Search Results (12,677)

Degraded underwater images decrease the accuracy of underwater object detection. Existing research uses image enhancement methods to improve the visual quality of images, which may not be beneficial in underwater image detection and lead to serious degradation in detector performance. To alleviate this problem, we proposed a bidirectional guided method for underwater object detection, referred to as BG-YOLO. In the proposed method, a network is organized by constructing an image enhancement branch and an object detection branch in a parallel manner. The image enhancement branch consists of a cascade of an image enhancement subnet and object detection subnet. The object detection branch only consists of a detection subnet. A feature-guided module connects the shallow convolution layers of the two branches. When training the image enhancement branch, the object detection subnet in the enhancement branch guides the image enhancement subnet to be optimized towards the direction that is most conducive to the detection task. The shallow feature map of the trained image enhancement branch is output to the feature-guided module, constraining the optimization of the object detection branch through consistency loss and prompting the object detection branch to learn more detailed information about the objects. This enhances the detection performance. During the detection tasks, only the object detection branch is reserved so that no additional computational cost is introduced. Extensive experiments demonstrate that the proposed method significantly improves the detection performance of the YOLOv5s object detection network (the mAP is increased by up to 2.9%) and maintains the same inference speed as YOLOv5s (132 fps). Full article

Synthetic aperture radar is widely applied to ship detection due to generating high-resolution images under diverse weather conditions and its penetration capabilities, making SAR images a valuable data source. However, detecting multi-scale ship targets in complex backgrounds leads to issues of false positives and missed detections, posing challenges for lightweight and high-precision algorithms. There is an urgent need to improve accuracy of algorithms and their deployability. This paper introduces LH-YOLO, a YOLOv8n-based, lightweight, and high-precision SAR ship detection model. We propose a lightweight backbone network, StarNet-nano, and employ element-wise multiplication to construct a lightweight feature extraction module, LFE-C2f, for the neck of LH-YOLO. Additionally, a reused and shared convolutional detection (RSCD) head is designed using a weight sharing mechanism. These enhancements significantly reduce model size and computational demands while maintaining high precision. LH-YOLO features only 1.862 M parameters, representing a 38.1% reduction compared to YOLOv8n. It exhibits a 23.8% reduction in computational load while achieving a mAP50 of 96.6% on the HRSID dataset, which is 1.4% higher than YOLOv8n. Furthermore, it demonstrates strong generalization on the SAR-Ship-Dataset with a mAP50 of 93.8%, surpassing YOLOv8n by 0.7%. LH-YOLO is well-suited for environments with limited resources, such as embedded systems and edge computing platforms. Full article

Wildfires increasingly threaten ecosystems and infrastructure, making accurate burn severity mapping (BSM) essential for effective disaster response and environmental management. Machine learning (ML) models utilizing satellite-derived vegetation indices are crucial for assessing wildfire damage; however, incorporating many indices can lead to multicollinearity, reducing classification accuracy. While principal component analysis (PCA) is commonly used to address this issue, its effectiveness relative to other feature extraction (FE) methods in BSM remains underexplored. This study aims to enhance ML classifier accuracy in BSM by evaluating various FE techniques that mitigate multicollinearity among vegetation indices. Using composite burn index (CBI) data from the 2014 Carlton Complex fire in the United States as a case study, we extracted 118 vegetation indices from seven Landsat-8 spectral bands. We applied and compared 13 different FE techniques—including linear and nonlinear methods such as PCA, t-distributed stochastic neighbor embedding (t-SNE), linear discriminant analysis (LDA), Isomap, uniform manifold approximation and projection (UMAP), factor analysis (FA), independent component analysis (ICA), multidimensional scaling (MDS), truncated singular value decomposition (TSVD), non-negative matrix factorization (NMF), locally linear embedding (LLE), spectral embedding (SE), and neighborhood components analysis (NCA). The performance of these techniques was benchmarked against six ML classifiers to determine their effectiveness in improving BSM accuracy. Our results show that alternative FE techniques can outperform PCA, improving classification accuracy and computational efficiency. Techniques like LDA and NCA effectively capture nonlinear relationships critical for accurate BSM. The study contributes to the existing literature by providing a comprehensive comparison of FE methods, highlighting the potential benefits of underutilized techniques in BSM. Full article

