGES
Original language description
Processing of satellite data with high and very high spatial resolution enables to monitor selected social activities in open land, such as the occurrence and movement of people, cars, ships and other transport means. Such information is important for mass event control, urban and transport planning, intelligent transport systems, emergency, epidemiologic control, military application etc. The pilot area is located in Prague, close to the Old Town Square. Panchromatic and multispectral imagery from WorldView3 was processed. After radiometric and atmospheric corrections pan-sharpening was applied to improve visual interpretation. Based on object classification approach the following algorithms for segmentation were tested - edge and intensity algorithms, merging using Full Lambda Schedule and Fast Lambda with variable settings. Created list of classes served as a source for selection of representative samples. Suitable spatial, spectral and texture attributes were selected for object-oriented classification using K Nearest Neighbor, Support Vector Machine and Principal Components Analysis. All object classification results were compared and evaluated. Also, convolutional neural networks (YOLOV5) were tested. Segmentation and classification results were validated using results of manual vectorization of target features (vehicles, people) performed by four persons. The detection of vehicles was relatively successful, especially in open places without shade or vegetation. A small size of people's vertical projection influencing only individual, often mixed, pixels caused difficulties in detection of people. While detection of individual people is not satisfactory yet, classification of people clusters provides promising results, especially in open land such as large streets or squares. (C) 2021 International Multidisciplinary Scientific Geoconference. All rights reserved.
Czech name
—
Czech description
—
Classification
Type
D - Article in proceedings
CEP classification
—
OECD FORD branch
10500 - Earth and related environmental sciences
Result continuities
Project
—
Continuities
S - Specificky vyzkum na vysokych skolach
Others
Publication year
2021
Confidentiality
S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů
Data specific for result type
Article name in the collection
21st International Multidisciplinary Scientific GeoConference SGEM 2021, 16 - 22 August, 2021, Albena, Bulgaria
ISBN
978-619-7603-22-4
ISSN
1314-2704
e-ISSN
—
Number of pages
9
Pages from-to
283-291
Publisher name
STEF92 Technology Ltd.
Place of publication
Sofia
Event location
Albena
Event date
Aug 16, 2021
Type of event by nationality
WRD - Celosvětová akce
UT code for WoS article
—