Real-time Performance Research Articles

Energy-aware dynamic rescheduling of flexible manufacturing system using edge-cloud collaborative decision-making method

ABSTRACT With the development of cloud computing technology, energy-aware scheduling based on cloud computing is an essential means to achieve energy saving and carbon reduction in manufacturing systems. Due to various production disturbances in flexible production, the scheduling based on cloud computing has data security problems and poor real-time performance. How to use real-time information to improve the security and responsiveness of cloud scheduling is a research gap. This paper establishes a cloud-edge collaborative dynamic flexible job-shop energy-aware rescheduling decision-making model. According to the characteristics of dynamic production in a flexible manufacturing system, create a learning model for cloud scheduling with Deep Q Network (DQN) and send the training model to the edge for scheduling decisions. Based on real-time production interference data, edge scheduling model real-time update scheduling scheme. To improve the robustness of the cloud scheduling model based on DQN, Dynamic scheduling data at the edge will be uploaded to update the cloud model. In addition, the effectiveness and practicability of the model are verified in an extrusion workshop. The experimental results show that this method can improve the comprehensive objective evaluation of minimum energy consumption and completion time by 3.6% −49.3% compared with the traditional scheduling rules.

Real-Time Intelligent Diagnosis of Co-frequency Vibration Faults in Rotating Machinery Based on Lightweight-Convolutional Neural Networks

The co-frequency vibration fault is one of the common faults in the operation of rotating equipment, and realizing the real-time diagnosis of the co-frequency vibration fault is of great significance for monitoring the health state and carrying out vibration suppression of the equipment. In engineering scenarios, co-frequency vibration faults are highlighted by rotational frequency and are difficult to identify, and existing intelligent methods require more hardware conditions and are exclusively time-consuming. Therefore, Lightweight-convolutional neural networks (LW-CNN) algorithm is proposed in this paper to achieve real-time fault diagnosis. The critical parameters are discussed and verified by simulated and experimental signals for the sliding window data augmentation method. Based on LW-CNN and data augmentation, the real-time intelligent diagnosis of co-frequency is realized. Moreover, a real-time detection method of fault diagnosis algorithm is proposed for data acquisition to fault diagnosis. It is verified by experiments that the LW-CNN and sliding window methods are used with high accuracy and real-time performance.

Synchronous Switching Strategy to Enhance the Real-Time Powering and Charging Performance of Triboelectric Nanogenerator.

Triboelectric nanogenerators (TENGs) are of great significance as sustainable power sources that harvest energy from the human body and environment. Nevertheless, due to TENG's impedance-dependent output voltage characteristics, in traditional strategy (TS), real-timely powering a sensor with TENG has a poor sensing on/off ratio (or response), and directly charging a capacitor with TENG shows a low charging efficiency. This degraded real-time powering and charging performance of TENG compared to a commercial constant voltage source has been a huge challenge of the TENG field for a long time. Here, we propose a synchronous switching strategy (SSS) for TENG to real-timely power sensors or charge capacitors without degrading its performance. Compared with TS, this new strategy enables sensors to have 5-7 times sensing on/off ratio enhancement when using TENG as a power source, reaching the powering ability of a commercial constant voltage source, it makes the powering performance of TENG stable under different driving frequency, improving the powering robustness of TENG. In addition, compared with TS, SSS can also enhance the charging efficiency of TENG in every charging cycle by up to 2.4 times when charging capacitors. This work contributes to real-timely powering or charging the distributed, mobile and wireless electronics using TENG. This article is protected by copyright. All rights reserved.

Evaluation of a real-time PCR assay performance to detect Mycobacterium tuberculosis, rifampicin, and isoniazid resistance in sputum specimens: a multicenter study in two major cities of Indonesia

