scholarly journals Automatic whale counting in satellite images with deep learning

2018 ◽  
Author(s):  
Emilio Guirado ◽  
Siham Tabik ◽  
Marga L. Rivas ◽  
Domingo Alcaraz-Segura ◽  
Francisco Herrera
Keyword(s):  
High Resolution ◽  
Open Data ◽  
Cost Effective ◽  
Google Earth ◽  
Deep Convolutional Neural Networks ◽  
Threat Status ◽  
High Resolution Imagery ◽  
Cost Effective Method ◽  
Great Performance ◽  
Conservation Actions

AbstractDespite their interest and threat status, the number of whales in world’s oceans remains highly uncertain. Whales detection is normally carried out from costly sighting surveys, acoustic surveys or through high-resolution orthoimages. Since deep convolutional neural networks (CNNs) achieve great performance in object-recognition in images, here we propose a robust and generalizable CNN-based system for automatically detecting and counting whales from space based on open data and tools. A test of the system on Google Earth images in ten global whale-watching hotspots achieved a performance (F1-measure) of 84% in detecting and 97% in counting 80 whales. Applying this cost-effective method worldwide could facilitate the assessment of whale populations to guide conservation actions. Free and global access to high-resolution imagery for conservation purposes would boost this process.

scholarly journals Whale counting in satellite and aerial images with deep learning

2019 ◽  
Vol 9 (1) ◽  
Author(s):  
Emilio Guirado ◽  
Siham Tabik ◽  
Marga L. Rivas ◽  
Domingo Alcaraz-Segura ◽  
Francisco Herrera
Keyword(s):  
High Resolution ◽  
Open Data ◽  
Cost Effective ◽  
Google Earth ◽  
Aerial Images ◽  
Deep Convolutional Neural Networks ◽  
Detection Model ◽  
Counting Approach ◽  
Threat Status ◽  
High Resolution Imagery

Abstract Despite their interest and threat status, the number of whales in world’s oceans remains highly uncertain. Whales detection is normally carried out from costly sighting surveys, acoustic surveys or through high-resolution images. Since deep convolutional neural networks (CNNs) are achieving great performance in several computer vision tasks, here we propose a robust and generalizable CNN-based system for automatically detecting and counting whales in satellite and aerial images based on open data and tools. In particular, we designed a two-step whale counting approach, where the first CNN finds the input images with whale presence, and the second CNN locates and counts each whale in those images. A test of the system on Google Earth images in ten global whale-watching hotspots achieved a performance (F1-measure) of 81% in detecting and 94% in counting whales. Combining these two steps increased accuracy by 36% compared to a baseline detection model alone. Applying this cost-effective method worldwide could contribute to the assessment of whale populations to guide conservation actions. Free and global access to high-resolution imagery for conservation purposes would boost this process.

scholarly journals Horizontal positional accuracy of Google Earth's imagery over rural areas: a study case in Tamaulipas, Mexico

2013 ◽  
Vol 19 (4) ◽  
pp. 588-601 ◽  
Author(s):  
Cutberto Uriel Paredes-Hernández ◽  
Wilver Enrique Salinas-Castillo ◽  
Francisco Guevara-Cortina ◽  
Xicoténcatl Martínez-Becerra
Keyword(s):  
High Resolution ◽  
Rural Areas ◽  
Google Earth ◽  
Positional Accuracy ◽  
High Resolution Imagery ◽  
Check Points ◽  
Class 1 ◽  
Study Case ◽  
Source Of Information

Due to the popularity of Google Earth (GE), users commonly assume that it is a credible and accurate source of information. Consequently, GE's imagery is frequently used in scientific and others projects. However, Google states that data available in their geographic products are only approximations and, therefore, their accuracy is not officially documented. In this paper, the horizontal positional accuracy of GE's imagery is assessed by means of comparing coordinates extracted from a rural cadastral database against coordinates extracted from well-defined and inferred check points in GE's imagery. The results suggest that if a large number of well-defined points are extracted from areas of high resolution imagery, GE's imagery over rural areas meets the horizontal accuracy requirements of the ASPRS for the production of "Class 1" 1:20,000 maps. Nonetheless, the results also show that georegistration and large horizontal errors occur in GE's imagery. Consequently, despite its overall horizontal positional accuracy, coordinates extracted from GE's imagery should be used with caution.

