Pub Date : 2023-08-06DOI: 10.1109/ICMA57826.2023.10216041
Jiahao Xu, Xiufen Ye
The cross-modality image registration task is becoming more and more important in the field of image fusion, thermal and visible image fusion can reduce the impact of environmental factors while maintaining the image’s texture and detail. However, the common registration methods always have the problem of overfitting and poor generalization when facing large modality span task. To register the thermal and visible images, a simple yet effective registration model called Dense-Cascade network is presented in this paper. Our network uses two independent branches to extract the cross-modality features respectively. In order to further reduce the regression error, we use a cascade structure by stacking the networks and using STN layer to achieve gradient backpropagation. Finally, to validate the robustness of our model, we add gaussian noise severity from 0-4 on thermal images to test the performance of our models, and the results show that the registration performances of our models are better than others.
{"title":"Dense-Cascade Neural Network for Thermal and Visible Image Registration","authors":"Jiahao Xu, Xiufen Ye","doi":"10.1109/ICMA57826.2023.10216041","DOIUrl":"https://doi.org/10.1109/ICMA57826.2023.10216041","url":null,"abstract":"The cross-modality image registration task is becoming more and more important in the field of image fusion, thermal and visible image fusion can reduce the impact of environmental factors while maintaining the image’s texture and detail. However, the common registration methods always have the problem of overfitting and poor generalization when facing large modality span task. To register the thermal and visible images, a simple yet effective registration model called Dense-Cascade network is presented in this paper. Our network uses two independent branches to extract the cross-modality features respectively. In order to further reduce the regression error, we use a cascade structure by stacking the networks and using STN layer to achieve gradient backpropagation. Finally, to validate the robustness of our model, we add gaussian noise severity from 0-4 on thermal images to test the performance of our models, and the results show that the registration performances of our models are better than others.","PeriodicalId":151364,"journal":{"name":"2023 IEEE International Conference on Mechatronics and Automation (ICMA)","volume":"55 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-08-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114117693","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-08-06DOI: 10.1109/ICMA57826.2023.10215820
Jia Guo, Kang Hu, Jinting Liu
To ensure reliable localization for Unmanned Aerial Vehicles(UAVs) in the presence of uncertain speed of the Unmanned Ground Vehicle (UGV), this study examines the relative localization problem of UAV-UGV using distance and bearing measurements. A Correlation of Bearing and Distance-based Relative Localization (CBDRL) algorithm is proposed in this paper under this scenario. The estimation of altitude, distance, and angle are simplified into a representation of the relative positioning between the UAV and UGV. The relative height difference is measured using the barometer in the algorithm. To determine the relative distance, Time of Arrival (TOA) ranging and Ultra Wide Band (UWB) communication are utilized. The relative direction measurement is then determined using the correlations of bearing and distance. We integrate these observations with height, direction, and distance data in an Extended Kalman Filter(EKF) to provide accurate and reliable relative position estimates that allow the UAV to track the target. The simulation results indicate that the CBDRL method developed in this study is superior to previous relative localization algorithms that rely on multi-sensor fusion, and can significantly enhance the accuracy of UAV positioning provided that range and angle measurements are precise enough.
{"title":"Relative Localization for UAV-UGV Based on Correlation of Bearing and Distance","authors":"Jia Guo, Kang Hu, Jinting Liu","doi":"10.1109/ICMA57826.2023.10215820","DOIUrl":"https://doi.org/10.1109/ICMA57826.2023.10215820","url":null,"abstract":"To ensure reliable localization for Unmanned Aerial Vehicles(UAVs) in the presence of uncertain speed of the Unmanned Ground Vehicle (UGV), this study examines the relative localization problem of UAV-UGV using distance and bearing measurements. A Correlation of Bearing and Distance-based Relative Localization (CBDRL) algorithm is proposed in this paper under this scenario. The estimation of altitude, distance, and angle are simplified into a representation of the relative positioning between the UAV and UGV. The relative height difference is measured using the barometer in the algorithm. To determine the relative distance, Time of Arrival (TOA) ranging and Ultra Wide Band (UWB) communication are utilized. The relative direction measurement is then determined using the correlations of bearing and distance. We integrate these observations with height, direction, and distance data in an Extended Kalman Filter(EKF) to provide accurate and reliable relative position estimates that allow the UAV to track the target. The simulation results indicate that the CBDRL method developed in this study is superior to previous relative localization algorithms that rely on multi-sensor fusion, and can significantly enhance the accuracy of UAV positioning provided that range and angle measurements are precise enough.","PeriodicalId":151364,"journal":{"name":"2023 IEEE International Conference on Mechatronics and Automation (ICMA)","volume":"40 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-08-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117123813","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-08-06DOI: 10.1109/ICMA57826.2023.10216066