BackgroundTuberculosis (TB) is one of the major global health issues due to its high mortality rate, especially in low- and middle-income countries. One of the key success points of the TB eradication program is early TB diagnosis, which requires rapid and accurate diagnostic testing. This study aimed to evaluate the performance of a newly developed RT-PCR kit (Indigen MTB/DR-TB RT-PCR) in a routine TB clinical setting.MethodA multi-fluorescence RT-PCR assay was designed and developed to detect regions within IS6110, rpoB, katG, and inhA of the Mycobacterium tuberculosis (MTB) genes. Sputum specimens were obtained from suspected TB patients who visited TB healthcare facilities in two major cities of Indonesia from September 2022 to May 2023. Specimens were assessed using Indigen MTB/DR-TB RT-PCR, acid-fast bacillus (AFB) smear microscopy, MTB culture, and drug susceptibility testing (DST) methods. Fisher’s exact test (χ2) was used to analyze the Indigen performance relative to culture methods.ResultThe performance of Indigen MTB/DR-TB RT-PCR to detect MTB was assessed using 610 sputum specimens obtained from suspected patients. The overall sensitivity and specificity were 94.12% (95% CI: 90.86–96.48%) and 98.32% (95% CI: 96.20–99.46%), respectively. When the analysis was performed on AFB smear-negative TB subjects (386 subjects), a lower sensitivity level was found at 78.57% (95% CI: 68.26–86.78%), while the specificity level remained similar at 98.34% (95% CI: 96.18–99.46%). The overall performance of Indigen MTB/DR-TB RT-PCR to detect MTB showed substantial agreement with the MTB culture method (kappa value 0.93). In comparison to DST, the sensitivity and specificity levels of Indigen to detect RIF resistance or INH resistance were 78.2% (95% CI: 61.8–90.2%) and 82.8% (95% CI: 64.2–94.2%), respectively, while the specificity level for both groups was at 100% (95% CI, 87.7–100%).ConclusionIndigen MTB/DR-TB RT-PCR demonstrated reliable performance for TB molecular diagnostic testing and can be implemented in routine TB diagnostic settings.

Multi-Perspective Adaptive Paperless Examination Cheating Detection System Based on Image Recognition

This paper proposes a multi-perspective adaptive examination cheating behavior detection method to meet the demand for automated monitoring throughout the entire process in paperless online exams. Unlike current dual-perspective cheating behavior detection methods, we expand the monitoring field of view by using three cameras with different perspectives: the overhead perspective, the horizontal perspective, and the face perspective. This effectively covers areas where cheating may occur. An adaptive cheating behavior detection system based on three perspectives is proposed, including a gaze direction recognition model based on Swin Transformer, a cheating tool detection model based on Lightweight-YOLOv5-Coordinate Attention, and a cheating behavior determination model based on Multilayer Perceptron. To reduce computational complexity and ensure efficient processing while expanding the monitoring field of view, the system uses the results of the gaze direction recognition model to adaptively select the cheating behavior detection model from different perspectives, reducing the three-perspective system to dual-perspective. In online simulation tests, our method achieves cheating behavior determination at 35 frames per second, with an average recognition rate of 95%. It has good real-time performance, accuracy, and a large monitoring range.

Building footprint extraction from aerial images using an edge-aware YOLO-v8 network

Abstract. Building footprint extraction is a critical indicator for assessing urban infrastructure, and extracting building footprints from remote sensing imagery can have significant practical applications. However, achieving rapid and accurate extraction of building footprints remains highly challenging, especially in scenarios with complex scenes, dense building distributions, and small targets. The instance segmentation models of the YOLO series offer strong real-time performance, reducing considerable time and effort in practical applications. Therefore, we propose building footprint extraction based on an enhance YOLO-v8 network. This study focuses on three enhancements to the YOLO-v8 network to improve extraction accuracy. Building upon the YOLO-v8 framework, we have incorporated the Feature Pyramid Network (FPN) module into feature maps at all scales to efficiently propagate high-level semantic information. Additionally, we introduce the Triple Feature Encoder (TFE) module, which integrates spatial detail information from feature maps at three different scales to enhance the network's ability to extract multi-scale information. Finally, we explore the integration of the Prewitt model, a conventional edge detection operator, to assist in extracting edge features in target regions of feature maps. This integration aims to reduce the jagged edges frequently seen in the outcomes of the original YOLO-v8. Furthermore, the Prewitt operator's noise suppression capability helps mitigate the influence of non-target areas in the feature maps. The proposed framework achieves an instance segmentation accuracy of mAP50 is 84.6% and mAP50:95 is 51.4% on public datasets, outperforming the original YOLO-v8 network.

Autonomous Landing Spot Detection for Unmanned Aerial Vehicles Based on Monocular Vision

Abstract. Enhancing the autonomous landing capability of unmanned aerial vehicles (UAVs) is of great significance for improving their operational efficiency and field survival capabilities. To this end, we propose a real-time autonomous landing spot detection method for UAVs. Firstly, the pose of the UAV at any given time and the initial three-dimensional point cloud of the scene are estimated using simultaneous localization and mapping (SLAM) techniques. Then, since the initial point cloud is sparse and cannot be used for terrain analysis, we generate a voxel-based elevation map of the scene, which can establish interconnectivity among the points. Finally, we propose a shift-box strategy to comprehensively analyze various terrain factors in the elevation map, determine the landing spot for UAVs, and update the landing spot in real time. The UAV flight experiments conducted in the real world have demonstrated the effectiveness and real-time performance of the proposed method.