Automatic modulation recognition (AMR) has wide applications in the fields of wireless communications, radar systems, and intelligent sensor networks. The existing deep learning-based modulation recognition models often focus on temporal features while overlooking the interrelations and spatio-temporal relationships among different types of signals. To overcome these limitations, a hybrid neural network based on a multimodal parallel structure, called the multimodal parallel hybrid neural network (MPHNN), is proposed to improve the recognition accuracy. The algorithm first preprocesses the data by parallelly processing the multimodal forms of the modulated signals before inputting them into the network. Subsequently, by combining Convolutional Neural Networks (CNN) and Bidirectional Gated Recurrent Unit (Bi-GRU) models, the CNN is used to extract spatial features of the received signals, while the Bi-GRU transmits previous state information of the time series to the current state to capture temporal features. Finally, the Convolutional Block Attention Module (CBAM) and Multi-Head Self-Attention (MHSA) are introduced as two attention mechanisms to handle the temporal and spatial correlations of the signals through an attention fusion mechanism, achieving the calibration of the signal feature maps. The effectiveness of this method is validated using various datasets, with the experimental results demonstrating that the proposed approach can fully utilize the information of multimodal signals. The experimental results show that the recognition accuracy of MPHNN on multiple datasets reaches 93.1%, and it has lower computational complexity and fewer parameters than other models. Full article

Digital elevation models (DEM) are widely used in many hydrologic applications, providing key information about the topography, which is a major driver of water flow in a landscape. Several open access DEMs with near-global coverage are currently available, however, they represent the elevation of the earth’s surface including all its elements, such as vegetation cover and buildings. These features introduce a positive elevation bias that can skew the water flow paths, impacting the extraction of hydrological features and the accuracy of hydrodynamic models. Many attempts have been made to reduce the effects of this bias over the years, leading to the generation of improved datasets based on the original global DEMs, such as MERIT DEM and, more recently, FABDEM. However, even after these corrections, the remaining bias still affects flow path delineation in a significant way. Aiming to improve on this aspect, a new vegetation bias correction method is proposed in this work. The method consists of subtracting from the Copernicus DEM elevations their respective forest height but adjusted by correction factors to compensate for the partial penetration of the SAR pulses into the vegetation cover during the Copernicus DEM acquisition process. These factors were calculated by a new approach where the slope around the pixels at the borders of each vegetation patch were analyzed. The forest height was obtained from a global dataset developed for the year 2019. Moreover, to avoid temporal vegetation cover mismatch between the DEM and the forest height dataset, we introduced a process where the latter is automatically adjusted to best match the Copernicus acquisition year. The correction method was applied for regions with different forest cover percentages and topographic characteristics, and the result was compared to the original Copernicus DEM and FABDEM, which was used as a benchmark for vegetation bias correction. The comparison method was hydrology-based, using drainage networks obtained from topographic maps as reference. The new corrected DEM showed significant improvements over both the Copernicus DEM and FABDEM in all tested scenarios. Moreover, a qualitative comparison of these DEMs was also performed through exhaustive visual analysis, corroborating these findings. These results suggest that the use of this new vegetation bias correction method has the potential to improve DEM-based hydrological applications worldwide. Full article

Land cover maps with high accuracy are essential for environmental protection and climate change research. The 30-meter-resolution maps, with their better resolution and longer historical records, are extensively utilized to assess changes in land cover and their effects on carbon storage, land–atmosphere energy balance, and water cycle processes. However, current data products use different classification methods, resulting in significant classification inconsistency and triggering serious disagreements among related studies. Here, we compared four mainstream land cover products in China, namely GLC_FCS30, CLCD, Globeland30, and CNLUCC. The result shows that only 50.34% of the classification results were consistent across the four datasets. The differences between pairs of datasets ranged from 21.10% to 37.53%. Importantly, most inconsistency occurs in transitional zones among land cover types sensitive to climate change and human activities. Based on the accuracy evaluation, CLCD is the most accurate land cover product, with an overall accuracy reaching 86.98 ± 0.76%, followed by CNLUCC (81.38 ± 0.87%) and GLC_FCS30 (77.83 ± 0.80%). Globeland30 had the lowest accuracy (75.24 ± 0.91%), primarily due to misclassification between croplands and forests. Misclassification diagnoses revealed that vegetation-related spectral confusion among land cover types contributed significantly to misclassifications, followed by slope, cloud cover, and landscape fragmentation, which affected satellite observation angles, data availability, and mixed pixels. Automated classification methods using the random forest algorithm can perform better than those that depend on traditional human–machine interactive interpretation or object-based approaches. However, their classification accuracy depends more on selecting training samples and feature variables. Full article