Oluwaleke Yusuf, MakiK . Habib
Hand Gesture Recognition (HGR) is a form of perceptual computing with applications in human-machine interaction, virtual/augmented reality, and human behavior analysis. Within the HGR domain, several frameworks have been developed with different combinations of input modalities and network architectures with varying levels of efficacy. Such frameworks maximized performance at the expense of increased hardware and computational requirements. These drawbacks can be tackled by transforming the relatively complex dynamic hand gesture recognition task into a simpler image classification task. This paper presents a skeleton-based HGR framework that implements data-level fusion for encoding spatiotemporal information from dynamic gestures into static representational images. Said static images are subsequently processed by a custom, end-to-end trainable multi-stream CNN architecture for gesture classification. Our framework reduces the hardware and computational requirements of the HGR task while remaining competitive with the state-of-the-art on the CNR, FPHA, LMDHG, SHREC2017, and DHG142S benchmark datasets. We demonstrated the practical utility of our framework by creating a lightweight real-time application that makes use of skeleton data extracted from RGB video streams captured by a standard inbuilt PC webcam. The application operates successfully with minimal CPU and RAM footprint while achieving 93.46% classification accuracy with approximately 2s latency at 15 frames per second.
{"title":"Development of a Lightweight Real-Time Application for Dynamic Hand Gesture Recognition","authors":"Oluwaleke Yusuf, MakiK . Habib","doi":"10.1109/ICMA57826.2023.10216066","DOIUrl":"https://doi.org/10.1109/ICMA57826.2023.10216066","url":null,"abstract":"Hand Gesture Recognition (HGR) is a form of perceptual computing with applications in human-machine interaction, virtual/augmented reality, and human behavior analysis. Within the HGR domain, several frameworks have been developed with different combinations of input modalities and network architectures with varying levels of efficacy. Such frameworks maximized performance at the expense of increased hardware and computational requirements. These drawbacks can be tackled by transforming the relatively complex dynamic hand gesture recognition task into a simpler image classification task. This paper presents a skeleton-based HGR framework that implements data-level fusion for encoding spatiotemporal information from dynamic gestures into static representational images. Said static images are subsequently processed by a custom, end-to-end trainable multi-stream CNN architecture for gesture classification. Our framework reduces the hardware and computational requirements of the HGR task while remaining competitive with the state-of-the-art on the CNR, FPHA, LMDHG, SHREC2017, and DHG142S benchmark datasets. We demonstrated the practical utility of our framework by creating a lightweight real-time application that makes use of skeleton data extracted from RGB video streams captured by a standard inbuilt PC webcam. The application operates successfully with minimal CPU and RAM footprint while achieving 93.46% classification accuracy with approximately 2s latency at 15 frames per second.","PeriodicalId":151364,"journal":{"name":"2023 IEEE International Conference on Mechatronics and Automation (ICMA)","volume":"109 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-08-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117165959","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Parallel-mechanism-based ultrasound robots bring the benefits of convenient robot-assisted remote ultrasound or automated scanning. During the use of this type of robot, accurate tracking and feedback of 6-DOF posture are critical for potential intelligent control and real-time navigation. However, this presents significant challenges due to the complex forward kinematics and environmental changes, making it difficult to apply in practical settings. To address this issue, we propose a novel 6-DOF pose tracking method based on an inertial measurement unit. This method directly senses the orientation and calculates the position using the intrinsic structural and kinematic characteristics of the parallel mechanisms. Therefore, it is possible to solve all posture information by only partial posture information. We conducted both simulation and physical experiments to evaluate effectiveness and robustness of the proposed method. The maximal tracking errors for the physical experiments are within the 2 mm and 0.6.°. The experimental results show promising performance, demonstrating the method’s potential for practical applications.