Path tracking with personalized handling stability constraints matched to driving styles

In the path-tracking control process, drivers have different expectations of the lateral response of vehicles. To meet drivers’ expectations of handling stability, this paper proposes a model predictive control (MPC)-based path-tracking controller with personalized handling stability constraints matched to driving style. Firstly, the categories of driving styles are obtained by applying the K-Means algorithm to the collected driving data. Then the sideslip angle-longitudinal velocity map is constructed based on the nonlinear dynamics model of the vehicle to generate the front-wheel steering angle constraints that are matched with the maximum sideslip of the vehicle expected by drivers of different styles. After that, an MPC-based path-tracking controller has been built, and the front-wheel steering angle is incorporated into the MPC controller. Furthermore, compared to the strategy of directly using sideslip angle constraints for handling stability control, this approach avoids introducing additional constraint variables, leading to improved real-time performance. Simulation experiments are conducted to compare the dynamic vehicle responses between the handling stability control that takes into account the driving style and the control system that does not consider the driving style. In addition, the subjective evaluation is conducted to verify the proposed path-tracking control method that considers the stability constraints of driving style. The evaluation results show that the proposed method can obtain better subjective evaluation results than a path-tracking controller that does not consider the driving style. Finally, through the hardware-in-the-loop experiments, the real-time computational capability of the proposed controller is verified.

Real-time prediction method of carbon concentration in carburized steel based on a BP neural network

ABSTRACT Gears are key components of mechanical transmission systems. They need to be carburized and quenched to meet the requirements of internal toughness and external hardness, to obtain higher hardness and wear resistance. Heat treatment engineers can calculate carbon concentration distribution through the finite element analysis method, facing the challenge of high computational cost and high memory requirements. To solve this problem, a real-time prediction method of 1D and 2D carburizing concentration based on a Back Propagation (BP) neural network (BPNN) is proposed. First, carburizing experiments were conducted to verify the accuracy of the carburizing numerical model. Then, by establishing an accurate carburizing model, 37,800 and 177,147 1D and 2D carburizing training samples were generated using the finite element model (FEM) method, respectively. Finally, for 1D carburizing, the average relative error of the training model on the test set is 0.2911%. For 2D carburizing, the average relative error of the reconstructed BPNN model on the test set was 0.4381%, and a real-time performance evaluation was performed. The carbon concentration prediction time for each set of process parameters is only 0.12 s, nearly 175 times faster than FEM calculation, which meets the accuracy and real-time requirements for real-time prediction of carburizing carbon concentration.

Visual SLAM for Unmanned Aerial Vehicles: Localization and Perception

Localization and perception play an important role as the basis of autonomous Unmanned Aerial Vehicle (UAV) applications, providing the internal state of movements and the external understanding of environments. Simultaneous Localization And Mapping (SLAM), one of the critical techniques for localization and perception, is facing technical upgrading, due to the development of embedded hardware, multi-sensor technology, and artificial intelligence. This survey aims at the development of visual SLAM and the basis of UAV applications. The solutions to critical problems for visual SLAM are shown by reviewing state-of-the-art and newly presented algorithms, providing the research progression and direction in three essential aspects: real-time performance, texture-less environments, and dynamic environments. Visual–inertial fusion and learning-based enhancement are discussed for UAV localization and perception to illustrate their role in UAV applications. Subsequently, the trend of UAV localization and perception is shown. The algorithm components, camera configuration, and data processing methods are also introduced to give comprehensive preliminaries. In this paper, we provide coverage of visual SLAM and its related technologies over the past decade, with a specific focus on their applications in autonomous UAV applications. We summarize the current research, reveal potential problems, and outline future trends from academic and engineering perspectives.