scholarly journals Comparison of multiple glacier inventories with a new inventory derived from high-resolution ALOS imagery in the Bhutan Himalaya

2016 ◽  
Vol 10 (1) ◽  
pp. 65-85 ◽  
Author(s):  
H. Nagai ◽  
K. Fujita ◽  
A. Sakai ◽  
T. Nuimura ◽  
T. Tadono
Keyword(s):  
High Resolution ◽  
Google Earth ◽  
Data Sets ◽  
Data Set ◽  
Glacier Inventory ◽  
Mountain Ranges ◽  
High Resolution Imagery ◽  
Overlap Ratio ◽  
Geographical Locations

Abstract. Digital glacier inventories are invaluable data sets for revealing the characteristics of glacier distribution and for upscaling measurements from selected locations to entire mountain ranges. Here, we present a new inventory of Advanced Land Observing Satellite (ALOS) imagery and compare it with existing inventories for the Bhutan Himalaya. The new inventory contains 1583 glaciers (1487 ± 235 km2), thereof 219 debris-covered glaciers (951 ± 193 km2) and 1364 debris-free glaciers (536 ± 42 km2). Moreover, we propose an index for quantifying consistency between two glacier outlines. Comparison of the overlap ratio demonstrates that the ALOS-derived glacier inventory contains delineation uncertainties of 10–20 % which depend on glacier size, that the shapes and geographical locations of glacier outlines derived from the fourth version of the Randolph Glacier Inventory have been improved in the fifth version, and that the latter is consistent with other inventories. In terms of whole glacier distribution, each data set is dominated by glaciers of 1.0–5.0 km2 area (31–34 % of the total area), situated at approximately 5400 m elevation (nearly 10 % in 100 m bin) with either north or south aspects (22 and 15 %). However, individual glacier outlines and their area exhibit clear differences among inventories. Furthermore, consistent separation of glaciers with inconspicuous termini remains difficult, which, in some cases, results in different values for glacier number. High-resolution imagery from Google Earth can be used to improve the interpretation of glacier outlines, particularly for debris-covered areas and steep adjacent slopes.

scholarly journals Road Extraction from Very High Resolution Images Using Weakly labeled OpenStreetMap Centerline

2019 ◽  
Vol 8 (11) ◽  
pp. 478 ◽  
Author(s):  
Songbing Wu ◽  
Chun Du ◽  
Hao Chen ◽  
Yingxiao Xu ◽  
Ning Guo ◽  
...  
Keyword(s):  
Deep Learning ◽  
High Resolution ◽  
Road Networks ◽  
Semantic Segmentation ◽  
Google Earth ◽  
Training Data ◽  
City Management ◽  
High Quality ◽  
High Resolution Imagery ◽  
Very High

Road networks play a significant role in modern city management. It is necessary to continually extract current road structure, as it changes rapidly with the development of the city. Due to the success of semantic segmentation based on deep learning in the application of computer vision, extracting road networks from VHR (Very High Resolution) imagery becomes a method of updating geographic databases. The major shortcoming of deep learning methods for road networks extraction is that they need a massive amount of high quality pixel-wise training datasets, which is hard to obtain. Meanwhile, a large amount of different types of VGI (volunteer geographic information) data including road centerline has been accumulated in the past few decades. However, most road centerlines in VGI data lack precise width information and, therefore, cannot be directly applied to conventional supervised deep learning models. In this paper, we propose a novel weakly supervised method to extract road networks from VHR images using only the OSM (OpenStreetMap) road centerline as training data instead of high quality pixel-wise road width label. Large amounts of paired Google Earth images and OSM data are used to validate the approach. The results show that the proposed method can extract road networks from the VHR images both accurately and effectively without using pixel-wise road training data.

scholarly journals Thickness estimation of supraglacial debris above ice cliff exposures using a high-resolution digital surface model derived from terrestrial photography

2017 ◽  
Vol 63 (242) ◽  
pp. 989-998 ◽  
Author(s):  
L. NICHOLSON ◽  
J. MERTES
Keyword(s):  
High Resolution ◽  
Cost Effective ◽  
Ablation Rate ◽  
Surface Model ◽  
Digital Surface Model ◽  
High Frequency Radar ◽  
Ice Surface ◽  
Cost Effective Method ◽  
Debris Cover ◽  
Glacier Ice

