A CNN-Based Method of Vehicle Detection from Aerial Images Using Hard Example Mining

Koga, Yohei; Miyazaki, Hiroyuki; Shibasaki, Ryosuke

doi:10.3390/rs10010124

Cited by 61 publications

(28 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To assess the accuracy of the results, the following accuracy coefficients were used: overall accuracy (OA), recall, precision, Cohen's kappa coefficient, and F1-score [42,43]. F1-score is the harmonic mean of precision and sensitivity and is typically used as an accuracy measure of a dichotomous model [44], so it is suitable for one-class delineation. In addition, the sum of the area resulting from errors of omission and commission (EO + EC) was analyzed to show the variant with the smallest area size error.…”

Section: Accuracy Assessment Of the Extent Of Trees And Shrubsmentioning

confidence: 99%

Using Canopy Height Model Obtained with Dense Image Matching of Archival Photogrammetric Datasets in Area Analysis of Secondary Succession

et al. 2019

View full text Add to dashboard Cite

One of the threats that has a significant impact on the conservation status and on the preservation of non-forest Natura 2000 habitats, is secondary succession, which is currently analyzed using airborne laser scanning (ALS) data. However, learning about the dynamics of this phenomenon in the past is only possible by using archival aerial photographs, which are often the only source of information about the past state of land cover. Algorithms of dense image matching developed in the last decade have provided a new quality of digital surface modeling. The aim of this study was to determine the extent of trees and shrubs, using dense image matching of aerial images. As part of a comprehensive research study, the testing of two software programs with different settings of image matching was carried out. An important step in this investigation was the quality assessment of digital surface models (DSM), derived from point clouds based on reference data for individual trees growing singly and in groups with high canopy closure. It was found that the detection of single trees provided worse results. The final part of the experiment was testing the impact of the height threshold value in elevation models on the accuracy of determining the extent of the trees and shrubs. It was concluded that the best results were achieved for the threshold value of 1.25–1.75 m (depending on the analyzed archival photos) with 10 to 30% error rate in determining the trees and shrubs cover.

show abstract

Section: Accuracy Assessment Of the Extent Of Trees And Shrubsmentioning

confidence: 99%

Using Canopy Height Model Obtained with Dense Image Matching of Archival Photogrammetric Datasets in Area Analysis of Secondary Succession

et al. 2019

View full text Add to dashboard Cite

show abstract

“…The PA, UA, OA and Kappa are commonly used measures in remote sensing [47]. The F1 score is the harmonic mean of precision and sensitivity and is usually used as an accuracy measure of a dichotomous model [48], which is suitable for one-class classification method.…”

Section: Accuracy Assessmentmentioning

confidence: 99%

A Tree Species Mapping Method from UAV Images over Urban Area Using Similarity in Tree-Crown Object Histograms

Feng

2019

Remote Sensing

View full text Add to dashboard Cite

Timely and accurate information about spatial distribution of tree species in urban areas provides crucial data for sustainable urban development, management and planning. Very high spatial resolution data collected by sensors onboard Unmanned Aerial Vehicles (UAV) systems provide rich data sources for mapping tree species. This paper proposes a method of tree species mapping from UAV images over urban areas using similarity in tree-crown object histograms and a simple thresholding method. Tree-crown objects are first extracted and used as processing units in subsequent steps. Tree-crown object histograms of multiple features, i.e., spectral and height related features, are generated to quantify within-object variability. A specific tree species is extracted by comparing similarity in histogram between a target tree-crown object and reference objects. The proposed method is evaluated in mapping four different tree species using UAV multispectral ortho-images and derived Digital Surface Model (DSM) data collected in Shanghai urban area, by comparing with an existing method. The results demonstrate that the proposed method outperforms the comparative method for all four tree species, with improvements of 0.61–5.81% in overall accuracy. The proposed method provides a simple and effective way of mapping tree species over urban area.

show abstract

“…To further compare the performance of the proposed method in this paper and other existing vehicle detection methods, a set of comparative experiments were conducted with the following nine existing vehicle detection methods: coupled region-based convolutional neural networks (CR-CNN) [49], hard example mining based convolutional neural networks (HEM-CNN) [23], affine invariant description and large-margin dimensionality reduction based method (AID-LDR) [17], bag-of-words and orientation aware scanning based method (BoW-OAS) [8], Viola-Jones based method (VJ) [18], enhanced Viola-Jones based method (EVJ) [19], fast binary detector based method (FBD) [21], YOLOv3 [59], and Faster R-CNN [36]. In the CR-CNN method, first, vehicle candidate regions are extracted based on a vehicle proposal network; then, a coupled region-based CNN is performed on the candidate regions to detect vehicles.…”

Section: Comparative Studiesmentioning

confidence: 99%

“…The output of the binary detector was further fed into a multiclass classifier for orientation and type analysis. Recently, disparity maps [22], hard example mining [23], catalog-based approach [24], and expert features [25] have also been studied for vehicle detection from remote sensing images.…”

Section: Introductionmentioning

confidence: 99%

Affine-Function Transformation-Based Object Matching for Vehicle Detection from Unmanned Aerial Vehicle Imagery

Cao

Guan

et al. 2019

Remote Sensing

View full text Add to dashboard Cite

Vehicle detection from remote sensing images plays a significant role in transportation related applications. However, the scale variations, orientation variations, illumination variations, and partial occlusions of vehicles, as well as the image qualities, bring great challenges for accurate vehicle detection. In this paper, we present an affine-function transformation-based object matching framework for vehicle detection from unmanned aerial vehicle (UAV) images. First, meaningful and non-redundant patches are generated through a superpixel segmentation strategy. Then, the affine-function transformation-based object matching framework is applied to a vehicle template and each of the patches for vehicle existence estimation. Finally, vehicles are detected and located after matching cost thresholding, vehicle location estimation, and multiple response elimination. Quantitative evaluations on two UAV image datasets show that the proposed method achieves an average completeness, correctness, quality, and F1-measure of 0.909, 0.969, 0.883, and 0.938, respectively. Comparative studies also demonstrate that the proposed method achieves compatible performance with the Faster R-CNN and outperforms the other eight existing methods in accurately detecting vehicles of various conditions.

show abstract

A CNN-Based Method of Vehicle Detection from Aerial Images Using Hard Example Mining

Cited by 61 publications

References 19 publications

Using Canopy Height Model Obtained with Dense Image Matching of Archival Photogrammetric Datasets in Area Analysis of Secondary Succession

Using Canopy Height Model Obtained with Dense Image Matching of Archival Photogrammetric Datasets in Area Analysis of Secondary Succession

A Tree Species Mapping Method from UAV Images over Urban Area Using Similarity in Tree-Crown Object Histograms

Affine-Function Transformation-Based Object Matching for Vehicle Detection from Unmanned Aerial Vehicle Imagery

Contact Info

Product

Resources

About