Terrestrial Images Research Articles

Multitemporal Quantification of the Geomorphodynamics on a Slope within the Cratère Dolomieu at the Piton de la Fournaise (La Réunion, Indian Ocean) Using Terrestrial LiDAR Data, Terrestrial Photographs, and Webcam Data

In this study, the geomorphological evolution of an inner flank of the Cratère Dolomieu at Piton de La Fournaise/La Réunion was investigated with the help of terrestrial laser scanning (TLS) data, terrestrial photogrammetric images, and historical webcam photographs. While TLS data and the terrestrial images were recorded during three field surveys, the study was also able to use historical webcam images that were installed for the monitoring of the volcanic activity inside the crater. Although the webcams were originally intended to be used only for visual monitoring of the area, at certain times they captured image pairs that could be analyzed using structure from motion (SfM) and subsequently processed to create digital terrain models (DTMs). With the help of all the data, the geomorphological evolution of selected areas of the crater was investigated in high temporal and spatial resolution. Surface changes were detected and quantified on scree slopes in the upper area of the crater as well as on scree slopes at the transition from the slope to the crater floor. In addition to their quantification, these changes could be assigned to individual geomorphological processes over time. The webcam photographs were a very important additional source of information here, as they allowed the observation period to be extended further into the past. Besides this, the webcam images made it possible to determine the exact dates at which geomorphological processes were active.

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.

SANS-CNN: An automated machine learning technique for spaceflight associated neuro-ocular syndrome with astronaut imaging data

Spaceflight associated neuro-ocular syndrome (SANS) is one of the largest physiologic barriers to spaceflight and requires evaluation and mitigation for future planetary missions. As the spaceflight environment is a clinically limited environment, the purpose of this research is to provide automated, early detection and prognosis of SANS with a machine learning model trained and validated on astronaut SANS optical coherence tomography (OCT) images. In this study, we present a lightweight convolutional neural network (CNN) incorporating an EfficientNet encoder for detecting SANS from OCT images titled “SANS-CNN.” We used 6303 OCT B-scan images for training/validation (80%/20% split) and 945 for testing with a combination of terrestrial images and astronaut SANS images for both testing and validation. SANS-CNN was validated with SANS images labeled by NASA to evaluate accuracy, specificity, and sensitivity. To evaluate real-world outcomes, two state-of-the-art pre-trained architectures were also employed on this dataset. We use GRAD-CAM to visualize activation maps of intermediate layers to test the interpretability of SANS-CNN’s prediction. SANS-CNN achieved 84.2% accuracy on the test set with an 85.6% specificity, 82.8% sensitivity, and 84.1% F1-score. Moreover, SANS-CNN outperforms two other state-of-the-art pre-trained architectures, ResNet50-v2 and MobileNet-v2, in accuracy by 21.4% and 13.1%, respectively. We also apply two class-activation map techniques to visualize critical SANS features perceived by the model. SANS-CNN represents a CNN model trained and validated with real astronaut OCT images, enabling fast and efficient prediction of SANS-like conditions for spaceflight missions beyond Earth’s orbit in which clinical and computational resources are extremely limited.

Toward viewing behavior for aerial scene categorization

Previous work has demonstrated similarities and differences between aerial and terrestrial image viewing. Aerial scene categorization, a pivotal visual processing task for gathering geoinformation, heavily depends on rotation-invariant information. Aerial image-centered research has revealed effects of low-level features on performance of various aerial image interpretation tasks. However, there are fewer studies of viewing behavior for aerial scene categorization and of higher-level factors that might influence that categorization. In this paper, experienced subjects’ eye movements were recorded while they were asked to categorize aerial scenes. A typical viewing center bias was observed. Eye movement patterns varied among categories. We explored the relationship of nine image statistics to observers’ eye movements. Results showed that if the images were less homogeneous, and/or if they contained fewer or no salient diagnostic objects, viewing behavior became more exploratory. Higher- and object-level image statistics were predictive at both the image and scene category levels. Scanpaths were generally organized and small differences in scanpath randomness could be roughly captured by critical object saliency. Participants tended to fixate on critical objects. Image statistics included in this study showed rotational invariance. The results supported our hypothesis that the availability of diagnostic objects strongly influences eye movements in this task. In addition, this study provides supporting evidence for Loschky et al.’s (Journal of Vision, 15(6), 11, 2015) speculation that aerial scenes are categorized on the basis of image parts and individual objects. The findings were discussed in relation to theories of scene perception and their implications for automation development.