ABSTRACTThe thickness of supraglacial debris cover controls how it impacts the ablation rate of underlying glacier ice, yet this quantity remains challenging to measure, particularly at glacier scales. We present a relatively straightforward, and cost-effective method to estimate debris thickness exposed above ice cliffs using simplified geometrical measurements from a high-resolution digital surface model (DSM), derived from a terrestrial photographic survey and a Structure from Motion with Multi-View Stereo workflow (SfM-MVS). As the ice surface relief beneath the debris cover is unknown, we assume it to be horizontal and provide error bounds based on characteristic ice-surface slope at the visible debris/ice interface. Debris thickness around the three sampled ice cliffs was highly variable (interquartile range of 0.80–2.85 m) and negatively skewed with a mean thickness of 2.08 ± 0.68 m. Manual, and high-frequency radar, determinations of debris thickness in the same area show similar thickness distributions, but statistically different mean debris thickness, due to local heterogeneity. Debris thickness values derived in this study all exceed estimates from satellite surface temperature inversions. Wider application of the method presented here would provide useful data for improving debris thickness approximations from satellite imagery.

Developing a Google Earth-Based Method to Measure Sight Distance for U-Turns at Unsignalized Intersections on Multilane Divided Highways

2021 ◽  
pp. 036119812110363
Author(s):  
Li Quan ◽  
Beijia Zhang ◽  
Huaguo Zhou ◽  
Pan Liu
Keyword(s):  
Grid Cell ◽  
Grid Method ◽  
Field Measurements ◽  
Cost Effective ◽  
Ground Level ◽  
Google Earth ◽  
Horizontal Curves ◽  
Geometric Conditions ◽  
Sight Distance ◽  
Cost Effective Method

The median U-turn is one of the most popular access management techniques implemented by state transportation agencies. Sight distance (SD) is among the key elements for the safe operation of median U-turns. The traditional method of measuring SD is using sight rods in the field, which is inconvenient and unsafe; thus, it is important to develop a convenient and safer method to measure SD for U-turn movement. The objective of this study is to develop a safe and cost-effective method to measure SD for U-turn vehicles, without going into the field, using existing tools in Google Earth and a perspective grid method that can be generated by Kinovea software. The method can be used in two different road geometric conditions: (i) a straight segment with crest curves; (ii) a roadway with a combination of crest curves and horizontal curves. For Condition (i), an empirical equation was developed to estimate SD as a function of distance and elevation along the sight line measured using Google Earth. For Condition (ii), a perspective grid was first applied using Google Earth ground-level view; then the number of broken lines and gaps in each perspective grid cell was used to measure the SD. All inputs can be collected using Google Earth. Field measurements of SD for U-turns at 10 selected locations were conducted to verify the validity of the method. The results show that the differences between the model estimates and field measurements are all less than 10%.

A cost-effective method for tranquility mapping using open environmental data

2018 ◽  
Vol 47 (3) ◽  
pp. 417-436 ◽  
Author(s):  
Maria Pafi ◽  
Christos Chalkias ◽  
Demetris Stathakis
Keyword(s):  
Analytic Functions ◽  
Cost Effective ◽  
Google Earth ◽  
Kappa Coefficient ◽  
Environmental Data ◽  
Geospatial Data ◽  
Preference Data ◽  
Geostatistical Methods ◽  
Cost Effective Method ◽  
Extended Zone

In this study, we propose a cost-effective method for tranquillity mapping using multi-criteria analysis and open geospatial data. We apply this method in an extended zone around a major Greek highway trespassing areas of high natural value. Composite criteria are developed through analytic functions and geostatistical methods to reflect either barriers or enablers of tranquillity. The results indicate that it is possible to identify tranquility zones which are spatially plausible. To verify the validity of the results, we calculate the Kappa coefficient (0.71) and the overall accuracy (80%) using preference data obtained from non-specialized photo-interpreters in a sample of places on Google Earth. We believe that this method can inform planning, especially in countries with a weak landscape policy.

Sign in / Sign up

Export Citation Format

Share Document