Unmanned Aerial Images Research Articles

Automatic Road Pavement Distress Recognition Using Deep Learning Networks from Unmanned Aerial Imagery

Detecting and recognizing distress types on road pavement is crucial for selecting the most appropriate methods to repair, maintain, prevent further damage, and ensure the smooth functioning of daily activities. However, this task presents challenges, such as dealing with crowded backgrounds, the presence of multiple distress types in images, and their small sizes. In this study, the YOLOv8 network, a cutting-edge single-stage model, is employed to recognize seven common pavement distress types, including transverse cracks, longitudinal cracks, alligator cracks, oblique cracks, potholes, repairs, and delamination, using a dataset comprising 5796 terrestrial and unmanned aerial images. The network’s optimized architecture and multiple convolutional layers facilitate the extraction of high-level semantic features, enhancing algorithm accuracy, speed, and robustness. By combining high and low semantic features, the network achieves improved accuracy in addressing challenges and distinguishing between different distress types. The implemented Convolutional Neural Network demonstrates a recognition precision of 77%, accuracy of 81%, mAP of 79%, f1-score of 74%, and recall of 75%, underscoring the model’s effectiveness in recognizing various pavement distress forms in both aerial and terrestrial images. These results highlight the model’s satisfactory performance and its potential for effectively recognizing and categorizing pavement distress for efficient infrastructure maintenance and management.

DEVELOPING A TECHNOLOGY FOR PRODUCING DRONE-BORNE HYPERSPECTRAL IMAGES TO MONITOR LARGE-AREA MIXED HERITAGE

Abstract. This study compared differences in the shooting and pre-processing techniques between manned and unmanned aerial images, and verified the precision of the images by comparing the ground sample distance between them. GSD of manned aerial image taken at high altitude could not discern tree shapes from 150 cm per pixel, but unmanned aerial image taken at low altitude (200 m) could distinguish trees individually with 30 cm per pixel. It, therefore, found that it is efficient and economically effective to produce unmanned hyperspectral images within the large-area mixed heritage. In addition, the unmanned aerial images have lower atmospheric errors and the ground sample distance that is high enough to distinguish individual trees, so they were found to be applied to the monitoring and the diagnosis for understanding the vegetation management and the health of the large-area mixed heritage.

Disease incidence and severity of Cercospora leaf spot in sugar beet assessed by multispectral unmanned aerial images and machine learning.

Disease incidence and metrics of disease severity are relevant parameters for decision making in plant protection and plant breeding. To develop automated and sensor-based routines, a sugar beet variety trial was inoculated with Cercospora beticola and monitored with a multispectral camera system mounted to an unmanned aerial vehicle (UAV) over the vegetation period. A pipeline based on machine learning methods was established for image data analysis and extraction of disease-relevant parameters. Features based on the digital surface model (DSM), vegetation indices, shadow condition and image resolution improved classification performance in comparison to using single multispectral channels in 12% and 6% of diseased and soil regions, respectively. With a post-processing step, area-related parameters were computed after classification. Results of this pipeline also included extraction of disease incidence (DI) and disease severity (DS) from UAV data. The calculated area under disease progress curve (AUDPC) of DS was 2810.4 to 7058.8 %.days for human visual scoring and 1400.5 to 4343.2 %.days for UAV-based scoring. Moreover, a sharper differentiation of varieties compared to visual scoring was observed in area-related parameters, like area of complete foliage (AF), area of healthy foliage (AH) and mean area of CLS lesion by unit of foliage (Ac/F). These advantages provide the option to replace the laborious work of visual disease assessments in the field with a more precise non-destructive assessment via multispectral data acquired by UAV flights.

Permafrost Degradation as a Supporting Factor for the Biodiversity of Tundra Ecosystems

Highly detailed images acquired using unmanned aerial imaging made it possible to describe the structural features of a certain type of thermokarst depression. These depressions developed in edoma deposits in the southern part of the typical tundra subzone in the Lena Delta. The geomorphology and vegetation of the model depression are characterized by a combination of GIS analysis and ground survey. Detailed mapping has shown that the surface of the depression bottom consists of a “baidjarakh” field covered by the complex vegetation of highly productive plant communities dominated by grasses and graminoids. A comparison of satellite images and topographic maps from different years has shown that the depression appeared from a shallow thermokarst lake drained no later than 40 years ago. The depression terrain and vegetation are determined by its origin and the composition of the surface geological substrate. Depression vegetation is quite a contrast to zonal tundra in regards to its structure and floristic composition. It contains few species which tend to be more typical for southern distribution. Such spots serve as feeding places for herbivorous animals and, in the past, may have been an important part of the Pleistocene megafauna forage. Thermokarst events were typical for the ice-complex through the whole period of its formation, but might differ in intensity and area depending on climate conditions. Nowadays, the size and lifetime of thermokarst lakes depends on their position in the terrain and the thermoerosion intensity. In the tundra landscape, thermokarst plays an important role in the creation of new habitats and biodiversity support for plants and animals.