An end-cloud collaboration approach for online state-of-health estimation of lithium-ion batteries based on multi-feature and transformer

State-of-Health (SOH) estimation of lithium-ion batteries is critical for reliable and efficient energy storage and usage. Traditional SOH estimation approaches either compromise on real-time performance or on estimation accuracy. This paper introduces an end-cloud collaboration approach that synergistically combines a cloud-based deep learning model with an end-based empirical model to achieve both high-accuracy and real-time SOH estimation. Utilizing the Transformer architecture, the cloud-side model employs multiple features extracted through signal analysis technology. On the other hand, the end-side employs a fast, iterative double exponential empirical model that provides real-time estimates. Two methods of end-cloud collaborative updating are presented, both leveraging Kalman Filter and Unscented Kalman Filter (UKF) to integrate and iteratively update the models. Evaluation shows that the cloud-side model achieves a Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) of around 0.8 %, while the end-cloud collaboration system maintains these metrics at 1 %, successfully capturing nonlinear capacity rebound phenomena. The work contributes to the domain by offering an approach that balances computational load and latency, making it suitable for online applications. Under the Cyber Hierarchy and Interactional Network (CHAIN) framework, this end-cloud collaboration architecture holds promise for high-accuracy, real-time health management of lithium-ion batteries across their full lifecycle.

A 5K Efficient Low-Light Enhancement Model by Estimating Increment between Dark Image and Transmission Map Based on Local Maximum Color Value Prior

Low-light enhancement (LLE) has seen significant advancements over decades, leading to substantial improvements in image quality that even surpass ground truth. However, these advancements have come with a downside as the models grew in size and complexity, losing their lightweight and real-time capabilities crucial for applications like surveillance, autonomous driving, smartphones, and unmanned aerial vehicles (UAVs). To address this challenge, we propose an exceptionally lightweight model with just around 5K parameters, which is capable of delivering high-quality LLE results. Our method focuses on estimating the incremental changes from dark images to transmission maps based on the low maximum color value prior, and we introduce a novel three-channel transmission map to capture more details and information compared to the traditional one-channel transmission map. This innovative design allows for more effective matching of incremental estimation results, enabling distinct transmission adjustments to be applied to the R, G, and B channels of the image. This streamlined approach ensures that our model remains lightweight, making it suitable for deployment on low-performance devices without compromising real-time performance. Our experiments confirm the effectiveness of our model, achieving high-quality LLE comparable to the IAT (local) model. Impressively, our model achieves this level of performance while utilizing only 0.512 GFLOPs and 4.7K parameters, representing just 39.1% of the GFLOPs and 23.5% of the parameters used by the IAT (local) model.

Improved YOLOv8-Pose Algorithm for Albacore Tuna (Thunnus alalunga) Fork Length Extraction and Weight Estimation

Aiming at the problems of large statistical error and the poor real-time performance of catch weight in the ocean fishing tuna industry, an algorithm based on improved YOLOv8-Pose for albacore tuna (Thunnus alalunga) fork length extraction and weight estimation is proposed, with reference to the human body’s pose estimation algorithm. Firstly, a lightweight module constructed using a heavy parameterization technique is used to replace the backbone network, and secondly, a weighted bidirectional feature pyramid network BIFPN is utilized. Finally, the upper and lower jaw and tail feature points of the albacore tuna (Thunnus alalunga) were extracted using the key point detection algorithm, and the weight of the albacore tuna (Thunnus alalunga) was estimated based on the fitted relationship between fork length and weight. The experimental results show that the improved YOLOv8-Pose algorithm reduces the number of model parameters by 13.63% and the number of floating-point operations by 14.03% compared with the baseline model without decreasing the accuracy of the target detection and key point detection and improves the model inference speed by 374%. At the same time, it reduces the drift of the key point detection, and the error of the comparison with the actual albacore tuna (Thunnus alalunga) body weight is not more than 10%. The improved key point detection algorithm has high detection accuracy and inference speed, which provides accurate yield data for pelagic fishing and is expected to solve the existing statistical problems and improve the accuracy and real-time performance of data in the fishing industry.

Research on online anomaly detection methods for bearing degradation

In industrial applications, rolling bearings operate under conditions of high precision and high speed, and their physical and mechanical characteristics change with the increase in operating time. Traditional diagnostic methods struggle to adapt well to the changing characteristics of bearings for online anomaly detection. Therefore, this research proposes an online anomaly detection method for rolling bearings based on time-density-weighted incremental support vector data description (TISVDD). A classification strategy is proposed to prevent samples misclassification in the updating process. The Detection Boundary is established based on SVDD decision boundary to enhance the recognition of abnormal samples in the process of model updating. A dual-screening mechanism update strategy for support vectors is proposed. It involves establishing a preliminary screening mechanism based on the Elimination Boundary. On this basis, an in-depth screening mechanism based on time density weight is introduced by considering spatiotemporal characteristics of samples, enhancing the real-time performance of online anomaly detection for bearings. Building upon the fused dual-boundary SVDD, a TISVDD framework for online anomaly detection is proposed, enabling the detection model to dynamically update in response to data changes over time. To validate the effectiveness of the proposed method, experiments were conducted using the XJTU-SY bearing dataset and real-time datasets collected on an online hardware platform. The results demonstrate the effectiveness and superiority of the method in practical applications.