Assessment of the impact of utility-scale photovoltaics on the surrounding environment in the Iberian Peninsula. Alternatives for the coexistence with agriculture

The rapid growth of photovoltaic solar energy, to achieve decarbonization, has been accompanied by increasing land occupation and the subsequent concern in the agroforestry sector. The increase in land area occupied has been of 20% in recent years, boosting solar electricity production to 5.9% of the total in Europe. This fact raises the question of the impact on vegetation greenness and moisture in the rural environment, something that has not always been considered.Image analysis is presented as one of the most effective tools to estimate the variation of vegetation greenness and moisture. For this, terrestrial images, like those from unmanned aerial vehicles, can be used; however, this limits the amount of information available and/or increases the cost. The use of satellite images in different bands is a relatively new tool that can be exploited for the analysis of solar plants impact. This work presents a new way to use Sentinel imagery to analyse the impact of utility-scale solar plants on vegetation and moisture of the surrounding areas. According to our results, a moderate decrease in weighted index for both moisture (5%) and vegetation (3%) occurred after solar plant installation. It is expected that these results can be of help for the design of new PV and agrivoltaic plants, originating the Ground-Integrated Photovoltaics (GIPV).

Transfer learning for galaxy feature detection: Finding giant star-forming clumps in low-redshift galaxies using Faster Region-based Convolutional Neural Network

Abstract Giant star-forming clumps (GSFCs) are areas of intensive star-formation that are commonly observed in high-redshift (z ≳ 1) galaxies but their formation and role in galaxy evolution remain unclear. Observations of low-redshift clumpy galaxy analogues are rare but the availability of wide-field galaxy survey data makes the detection of large clumpy galaxy samples much more feasible. Deep Learning (DL), and in particular Convolutional Neural Networks (CNNs), have been successfully applied to image classification tasks in astrophysical data analysis. However, one application of DL that remains relatively unexplored is that of automatically identifying and localizing specific objects or features in astrophysical imaging data. In this paper, we demonstrate the use of DL-based object detection models to localize GSFCs in astrophysical imaging data. We apply the Faster Region-based Convolutional Neural Network object detection framework (FRCNN) to identify GSFCs in low-redshift (z ≲ 0.3) galaxies. Unlike other studies, we train different FRCNN models on observational data that was collected by the Sloan Digital Sky Survey and labelled by volunteers from the citizen science project ‘Galaxy Zoo: Clump Scout’. The FRCNN model relies on a CNN component as a ‘backbone’ feature extractor. We show that CNNs, that have been pre-trained for image classification using astrophysical images, outperform those that have been pre-trained on terrestrial images. In particular, we compare a domain-specific CNN – ‘Zoobot’ – with a generic classification backbone and find that Zoobot achieves higher detection performance. Our final model is capable of producing GSFC detections with a completeness and purity of ≥0.8 while only being trained on ∼5000 galaxy images.

Automatic detection of urban infrastructure elements from terrestrial images using deep learning

Automatic detection of urban infrastructure elements from terrestrial images using deep learning

3D Modelling of Sanggrahan Temple Using UAV Imagery and Terrestrial Photogrammetry Method

As evidence of ancient Majapahit civilization existence, Sanggrahan Temple has been officially recognized as a cultural heritage building by Indonesian government in 2019. Sanggrahan temple is located in Boyolangu District, Tulungagung Regency, East Java. Conservation efforts in Sanggrahan temple have been done by BPK Wilayah XI Jawa Timur. Many parts of the temple have been damaged and are believed in different condition from its past original design. Therefore, a 3D documentation of Sanggrahan Temple needs to be carried out to provide sufficient data for future conservation efforts. The photogrammetry method which utilizes UAV (Unmanned Aerial Vehicles) and terrestrial imagery offers great efficiency in terms of fieldwork and cost. The technique applied close range photogrammetry for image-acquisition and Structure from Motion (SfM) processing. There are roughly 500 images captured for 3D modelling purposes. For georeferencing and geometric accuracy assessment, the Ground Control Point measurement was also carried out using GPS receiver and total station. The result show that the combination of aerial (UAV) and terrestrial images provide complete 3D model. The generated 3D model can provide architectural views of the temple from all sides which help architects and archaeologists to interpret architectural design and detail of the building. Accuracy assessment based on the coordinate measurement shows that the 3D model results meet the specified accuracy standard of 10 cm which is suitable for many architectural drawing applications.