{"title":"A Novel IMU-Based 6-DOF Pose Tracking Method for Parallel-Mechanism-Based Ultrasound Robot","authors":"Zhaokun Deng, Xilong Hou, Peng-Cheng Zhang, Mingrui Hao, Cheng Chen, Shuangyi Wang","doi":"10.1109/ICMA57826.2023.10216060","DOIUrl":"https://doi.org/10.1109/ICMA57826.2023.10216060","url":null,"abstract":"Parallel-mechanism-based ultrasound robots bring the benefits of convenient robot-assisted remote ultrasound or automated scanning. During the use of this type of robot, accurate tracking and feedback of 6-DOF posture are critical for potential intelligent control and real-time navigation. However, this presents significant challenges due to the complex forward kinematics and environmental changes, making it difficult to apply in practical settings. To address this issue, we propose a novel 6-DOF pose tracking method based on an inertial measurement unit. This method directly senses the orientation and calculates the position using the intrinsic structural and kinematic characteristics of the parallel mechanisms. Therefore, it is possible to solve all posture information by only partial posture information. We conducted both simulation and physical experiments to evaluate effectiveness and robustness of the proposed method. The maximal tracking errors for the physical experiments are within the 2 mm and 0.6.°. The experimental results show promising performance, demonstrating the method’s potential for practical applications.","PeriodicalId":151364,"journal":{"name":"2023 IEEE International Conference on Mechatronics and Automation (ICMA)","volume":"49 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-08-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115839689","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-08-06DOI: 10.1109/ICMA57826.2023.10215522
Xiulong Cui, Jizhuang Fan, Zhihui Dong, Yue Ou, Jie Zhao
In this paper, we made an improved design on the traditional quadruped robot structure and developed a new small wheel-legged mobile robot structure. It has four legged configurations, which can be selected according to different situations, improving the mobility and operability of the robot. In wheeled mode, the robot can adjust its center of gravity according to actual needs. Including the rotational degree of freedom of the wheel, each leg has four degrees of freedom. Each leg adopts a series structure and moves the knee drive motor to the fuselage position using a parallel four-link mechanism and is symmetrically arranged with the hip drive motor. A wheel is installed at the end of the shank, and a DC brushless motor is built inside it, enabling the robot to have the function of wheeled motion. This article introduces the details of the wheel-legged robot, emphasizing the innovative design of the physical structure and the diversity of deformation. The forward kinematics was analyzed. Based on the full-elbow configuration (a type of legged configuration), a cycloidal motion optimization simulation was completed for the single leg structure. Simulation experiment has proven that the optimized robot’s single leg motion is smoother. Finally, we conducted a typical experiment using an embedded controller to control single leg motion. The experimental results verify that the optimized cycloidal motion of the robot meets the desired motion effect.
{"title":"Structural Design and Experiment of a Small Wheel-legged Mobile Robot with Variable Configuration","authors":"Xiulong Cui, Jizhuang Fan, Zhihui Dong, Yue Ou, Jie Zhao","doi":"10.1109/ICMA57826.2023.10215522","DOIUrl":"https://doi.org/10.1109/ICMA57826.2023.10215522","url":null,"abstract":"In this paper, we made an improved design on the traditional quadruped robot structure and developed a new small wheel-legged mobile robot structure. It has four legged configurations, which can be selected according to different situations, improving the mobility and operability of the robot. In wheeled mode, the robot can adjust its center of gravity according to actual needs. Including the rotational degree of freedom of the wheel, each leg has four degrees of freedom. Each leg adopts a series structure and moves the knee drive motor to the fuselage position using a parallel four-link mechanism and is symmetrically arranged with the hip drive motor. A wheel is installed at the end of the shank, and a DC brushless motor is built inside it, enabling the robot to have the function of wheeled motion. This article introduces the details of the wheel-legged robot, emphasizing the innovative design of the physical structure and the diversity of deformation. The forward kinematics was analyzed. Based on the full-elbow configuration (a type of legged configuration), a cycloidal motion optimization simulation was completed for the single leg structure. Simulation experiment has proven that the optimized robot’s single leg motion is smoother. Finally, we conducted a typical experiment using an embedded controller to control single leg motion. The experimental results verify that the optimized cycloidal motion of the robot meets the desired motion effect.","PeriodicalId":151364,"journal":{"name":"2023 IEEE International Conference on Mechatronics and Automation (ICMA)","volume":"60 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-08-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115167205","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-08-06DOI: 10.1109/ICMA57826.2023.10216262
Xinru Lin, Liwei Shi, Shuxiang Guo
With the increased demand for the exploration of the ocean, the focus of research in the field of robotics has changed. Amphibious robots show broad application prospects. In order to design an amphibious robot adapted to multiple environments, a feet-legs device inspired by natatores was designed. Each webbed foot is made up of three sets of four-linked rods that allow the toe bones to fold and shrink, enabling better imitation of the webbed feet motion of natatores paddling. In order to analyse and optimize the mechanical strength of the critical components of the device during operation, static analysis and structural improvements are made to the webbed feet and the roller of the cylindrical cam. The results show that the static strength of improved webbed feet and the cam roller meet the requirements.