Availability of High-resolution Unmanned Aerial Image for Estimation of Shorelines

Park, J.K. and Lee, K.W., 2019. Availability of high-resolution unmanned aerial image for estimation of shorelines. In: Lee, J.L.; Yoon, J.-S.; Cho, W.C.; Muin, M., and Lee, J. (eds.), The 3rd International Water Safety Symposium. Journal of Coastal Research, Special Issue No. 91, pp. 411-415. Coconut Creek (Florida), ISSN 0749-0208.The shoreline is important as a baseline for national boundaries. Shoreline surveys are necessary to determine the length, area, and shape of a country and should be periodically investigated. Existing shoreline surveys were conducted based on surveying equipment such as the GNSS (Global Navigation Satellite System) and total stations and depth measurements. However, this method requires a lot of manpower and time. In this study, coastline survey was performed using high resolution image of UAV (Unmanned Aerial Vehicle). For the study, the west sea island was selected as the study area, and about 300 images were acquired and DSM (Digital Surface Model) and ortho image were obtained through data processing. The accuracy of the result was evaluated and the accuracy was confirmed to be less than 10 cm. In addition, DSM-based contour lines were created to provide basic data for determining shoreline. Shoreline surveying using UAV will greatly increase the efficiency of related work, and it will be possible to determine the shoreline by linking with tidal observations data.

High-density stereo image matching using intrinsic curves

One of the most active areas of research in photogrammetry and computer vision is dense three-dimensional (3D) reconstruction of the environment via high-density image matching. This research interest is mainly driven by the growing popularity of unconventional imaging solutions such as images captured from unmanned aerial vehicles. With such data, problems like large disparity search space, occlusion, noise, and matching ambiguity become more pronounced. In this paper, we present a dense matching method to deal with these issues partially. The proposed method uses the concepts of intrinsic curves (IC) and derives useful matching information from their geometric features. First, we propose sparse disparity hypotheses for each pixel based on the orientations of the curves. These hypotheses are propagated to the neighboring pixels based on the proximity in the IC space; a solution which adaptively considers the intensity-variations in the neighborhood of a pixel to enlarge the set of its possible disparities. Then, a global matching energy function is formed and minimized, in which occlusions are explicitly taken into account based on curvature similarities of the ICs. The proposed technique is extensively tested on close-range and unmanned aerial images. Its performance is also compared to the state-of-the-art of dense-matching, such as hierarchical semi-global matching. Evaluations by the Middlebury Computer Vision Stereo Benchmark also show that the proposed technique results in average 3% error (percentage of pixels which are matched with more than 1-pixel error). The proposed framework could achieve high levels of accuracy (averagely 92%) as well as high efficiency by reducing the disparity search space up to 98% with an average confidence of 92% that the correct match, if existing, is still in the reduced search space.

SATELLITE IMAGE CLASSIFICATION OF BUILDING DAMAGES USING AIRBORNE AND SATELLITE IMAGE SAMPLES IN A DEEP LEARNING APPROACH

Abstract. The localization and detailed assessment of damaged buildings after a disastrous event is of utmost importance to guide response operations, recovery tasks or for insurance purposes. Several remote sensing platforms and sensors are currently used for the manual detection of building damages. However, there is an overall interest in the use of automated methods to perform this task, regardless of the used platform. Owing to its synoptic coverage and predictable availability, satellite imagery is currently used as input for the identification of building damages by the International Charter, as well as the Copernicus Emergency Management Service for the production of damage grading and reference maps. Recently proposed methods to perform image classification of building damages rely on convolutional neural networks (CNN). These are usually trained with only satellite image samples in a binary classification problem, however the number of samples derived from these images is often limited, affecting the quality of the classification results. The use of up/down-sampling image samples during the training of a CNN, has demonstrated to improve several image recognition tasks in remote sensing. However, it is currently unclear if this multi resolution information can also be captured from images with different spatial resolutions like satellite and airborne imagery (from both manned and unmanned platforms). In this paper, a CNN framework using residual connections and dilated convolutions is used considering both manned and unmanned aerial image samples to perform the satellite image classification of building damages. Three network configurations, trained with multi-resolution image samples are compared against two benchmark networks where only satellite image samples are used. Combining feature maps generated from airborne and satellite image samples, and refining these using only the satellite image samples, improved nearly 4 % the overall satellite image classification of building damages.