EVALUATING MONOCULAR DEPTH ESTIMATION METHODS

Abstract. Depth estimation from monocular images has become a prominent focus in photogrammetry and computer vision research. Monocular Depth Estimation (MDE), which involves determining depth from a single RGB image, offers numerous advantages, including applications in simultaneous localization and mapping (SLAM), scene comprehension, 3D modeling, robotics, and autonomous driving. Depth information retrieval becomes especially crucial in situations where other sources like stereo images, optical flow, or point clouds are not available. In contrast to traditional stereo or multi-view methods, MDE techniques require fewer computational resources and smaller datasets. This research work presents a comprehensive analysis and evaluation of some state-of-the-art MDE methods, considering their ability to infer depth information in terrestrial images. The evaluation includes quantitative assessments using ground truth data, including 3D analyses and inference time.

Underwater image enhancement utilizing adaptive color correction and model conversion for dehazing

Underwater imaging is typically affected by absorption and scattering, which can lead to color distortion and image blurring. However, no universal color constancy algorithm for correcting underwater images exists due to the wide variation in underwater scenes. Furthermore, methods based on learning to improve underwater image quality often suffer from a lack of ground truth, which can compromise the validity and authenticity of restored images. In this study, we propose an adaptive underwater image enhancement technique that corrects image color and reduces haze using a valid dataset with ground truth. Specifically, we first use hue channel statistics from underwater images to build a dataset of color-corrected images from different underwater scenes. We use this dataset to train an Unet-like network for adaptive color correction. After color correction, underwater images often exhibit characteristics similar to hazy terrestrial images. We refer to this phenomenon as “model conversion”. The hazy terrestrial images have corresponding ground truth. Therefore, we train a Transformer-like network with hazy terrestrial image datasets to remove haze in underwater images. Our method is more robust for different underwater scenes and elegantly solves the problem of lacking ground truth. A series of experiments demonstrate that our method achieves superior performance compared to state-of-the-art methods in terms of both visual quality and quantitative metrics.

A close-range photogrammetric model for tracking and performance-based forecasting earthmoving operations

PurposeThis paper aims to present a highly accessible and affordable tracking model for earthmoving operations in an attempt to overcome some of the limitations of current tracking models.Design/methodology/approachThe proposed methodology involves four main processes: acquiring onsite terrestrial images, processing the images into 3D scaled cloud data, extracting volumetric measurements and crew productivity estimations from multiple point clouds using Delaunay triangulation and conducting earned value/schedule analysis and forecasting the remaining scope of work based on the estimated performance. For validation, the tracking model was compared with an observation-based tracking approach for a backfilling site. It was also used for tracking a coarse base aggregate inventory for a road construction project.FindingsThe presented model has proved to be a practical and accurate tracking approach that algorithmically estimates and forecasts all performance parameters from the captured data.Originality/valueThe proposed model is unique in extracting accurate volumetric measurements directly from multiple point clouds in a developed code using Delaunay triangulation instead of extracting them from textured models in modelling software which is neither automated nor time-effective. Furthermore, the presented model uses a self-calibration approach aiming to eliminate the pre-calibration procedure required before image capturing for each camera intended to be used. Thus, any worker onsite can directly capture the required images with an easily accessible camera (e.g. handheld camera or a smartphone) and can be sent to any processing device via e-mail, cloud-based storage or any communication application (e.g. WhatsApp).

