Wild animal survey using UAS imagery and deep learning: modified Faster R-CNN for kiang detection in Tibetan Plateau

Peng, Jinbang; Wang, Dongliang; Liu, Xiaohan; Shao, Quanqin; Sun, Zhigang; Yue, Huanyin; Ye, Huping

doi:10.1016/j.isprsjprs.2020.08.026

Cited by 57 publications

(32 citation statements)

References 54 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Deep learning is a type of advanced machine-learning technique characterized by multilayer neural networks [61]. It is convenient and efficient to operate on images under different illuminations, magnifications, complex terrains, and climatic conditions, since the classification rules can be automatically learned through the continuous optimization of loss functions.…”

Section: U-net-based Deep-learning Modelmentioning

confidence: 99%

See 1 more Smart Citation

Comparing Methods for Segmenting Supra-Glacial Lakes and Surface Features in the Mount Everest Region of the Himalayas Using Chinese GaoFen-3 SAR Images

Chen

2021

Remote Sensing

View full text Add to dashboard Cite

Glaciers and numerous glacial lakes that are produced by glacier melting are key indicators of climate change. Often overlooked, supra-glacial lakes develop in the melting area in the low-lying part of a glacier and appear to be highly variable in their size, shape, and location. The lifespan of these lakes is thought to be quite transient, since the lakes may be completely filled by water and burst out within several weeks. Changes in supra-glacial lake outlines and other surface features such as supra-glacial rivers and crevasses on the glaciers are useful indicators for the direct monitoring of glacier changes. Synthetic aperture radar (SAR) is not affected by weather and climate, and is an effective tool for study of glaciated areas. The development of the Chinese GaoFen-3 (GF-3) SAR, which has high spatial and temporal resolution and high-precision observation performance, has made it possible to obtain dynamic information about glaciers in more detail. In this paper, the classical Canny operator, the variational B-spline level-set method, and U-Net-based deep-learning model were applied and compared to extract glacial lake outlines and other surface features using different modes and Chinese GF-3 SAR imagery in the Mount Everest Region of the Himalayas. Particularly, the U-Net-based deep-learning method, which was independent of auxiliary data and had a high degree of automation, was used for the first time in this context. The experimental results showed that the U-Net-based deep-learning model worked best in the segmentation of supra-glacial lakes in terms of accuracy (Precision = 98.45% and Recall = 95.82%) and segmentation efficiency, and was good at detecting small, elongated, and ice-covered supra-glacial lakes. We also found that it was useful for accurately identifying the location of supra-glacial streams and ice crevasses on glaciers, and quantifying their width. Finally, based on the time series of the mapping results, the spatial characteristics and temporal evolution of these features over the glaciers were comprehensively analyzed. Overall, this study presents a novel approach to improve the detection accuracy of glacier elements that could be leveraged for dynamic monitoring in future research.

show abstract

Section: U-net-based Deep-learning Modelmentioning

confidence: 99%

“…The first three indicators, specifically Precision (P), Recall (R), and F1 score (F1), were generated based on the True Positive (TP), True Negative (TN), False Positive (FP), and False Negative (FN) [61,70]. Precision shows the ratio of the correctly classified classes that are positive for each class.…”

Section: Accuracy Assessmentmentioning

confidence: 99%

Comparing Methods for Segmenting Supra-Glacial Lakes and Surface Features in the Mount Everest Region of the Himalayas Using Chinese GaoFen-3 SAR Images

Chen

2021

Remote Sensing

View full text Add to dashboard Cite

show abstract

“…e research of automatic detection of steel surface defects is focused on in this paper. At the first, the ResNet-50 [21][22][23][24] network is reconstructed by deformable convolution as the prefeature extraction network of Faster R-CNN [25][26][27].…”

Section: Introductionmentioning

confidence: 99%

A New Steel Defect Detection Algorithm Based on Deep Learning

Zhao

Feng

Huang

et al. 2021

Computational Intelligence and Neuroscience

124

View full text Add to dashboard Cite

In recent years, more and more scholars devoted themselves to the research of the target detection algorithm due to the continuous development of deep learning. Among them, the detection and recognition of small and complex targets are still a problem to be solved. The authors of this article have understood the shortcomings of the deep learning detection algorithm in detecting small and complex defect targets and would like to share a new improved target detection algorithm in steel surface defect detection. The steel surface defects will affect the quality of steel seriously. We find that most of the current detection algorithms for NEU-DET dataset detection accuracy are low, so we choose to verify a steel surface defect detection algorithm based on machine vision on this dataset for the problem of defect detection in steel production. A series of improvement measures are carried out in the traditional Faster R-CNN algorithm, such as reconstructing the network structure of Faster R-CNN. Based on the small features of the target, we train the network with multiscale fusion. For the complex features of the target, we replace part of the conventional convolution network with a deformable convolution network. The experimental results show that the deep learning network model trained by the proposed method has good detection performance, and the mean average precision is 0.752, which is 0.128 higher than the original algorithm. Among them, the average precision of crazing, inclusion, patches, pitted surface, rolled in scale and scratches is 0.501, 0.791, 0.792, 0.874, 0.649, and 0.905, respectively. The detection method is able to identify small target defects on the steel surface effectively, which can provide a reference for the automatic detection of steel defects.

show abstract

“…Although drones can rapidly obtain large amounts of data, processing these data is human-centric (see table 3.3 in Wich and Koh (2018)) leading to high-cost, greater time investments and difficulties in near real-time detections. As a result, there have been efforts to automate the detection of animals, humans, cars, and other objects such as nests (Longmore et al 2017;Bondi et al 2019;Bondi et al 2018;Fang et al 2016;Maire, Alvarez, and Hodgson 2015;van Gemert et al 2014;Lamba et al 2019;Peng et al 2020).…”

Section: Introductionmentioning

confidence: 99%

Video analysis for the detection of animals using convolutional neural networks and consumer-grade drones

Chalmers

Fergus

Montañez

et al. 2021

J. Unmanned Veh. Sys.

View full text Add to dashboard Cite

show abstract

Wild animal survey using UAS imagery and deep learning: modified Faster R-CNN for kiang detection in Tibetan Plateau

Cited by 57 publications

References 54 publications

Comparing Methods for Segmenting Supra-Glacial Lakes and Surface Features in the Mount Everest Region of the Himalayas Using Chinese GaoFen-3 SAR Images

Comparing Methods for Segmenting Supra-Glacial Lakes and Surface Features in the Mount Everest Region of the Himalayas Using Chinese GaoFen-3 SAR Images

A New Steel Defect Detection Algorithm Based on Deep Learning

Video analysis for the detection of animals using convolutional neural networks and consumer-grade drones

Contact Info

Product

Resources

About