FCS-YOLO: an efficient algorithm for detecting steel surface defects

Due to the imperfection of the manufacturing process and external factors, there may be some defects on the steel surface, which seriously affects the life and availability of the steel. Aiming at the problems of low precision and poor real-time performance in the inspection of steel surface defects in the past, an algorithm named FCS-you only look once (YOLO) was proposed for the inspection of steel surface defects. Firstly, the backbone feature extraction network is reconstructed using FasterNet to reduce the number of model parameters and weight size. Secondly, a new C2f-global attention mechanism module is proposed, which can focus image features more accurately by dynamically adjusting feature response. Finally, the attention mechanism SK is embedded in the feature extraction network to adjust the receptive field of the model, strengthen its feature recognition ability, and effectively improve the detection accuracy of steel surface defects. The test results show that FCS-YOLO has the best detection effect and the fastest detection speed among the comparison models. On the NEU-DET Dataset, the detection accuracy reached 81.8% and the detection speed reached 129.87 frames per second, which met the requirements of accurate and fast detection of strip surface defects.

A lightweight network model designed for alligator gar detection

When using advanced detection algorithms to monitor alligator gar in real-time in wild waters, the efficiency of existing detection algorithms is subject to certain limitations due to turbid water quality, poor underwater lighting conditions, and obstruction by other objects. In order to solve this problem, we developed a lightweight real-time detection network model called ARD-Net, from the perspective of reducing the amount of calculation and obtaining more feature map patterns. We introduced a cross-domain grid matching strategy to accelerate network convergence, and combined the involution operator and dual-channel attention mechanism to build a more lightweight feature extractor and multi-scale detection reasoning network module to enhance the network’s response to different semantics. Compared with the yoloV5 baseline model, our method performs equivalently in terms of detection accuracy, but the model is smaller, the detection speed is increased by 1.48 times, When compared with the latest State-of-the-Art (SOTA) method, YOLOv8, our method demonstrates clear advantages in both detection efficiency and model size,and has good real-time performance. Additionally, we created a dataset of alligator gar images for training.

Bio-inspired mobile robot design and autonomous exploration strategy for underground special space

Purpose To make the robot that have real autonomous ability is always the goal of mobile robot research. For mobile robots, simultaneous localization and mapping (SLAM) research is no longer satisfied with enabling robots to build maps by remote control, more needs will focus on the autonomous exploration of unknown areas, which refer to the low light, complex spatial features and a series of unstructured environment, lick underground special space (dark and multiintersection). This study aims to propose a novel robot structure with mapping and autonomous exploration algorithms. The experiment proves the detection ability of the robot. Design/methodology/approach A small bio-inspired mobile robot suitable for underground special space (dark and multiintersection) is designed, and the control system is set up based on STM32 and Jetson Nano. The robot is equipped with double laser sensor and Ackerman chassis structure, which can adapt to the practical requirements of exploration in underground special space. Based on the graph optimization SLAM method, an optimization method for map construction is proposed. The Iterative Closest Point (ICP) algorithm is used to match two frames of laser to recalculate the relative pose of the robot, which improves the sensor utilization rate of the robot in underground space and also increase the synchronous positioning accuracy. Moreover, based on boundary cells and rapidly-exploring random tree (RRT) algorithm, a new Bio-RRT method for robot autonomous exploration is proposed in addition. Findings According to the experimental results, it can be seen that the upgraded SLAM method proposed in this paper achieves better results in map construction. At the same time, the algorithm presents good real-time performance as well as high accuracy and strong maintainability, particularly it can update the map continuously with the passing of time and ensure the positioning accuracy in the process of map updating. The Bio-RRT method fused with the firing excitation mechanism of boundary cells has a more purposeful random tree growth. The number of random tree expansion nodes is less, and the amount of information to be processed is reduced, which leads to the path planning time shorter and the efficiency higher. In addition, the target bias makes the random tree grow directly toward the target point with a certain probability, and the obtained path nodes are basically distributed on or on both sides of the line between the initial point and the target point, which makes the path length shorter and reduces the moving cost of the mobile robot. The final experimental results demonstrate that the proposed upgraded SLAM and Bio-RRT methods can better complete the underground special space exploration task. Originality/value Based on the background of robot autonomous exploration in underground special space, a new bio-inspired mobile robot structure with mapping and autonomous exploration algorithm is proposed in this paper. The robot structure is constructed, and the perceptual unit, control unit, driving unit and communication unit are described in detail. The robot can satisfy the practical requirements of exploring the underground dark and multiintersection space. Then, the upgraded graph optimization laser SLAM algorithm and interframe matching optimization method are proposed in this paper. The Bio-RRT independent exploration method is finally proposed, which takes shorter time in equally open space and the search strategy for multiintersection space is more efficient. The experimental results demonstrate that the proposed upgrade SLAM and Bio-RRT methods can better complete the underground space exploration task.