Extreme Underwater Image Compression Using Physical Priors

Underwater images (UWIs) require higher compression ratio than terrestrial images due to the limited bandwidth of underwater wireless acoustic channel. In many studies such as marine species, the foreground objects (FGOs) in UWIs need to be observed in detail, while the background only needs to be viewed in general. However, existing image compression methods achieve limited compression ratio and reconstruction quality, which cannot fulfill these practical applications since they do not consider the unique underwater physical priors. To overcome the limitation, we propose an underwater physical prior-based extreme compression network (PPECN) for UWIs compression, which includes an underwater physical prior-guided FGOs autoencoder (UPGAE) and a FGOs-assisted background generator (FG-BGGAN). Specifically, we design an underwater physical prior guidance structure that simulates the data flow in the underwater physical imaging process to adaptively adjust the distribution of received Gaussian features in the UPGAE to be more consistent with real UWIs. During the adjustment, some basic UWI properties can be reconstructed, which can improve the reconstruction quality and implicitly reduce bits through the end-to-end training. Furthermore, the background is generated from simple semantic map under the constraint of the perceptual consistency between background and FGOs, significantly saving coding bits and improving the perceptual quality of the generated background. Extensive experimental results on four underwater image datasets verify that, compared with state-of-the-art compression methods, our PPECN achieves both impressive improvement in the perceptual quality of the whole image and significant gain in the pixel fidelity of the FGOs at the similar low bitrate.

Pseudo-Retinex decomposition-based unsupervised underwater image enhancement and beyond

Underwater images suffer from color casts and low contrast degraded due to wavelength-dependent light scatter and abortion of the underwater environment. To effectively improve the quality of the underwater images, deep learning-based underwater image enhancement methods have been widely proposed. However, most deep learning-based underwater image enhancement methods rely heavily on paired datasets. Actually, obtaining distortion-free images as reference images is difficult in underwater imaging. To address this problem, a fully Unsupervised convolution neural network-based Underwater Image Enhancement (UUIE) is proposed by pseudo-Retinex decomposition. The innovation of the proposed UUIE is to establish a relationship between the underwater imaging model and the Retinex model, then use terrestrial images to replace underwater images for training and estimate pseudo-illumination and pseudo-reflection maps through self-supervision using the pseudo-Retinex decomposition. The pseudo-reflection image and pseudo-illumination image are reconstructed by the pseudo-Retinex decomposition to obtain the enhanced image. Additionally, the proposed UUIE can also be extended to image dehazing and low-light enhancement with only one trained model. Experimental results on synthetic and real-world datasets demonstrate the effectiveness of the proposed UUIE quantitatively and qualitatively.

Tie Point Matching between Terrestrial and Aerial Images Based on Patch Variational Refinement

To produce highly detailed 3D models of architectural scenes, both aerial and terrestrial images are usually captured. However, due to the different viewpoints of each set of images, visual entities in cross-view images show dramatic changes. The perspective distortion makes it difficult to obtain correspondences between aerial–terrestrial image pairs. To solve this problem, a tie point matching method based on variational patch refinement is proposed. First, aero triangulation is performed on aerial images and terrestrial images, respectively; then, patches are created based on sparse point clouds. Second, the patches are optimized to be close to the surface of the object by variational patch refinement. The perspective distortion and scale difference of the terrestrial and aerial images projected onto the patches are reduced. Finally, tie points between aerial and terrestrial images can be obtained through patch-based matching. Experimental evaluations using four datasets from the ISPRS benchmark datasets and Shandong University of Science and Technology datasets reveal the satisfactory performance of the proposed method in terrestrial–aerial image matching. However, matching time is increased, because point clouds need to be generated. Occlusion in an image, such as that caused by a tree, can influence the generation of point clouds. Therefore, future research directions include the optimization of time complexity and the processing of occluded images.

Conversion from multi-spectral data into SAR data with Deep Convolution Neural Architecture using Generative Adversarial Network

Synthetic Aperture Radar (SAR) is a high-resolution radar extensively used for aerial and terrestrial images in numerous applications. The SAR images are all weather and day/night capable. It is extensively used in defense/military systems for surveillance, target detection and target tracking. In current times, the need for deep learning algorithms has increased due to its capability of generating SAR images, as it is very intensive and expensive task for the acquisition and formation of SAR images through sensors. Thus, alternative methods are required to produce SAR image data. In this paper, generative adversarial networks (GAN) are used to generate SAR images from a limited pre-existing dataset. This paper presents a novel way of translating optical images to SAR images, for the generation of paired, unpaired data set. Due to the unavailability of an open-source paired, unpaired image dataset, a synthetic paired image dataset is created for proving the concepts of CycleGAN and Pix-to-Pix algorithm by fine tuning the algorithms using generated datasets. The paper tried tackling this problem in two ways, both paired and unpaired image-to-image translation.