{"title":"Design and Static Analysis of Feet-legs Device of a Natatores-like Amphibious Robot","authors":"Xinru Lin, Liwei Shi, Shuxiang Guo","doi":"10.1109/ICMA57826.2023.10216262","DOIUrl":"https://doi.org/10.1109/ICMA57826.2023.10216262","url":null,"abstract":"With the increased demand for the exploration of the ocean, the focus of research in the field of robotics has changed. Amphibious robots show broad application prospects. In order to design an amphibious robot adapted to multiple environments, a feet-legs device inspired by natatores was designed. Each webbed foot is made up of three sets of four-linked rods that allow the toe bones to fold and shrink, enabling better imitation of the webbed feet motion of natatores paddling. In order to analyse and optimize the mechanical strength of the critical components of the device during operation, static analysis and structural improvements are made to the webbed feet and the roller of the cylindrical cam. The results show that the static strength of improved webbed feet and the cam roller meet the requirements.","PeriodicalId":151364,"journal":{"name":"2023 IEEE International Conference on Mechatronics and Automation (ICMA)","volume":"80 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-08-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123090021","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-08-06DOI: 10.1109/ICMA57826.2023.10216153
Panhong Zhang, Tao Ni
In this paper, a human-machine collaborative motion planning method is put forward to deal with the matter that there is a strong dependence on the operator and the operation load brought by it in the course of task execution. In this paper, the human intention obtained by the hidden markov model and the communication delay in the teleoperation system constitute the target point of the heuristic motion planning. Then a heuristic method for vehicle motion planning in intricate environment based on the target point is presented, so that the vehicle can also move successfully after the change of human intention is monitored. The heuristic motion planning method based on human intention framework is confirmed in the numerical experiment, then its outstanding performance in the intricate environment is evaluated, and the feasibility of the proposed method has been verified.
{"title":"A Heuristic Motion Planning Method based on Human Intention for Teleoperation Systems","authors":"Panhong Zhang, Tao Ni","doi":"10.1109/ICMA57826.2023.10216153","DOIUrl":"https://doi.org/10.1109/ICMA57826.2023.10216153","url":null,"abstract":"In this paper, a human-machine collaborative motion planning method is put forward to deal with the matter that there is a strong dependence on the operator and the operation load brought by it in the course of task execution. In this paper, the human intention obtained by the hidden markov model and the communication delay in the teleoperation system constitute the target point of the heuristic motion planning. Then a heuristic method for vehicle motion planning in intricate environment based on the target point is presented, so that the vehicle can also move successfully after the change of human intention is monitored. The heuristic motion planning method based on human intention framework is confirmed in the numerical experiment, then its outstanding performance in the intricate environment is evaluated, and the feasibility of the proposed method has been verified.","PeriodicalId":151364,"journal":{"name":"2023 IEEE International Conference on Mechatronics and Automation (ICMA)","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-08-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124473551","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-08-06DOI: 10.1109/ICMA57826.2023.10215794
Zhenzhuo Yan, Xifeng Gao, Yifan Li, Pengyue Zhao, Z. Deng
The topic of position control for continuum manipulators (CMs) remains open and yet to be well explored and developed. Current applications of CMs focus on employing static or quasi-dynamic controllers built upon kinematic models or linearity in the joint space, resulting in a loss of the rich dynamics of a system. This paper presents a model-based reinforcement learning scheme for position control of a class of CMs with strong nonlinearity and input coupling, which includes a probabilistic dynamics model as the dynamic forward model and a policy update approach for the closed-loop policy. The effectiveness of the scheme is verified on a dual-segment CM actuated by pneumatic artificial muscles, and the experimental results confirm that such scheme can obtain good results with only a limited number of samples and interactions.