Recognizing ripe tomatoes is a crucial aspect of tomato picking. To ensure the accuracy of inspection results, You Only Look Once version 9 (YOLOv9) has been explored as a fruit detection algorithm. To tackle the challenge of identifying tomatoes and the low accuracy of small object detection in complex environments, we propose a ripe tomato recognition algorithm based on an enhanced YOLOv9-C model. After collecting tomato data, we used Mosaic for data augmentation, which improved model robustness and enriched experimental data. Improvements were made to the feature extraction and down-sampling modules, integrating HGBlock and SPD-ADown modules into the YOLOv9 model. These measures resulted in high detection performance with precision and recall rates of 97.2% and 92.3% in horizontal and vertical experimental comparisons, respectively. The module-integrated model improved accuracy and recall by 1.3% and 1.1%, respectively, and also reduced inference time by 1 ms compared to the original model. The inference time of this model was 14.7 ms, which is 16 ms better than the RetinaNet model. This model was tested accurately with [email protected] (%) up to 98%, which is 9.6% higher than RetinaNet. Its increased speed and accuracy make it more suitable for practical applications. Overall, this model provides a reliable technique for recognizing ripe tomatoes during the picking process. Full article

Detecting cracks in cylinder heads traditionally relies on manual inspection, which is time-consuming and susceptible to human error. As an alternative, automated object detection utilizing computer vision and machine learning models has been explored. However, these methods often face challenges due to a lack of sufficiently annotated training data, limited image diversity, and the inherently small size of cracks. Addressing these constraints, this paper introduces a novel automated crack-detection method that enhances data availability through a synthetic data generation technique. Unlike general data augmentation practices, our method involves copying cracks from one location to another, guided by both random and informed engineering decisions about likely crack formations due to cyclic thermomechanical loads. The innovative aspect of our approach lies in the integration of domain-specific engineering knowledge into the synthetic generation process, which substantially improves detection accuracy. We evaluate our method’s effectiveness using two metrics: the F2 score, which emphasizes recall to prioritize detecting all potential cracks, and mean average precision (MAP), a standard measure in object detection. Experimental results demonstrate that, without engineering insights, our method increases the F2 score from 0.40 to 0.65, while maintaining a stable MAP. Incorporating detailed engineering knowledge further enhances the F2 score to 0.70 and improves MAP to 0.57, representing increases of 63% and 43%, respectively. These results confirm that our approach not only mitigates the limitations of traditional data augmentation but also significantly advances the reliability and precision of crack detection in industrial settings. Full article

Tuna accounts for 20% of the output value of global marine capture fisheries, and it plays a crucial role in maintaining ecosystem stability, ensuring global food security, and supporting economic stability. However, improper management has led to significant overfishing, resulting in a sharp decline in tuna populations. For sustainable tuna fishing, it is essential to accurately identify the species of tuna caught and to count their numbers, as these data are the foundation for setting scientific catch quotas. The traditional manual identification method suffers from several limitations and is prone to errors during prolonged operations, especially due to factors like fatigue, high-intensity workloads, or adverse weather conditions, which ultimately compromise its accuracy. Furthermore, the lack of transparency in the manual process may lead to intentional underreporting, which undermines the integrity of fisheries’ data. In contrast, an intelligent, real-time identification system can reduce the need for human labor, assist in more accurate identification, and enhance transparency in fisheries’ management. This system not only provides reliable data for refined management but also enables fisheries’ authorities to dynamically adjust fishing strategies in real time, issue timely warnings when catch limits are approached or exceeded, and prevent overfishing, thus ultimately contributing to sustainable tuna management. In light of this need, this article proposes the RSNC-YOLO algorithm, an intelligent model designed for recognizing tuna in complex scenarios on fishing vessels. Based on YOLOv8s-seg, RSNC-YOLO integrates Reparameterized C3 (RepC3), Selective Channel Down-sampling (SCDown), a Normalization-based Attention Module (NAM), and C2f-DCNv3-DLKA modules. By utilizing a subset of images selected from the Fishnet Open Image Database, the model achieves a 2.7% improvement in [email protected] and a 0.7% improvement in [email protected]:0.95. Additionally, the number of parameters is reduced by approximately 30%, and the model’s weight size is reduced by 9.6 MB, while maintaining an inference speed comparable to that of YOLOv8s-seg. Full article