Automatic orientation of historical terrestrial images in mountainous terrain using the visible horizon

Historical terrestrial images are the only visual sources documenting alpine environments shortly after the end of the Little Ice Age. Despite their unique value, they are largely unused for quantifying environmental changes because of the difficult and time-consuming estimation of the unknown camera parameters. For most images large parts of the captured scenery have vastly changed over time, making automatic feature point matching infeasible. In contrast, the visible image horizon seems to remain stable over time and hence, appears to be a suitable feature for image orientation. Since the focal length is unknown for historical terrestrial images, existing methods, focusing solely on estimating the exterior orientation of recent imagery, can not be applied. Accordingly, it was investigated if the horizon is suitable to estimate both the interior and exterior orientation of historical terrestrial images, with an accuracy comparable to manually oriented images. In a first step, the whole horizon was used to approximate the unknown camera parameters, reducing the potential search space. In the subsequent spatial resection these approximations were further refined using salient points along the horizon. We evaluated our approach using 204 manually oriented reference images. With the proposed method the accuracy of the estimated exterior orientation could be significantly improved compared to previous works. Additionally, the unknown focal length was estimated within 5% of the true focal length for 75% of the images. As historical terrestrial images are commonly used for monoplotting, the accuracy for 2400 manually selected checkpoints was evaluated. This analysis showed that for 63% of the images the same accuracy as with manually oriented images was achieved. For additional 22% the estimated camera parameters were still accurate enough to serve as initial estimates for a subsequent manual orientation. In 15% of the images our method completely failed. Due to the vastly changing scenery and oblique viewing geometry, finding the initial camera parameters, in our experience, is often the most challenging and time consuming step during manual orientation of historical images. Hence, in 85% of the images this initial step can be replaced with our method, leading to a significantly reduced effort for orienting whole collections of historical terrestrial images.

Reinstating ‘the Value of Solitude’: Gaston Bachelard on the Imagination and Moral Life

ABSTRACT The aim of this article is to show that what Gaston Bachelard called the “psychology of the imagination” often doubles as moral psychology. In Water and Dreams, for example, Bachelard presents “water’s morality,” which is a morality attained by an imagination of water’s purity. Similarly, in Air and Dreams, he explores the aerial imagination that forms the moral thought in Thus Spoke Zarathustra and calls the will to dynamism in Nietzschean philosophy “an experimental physics of moral life.” In Earth and Reveries of Will, numerous terrestrial images are considered and the moralities they embody are discussed—among them the supreme hardness of the oak tree in Virginia Woolf’s Orlando and the “counsel of solidity” it gives. As Bachelard becomes more inclined to call his endeavor a “phenomenology of the imagination,” concern for morality becomes less pronounced. And yet in two of his later works, The Poetics of Space and The Poetics of Reverie, Bachelard continues to delve into the ways imagination allies with moral life. Sensitively heeding the solicitations from the material world and our responses to them, Bachelard’s psychology of the imagination is a project of self-knowledge of an unusual sort: it uncovers “the autobiography of lost possibilities” and revives confidence in human strength. Finally, a comparison between Bachelard’s moral psychology with the work of Theodor Adorno and of John McDowell shows that its philosophical significance is best viewed as a re-enchantment project—the re-enchantment of nature and our relationship with it.

Ecological Assessment Based on Remote Sensing Ecological Index: A Case Study of the “Three-Lake” Basin in Yuxi City, Yunnan Province, China