Data-driven traffic signal adaptive control algorithm integrating vehicle perception and traffic flow data

Due to the equipment error, environmental interference and data transmission delay of vehicle flow detection, the accuracy and real-time performance of vehicle perception and traffic flow data will be affected to some extent, resulting in poor traffic signal control effect. Therefore, a data-driven traffic signal adaptive control algorithm is designed by integrating vehicle perception and traffic flow data. To complete the modeling of urban traffic, the discrete distribution and continuous distribution of traffic are obtained. Based on this research environment, the DV-hop localization algorithm is improved to sense the vehicle position. Based on the phase space reconstruction of traffic flow time series and vehicle location information, traffic flow data is predicted. Based on the driving of traffic data, the vehicle types are divided into small, medium and large three categories, and the impact weights are assigned respectively, and the weight values affecting the final allocation of green time are obtained to realize the allocation of green time. The experimental results show that: The research algorithm can not only predict the traffic flow intensity effectively, but also the predicted results are highly coincident with the actual traffic flow intensity. Vehicle arrival rates are higher, vehicle delays are shorter, and vehicles stop fewer times on average.

A Novel Approach for Real-Time Tire-Ground Enveloping Modeling

Abstract Over the past decades, simulation has become vital in vehicle development and virtual testing, especially for autonomous vehicles. High-performance hard real-time simulators, which are crucial for those undertakings, require efficient algorithms to accurately model vehicle behavior within virtual environments. A prime example is tire-ground contact modeling, which is pivotal if we aim to achieve a high level of realism when simulating wheeled vehicles. Contact modeling focuses on an accurate estimation of the parameters needed to compute the forces and torques generated by vehicle-ground interaction. However, the complexity of this task is compounded by the fact that tire-ground contact is a highly non-linear phenomenon, which is further exacerbated by the need to perform tests to fine-tune state-of-the-art tire-ground contact models. To tackle those challenges, we have developed a novel enveloping model that does not require any fitting of experimental data and is based on the 3D geometry of the intersection between undeformed volumes. In this manuscript, we provide a detailed description of the algorithm's formulation, the current software implementation (available as an open-source library), as well as the achieved scalability and real-time performance.

BD-YOLO: detection algorithm for high-resolution remote sensing images

This paper focuses on detecting extremely small targets in aerial images. Compared to common datasets, the average size of targets in remote sensing images is only 12.8 pixels, significantly smaller than those in common datasets. Therefore, directly applying existing detectors to aerial images is ineffective. To address this issue and ensure real-time performance, This paper propose BD-YOLO, which incorporates five key innovations. A dual Backbone route was proposed to maintain data integrity and achieve high-resolution aerial remote sensing detection. Additionally, a new feature fusion method was developed to comprehensively merge shallow and deep information. To supplement small-sized target information, a new network structure was proposed. The detector strategy used by BD-YOLO considers the detection accuracy of objects with different sizes. Furthermore, a lightweight method was adopted to ensure real-time performance of the algorithm. BD-YOLO outperformed YOLOv8s on the AI-TOD dataset, achieving a higher mAP by 2.4%. Similarly, on the Visdrone dataset, BD-YOLO achieved a 2.5% higher mAP compared to YOLOv8s. Additionally, on the Tinyperson dataset, BD-YOLO achieved a 0.6% higher mAP than YOLOv8s. Notably, BD-YOLO maintains real-time performance while ensuring accurate object detection.