To improve the accuracy of detecting small and long-distance objects while self-driving cars are in motion, in this paper, we propose a 3D object detection method, Att-BEVFusion, which fuses camera and LiDAR data in a bird’s-eye view (BEV). First, the transformation from the camera view to the BEV space is achieved through an implicit supervision-based method, and then the LiDAR BEV feature point cloud is voxelized and converted into BEV features. Then, a channel attention mechanism is introduced to design a BEV feature fusion network to realize the fusion of camera BEV feature space and LiDAR BEV feature space. Finally, regarding the issue of insufficient global reasoning in the BEV fusion features generated by the channel attention mechanism, as well as the challenge of inadequate interaction between features. We further develop a BEV self-attention mechanism to apply global operations on the features. This paper evaluates the effectiveness of the Att-BEVFusion fusion algorithm on the nuScenes dataset, and the results demonstrate that the algorithm achieved 72.0% mean average precision (mAP) and 74.3% nuScenes detection score (NDS), with an advanced detection accuracy of 88.9% and 91.8% for single-item detection of automotive and pedestrian categories, respectively. Full article

This work presents a novel RGB-D dynamic simultaneous localization and mapping (SLAM) method that improves accuracy, stability, and efficiency of localization while relying on deep learning in a dynamic environment, in contrast to traditional static scene-based visual SLAM methods. Based on the classic framework of traditional visual SLAM, we propose a method that replaces the traditional feature extraction method with a convolutional neural network approach, aiming to enhance the accuracy of feature extraction and localization, as well as to improve the algorithm’s ability to capture and represent the characteristics of the entire scene. Subsequently, the semantic segmentation thread was utilized in a target detection network combined with geometric methods to identify potential dynamic areas in the image and generate masks for dynamic objects. Finally, the standard deviation of the depth information of potential dynamic points was calculated to identify true dynamic feature points, to guarantee that static feature points were used for position estimation. We performed experiments based on the public datasets to validate the feasibility of the proposed algorithm. The experimental results indicate that the improved SLAM algorithm, which boasts a reduction in absolute trajectory error (ATE) by approximately 97% compared to traditional static visual SLAM and about 20% compared to traditional dynamic visual SLAM, also exhibited a 68% decrease in computation time compared to well-known dynamic visual SLAM, thereby possessing absolute advantages in both positioning accuracy and operational efficiency. Full article

Mapping shallow-water bathymetry and morphology represents a technical challenge. In fact, acoustic surveys are limited by water depths reachable by boat, and airborne surveys have high costs. Photogrammetric approaches (either via drone or from the sea surface) have opened up the possibility to perform shallow-water surveys easily and at accessible costs. This work presents a simple, low-cost, and highly portable platform that allows gathering sequential photos and echosounder depth values of shallow-water sites (up to 5 m depth). The photos are then analysed in conjunction with photogrammetric techniques to obtain digital bathymetric models and orthomosaics of the seafloor. The workflow was tested on four repeated surveys of the same area in the Western Mediterranean and allowed obtaining digital bathymetric models with centimetric average accuracy and precision and root mean square errors within a few decimetres. The platform presented in this work can be employed to obtain first-order bathymetric products, enabling the contextual establishment of the depth accuracy of the final products. Full article