With continuous urbanization, human activities have left considerable impacts on the ecology. Therefore, it is necessary to perform timely and objective monitoring and evaluation of the ecology. With the basin of three highland lakes (Fuxian Lake, Xinyun Lake, and Qilu Lake) in Yunnan Province as the study case, four indices, i.e., the Normalized Difference Vegetation Index (NDVI), the Wet Index (WET), the Normalized Differential Build-Up And Bare Soil Index (NDBSI), and the Land Surface Temperature (LST), which indicate, respectively, greenness, humidity, dryness, and heat of the study area, were extracted. On the basis of five sets of terrestrial images of the areas around the three lakes from 2001 to 2021, principal component analysis (PCA) was performed on these four indices; the more informative principal component contribution was selected as the weight to establish a remote sensing ecological index (RSEI) evaluation model to evaluate the ecological environment quality of the study area; the Mann–Kendall test combined with Sen’s slope (Sen + MK) and the Hurst exponent were employed to explore the ecological conditions and development trends of the “three-lake” basin. The results showed that the ecological quality of the study area improved and then deteriorated from 2001 to 2021. The ecological quality classes in the study area were fair, medium, and good. The ecological quality has been greatly improved, but poor ecological quality was still observed in some regions such as Chengjiang. Eighty-eight percent of the study area witnessed a stable trend in the ecological quality over the 20 years; in 2021, the area of built-up land with fair and poor ecological quality reached 140.97 km2, which occupies 68.1% of the total area under the same land use type. Analysis shows that urban area expansion and human activities have exacerbated ecological problems of towns and built-up land in the study area. In the selected indicators, both greenness and humidity are positive indicators to ecological quality, and the R2 value of the two in 5-year regression was both greater than 0.99, which validated the reliability of the selected model indicators. The research findings are expected to provide a basis for scientific ecological planning and restoration of lake basins.

Halving of Swiss glacier volume since 1931 observed from terrestrial image photogrammetry

Abstract. The monitoring of glaciers in Switzerland has a long tradition, yet glacier changes during the 20th century are only known through sparse observations. Here, we estimate a halving of Swiss glacier volumes between 1931 and 2016 by mapping historical glacier elevation changes at high resolution. Our analysis relies on a terrestrial image archive known as TerrA, which covers about 86 % of the Swiss glacierised area with 21 703 images acquired during the period 1916–1947 (with a median date of 1931). We developed a semi-automated workflow to generate digital elevation models (DEMs) from these images, resulting in a 45 % total glacier coverage. Using the geodetic method, we estimate a Swiss-wide glacier mass balance of −0.52 ± 0.09 m w.e. a−1 between 1931 and 2016. This equates to a 51.5 ± 8.0 % loss in glacier volume. We find that low-elevation, high-debris-cover, and gently sloping glacier termini are conducive to particularly high mass losses. In addition to these glacier-specific, quasi-centennial elevation changes, we present a new inventory of glacier outlines with known timestamps and complete attributes from around 1931. The fragmented spatial coverage and temporal heterogeneity of the TerrA archive are the largest sources of uncertainty in our glacier-specific estimates, reaching up to 0.50 m w.e. a−1. We suggest that the high-resolution mapping of historical surface elevations could also unlock great potential for research fields other than glaciology.

Real-Time Ground-Level Building Damage Detection Based on Lightweight and Accurate YOLOv5 Using Terrestrial Images

Real-time building damage detection effectively improves the timeliness of post-earthquake assessments. In recent years, terrestrial images from smartphones or cameras have become a rich source of disaster information that may be useful in assessing building damage at a lower cost. In this study, we present an efficient method of building damage detection based on terrestrial images in combination with an improved YOLOv5. We compiled a Ground-level Detection in Building Damage Assessment (GDBDA) dataset consisting of terrestrial images with annotations of damage types, including debris, collapse, spalling, and cracks. A lightweight and accurate YOLOv5 (LA-YOLOv5) model was used to optimize the detection efficiency and accuracy. In particular, a lightweight Ghost bottleneck was added to the backbone and neck modules of the YOLOv5 model, with the aim to reduce the model size. A Convolutional Block Attention Module (CBAM) was added to the backbone module to enhance the damage recognition effect. In addition, regarding the scale difference of building damage, the Bi-Directional Feature Pyramid Network (Bi-FPN) for multi-scale feature fusion was used in the neck module to aggregate features with different damage types. Moreover, depthwise separable convolution (DSCONV) was used in the neck module to further compress the parameters. Based on our GDBDA dataset, the proposed method not only achieved detection accuracy above 90% for different damage targets, but also had the smallest weight size and fastest detection speed, which improved by about 64% and 24%, respectively. The model performed well on datasets from different regions. The overall results indicate that the proposed model realizes rapid and accurate damage detection, and meets the requirement of lightweight embedding in the future.