Composite robots often encounter difficulties due to changes in illumination, external disturbances, reflective surface effects, and cumulative errors. These challenges significantly hinder their capabilities in environmental perception and the accuracy and reliability of pose estimation. We propose a nonlinear optimization approach to overcome these issues to develop an integrated localization and navigation framework, IIVL-LM (IMU, Infrared, Vision, and LiDAR Fusion for Localization and Mapping). This framework achieves tightly coupled integration at the data level using inputs from an IMU (Inertial Measurement Unit), an infrared camera, an RGB (Red, Green and Blue) camera, and LiDAR. We propose a real-time luminance calculation model and verify its conversion accuracy. Additionally, we designed a fast approximation method for the nonlinear weighted fusion of features from infrared and RGB frames based on luminance values. Finally, we optimize the VIO (Visual-Inertial Odometry) module in the R3LIVE++ (Robust, Real-time, Radiance Reconstruction with LiDAR-Inertial-Visual state Estimation) framework based on the infrared camera’s capability to acquire depth information. In a controlled study, using a simulated indoor rescue scenario dataset, the IIVL-LM system demonstrated significant performance enhancements in challenging luminance conditions, particularly in low-light environments. Specifically, the average RMSE ATE (Root Mean Square Error of absolute trajectory Error) improved by 23% to 39%, with reductions from 0.006 to 0.013. At the same time, we conducted comparative experiments using the publicly available TUM-VI (Technical University of Munich Visual-Inertial Dataset) without the infrared image input. It was found that no leading results were achieved, which verifies the importance of infrared image fusion. By maintaining the active engagement of at least three sensors at all times, the IIVL-LM system significantly boosts its robustness in both unknown and expansive environments while ensuring high precision. This enhancement is particularly critical for applications in complex environments, such as indoor rescue operations. Full article

The tasseling stage of maize, as a critical period of maize cultivation, is essential for predicting maize yield and understanding the normal condition of maize growth. However, the branches overlap each other during the growth of maize seedlings and cannot be used as an identifying feature. However, during the tasseling stage, its apical ear blooms and has distinctive features that can be used as an identifying feature. However, the sizes of the maize tassels are small, the background is complex, and the existing network has obvious recognition errors. Therefore, in this paper, unmanned aerial vehicle (UAV) RGB images and an improved YOLOv8 target detection network are used to enhance the recognition accuracy of maize tassels. In the new network, a microscale target detection head is added to increase the ability to perceive small-sized maize tassels; In addition, Spatial Pyramid Pooling—Fast (SPPF) is replaced by the Spatial Pyramid Pooling with Efficient Layer Aggregation Network (SPPELAN) in the backbone network part to connect different levels of detailed features and semantic information. Moreover, a dual-attention module synthesized by GAM-CBAM is added to the neck part to reduce the loss of features of maize tassels, thus improving the network’s detection ability. We also labeled the new maize tassels dataset in VOC format as the training and validation of the network model. In the final model testing results, the new network model’s precision reached 93.6% and recall reached 92.5%, which was an improvement of 2.8–12.6 percentage points and 3.6–15.2 percentage points compared to the mAP₅₀ and F1-score values of other models. From the experimental results, it is shown that the improved YOLOv8 network, with high performance and robustness in small-sized maize tassel recognition, can accurately recognize maize tassels in UAV images, which provides technical support for automated counting, accurate cultivation, and large-scale intelligent cultivation of maize seedlings. Full article

Clubbing finger is a significant clinical indicator, and its early detection is essential for the diagnosis and treatment of associated diseases. However, traditional diagnostic methods rely heavily on the clinician’s subjective assessment, which can be prone to biases and may lack standardized tools. Unlike other diagnostic challenges, the characteristic changes of clubbing finger are subtle and localized, necessitating high-precision feature extraction. Existing models often fail to capture these delicate changes accurately, potentially missing crucial diagnostic features or generating false positives. Furthermore, these models are often not suited for accurate clinical diagnosis in resource-constrained settings. To address these challenges, we propose MSG-YOLO, a lightweight clubbing finger detection model based on YOLOv8n, designed to enhance both detection accuracy and efficiency. The model first employs a multi-scale dilated residual module, which expands the receptive field using dilated convolutions and residual connections, thereby improving the model’s ability to capture features across various scales. Additionally, we introduce a Selective Feature Fusion Pyramid Network (SFFPN) that dynamically selects and enhances critical features, optimizing the flow of information while minimizing redundancy. To further refine the architecture, we reconstruct the YOLOv8 detection head with group normalization and shared-parameter convolutions, significantly reducing the model’s parameter count and increasing computational efficiency. Experimental results indicate that the model maintains high detection accuracy with reduced parameter and computational requirements. Compared to YOLOv8n, MSG-YOLO achieves a 48.74% reduction in parameter count and a 24.17% reduction in computational load, while improving the mAP0.5 score by 2.86%, reaching 93.64%. This algorithm strikes a balance between accuracy and lightweight design, offering efficient and reliable clubbing finger detection even in resource-constrained environments. Full article