{"title":"Model-Based Reinforcement Learning for Position Control of Continuum Manipulators Actuated by Pneumatic Artificial Muscles","authors":"Zhenzhuo Yan, Xifeng Gao, Yifan Li, Pengyue Zhao, Z. Deng","doi":"10.1109/ICMA57826.2023.10215794","DOIUrl":"https://doi.org/10.1109/ICMA57826.2023.10215794","url":null,"abstract":"The topic of position control for continuum manipulators (CMs) remains open and yet to be well explored and developed. Current applications of CMs focus on employing static or quasi-dynamic controllers built upon kinematic models or linearity in the joint space, resulting in a loss of the rich dynamics of a system. This paper presents a model-based reinforcement learning scheme for position control of a class of CMs with strong nonlinearity and input coupling, which includes a probabilistic dynamics model as the dynamic forward model and a policy update approach for the closed-loop policy. The effectiveness of the scheme is verified on a dual-segment CM actuated by pneumatic artificial muscles, and the experimental results confirm that such scheme can obtain good results with only a limited number of samples and interactions.","PeriodicalId":151364,"journal":{"name":"2023 IEEE International Conference on Mechatronics and Automation (ICMA)","volume":"75 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-08-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121728629","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-08-06DOI: 10.1109/ICMA57826.2023.10216053
Jiale Tian, Peng Shang, Jian Zhou, Lie Yu
The rotor’s high-speed development trend has brought new challenges to the rotor’s dynamic balancing which has become a key factor affecting the safe and stable operation of the rotating machinery. However, there is still a lack of guidance for the dynamic balancing design for high-speed rotor bearing system. In this paper, the compliant foil gas bearings are analyzed considering the nonlinear deflection of the top foil, and the static and dynamic characteristics are obtained. Then, the rotor’s dynamic responses to various unbalance values are calculated theoretically. Finally, the maximum allowable unbalance of a high-speed rotor-bearing system is given based on the dynamic load capacity of the compliant foil gas bearing. This study can provide theoretical support for the high-speed rotor’s dynamic balancing design.
{"title":"Study on the Maximum Allowable Unbalance of the High-speed Rotor Based on the Dynamic Load Capacity of Gas Foil Bearings","authors":"Jiale Tian, Peng Shang, Jian Zhou, Lie Yu","doi":"10.1109/ICMA57826.2023.10216053","DOIUrl":"https://doi.org/10.1109/ICMA57826.2023.10216053","url":null,"abstract":"The rotor’s high-speed development trend has brought new challenges to the rotor’s dynamic balancing which has become a key factor affecting the safe and stable operation of the rotating machinery. However, there is still a lack of guidance for the dynamic balancing design for high-speed rotor bearing system. In this paper, the compliant foil gas bearings are analyzed considering the nonlinear deflection of the top foil, and the static and dynamic characteristics are obtained. Then, the rotor’s dynamic responses to various unbalance values are calculated theoretically. Finally, the maximum allowable unbalance of a high-speed rotor-bearing system is given based on the dynamic load capacity of the compliant foil gas bearing. This study can provide theoretical support for the high-speed rotor’s dynamic balancing design.","PeriodicalId":151364,"journal":{"name":"2023 IEEE International Conference on Mechatronics and Automation (ICMA)","volume":"30 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-08-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122171737","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-08-06DOI: 10.1109/ICMA57826.2023.10215969
Yongchang Zhu, Enzeng Dong, Jigang Tong, Sen Yang, Zufeng Zhang, Wenyu Li
Target detection is a significant research direction in the realm of computer vision, aimed at analyzing and localizing targets in input images to obtain their categories and locations. However, this task is faced with several challenges in complex scenarios, such as large variations in target scales, similarity to the background, dense number, and overlapping with each other, making it difficult for the target detection model to extract different scales and discriminative target features adaptively, leading to low detection accuracy and weak generalization ability. To tackle the aforementioned problems, this study proposes a novel approach by optimizing the detection head of the model, invoking decoupled detection heads, dividing the classification and localization tasks, and optimizing for small targets. This approach effectively reduces the false detection rate and the missed detection rate, enabling the network to better discriminate and recognize distant objects as well as objects of relatively small size. Moreover, the proposed approach outperforms the yolov7 model by 0.8% in the COCO dataset, not only improving the accuracy of detecting distant and small targets but also exhibiting better performance in conventional target detection. Additionally, this approach is capable of handling small targets, overlapping targets, or relatively dense targets, significantly expanding its practical use range.
{"title":"Deep Neural Network Based Object Detection Algorithm With optimized Detection Head for Small Targets","authors":"Yongchang Zhu, Enzeng Dong, Jigang Tong, Sen Yang, Zufeng Zhang, Wenyu Li","doi":"10.1109/ICMA57826.2023.10215969","DOIUrl":"https://doi.org/10.1109/ICMA57826.2023.10215969","url":null,"abstract":"Target detection is a significant research direction in the realm of computer vision, aimed at analyzing and localizing targets in input images to obtain their categories and locations. However, this task is faced with several challenges in complex scenarios, such as large variations in target scales, similarity to the background, dense number, and overlapping with each other, making it difficult for the target detection model to extract different scales and discriminative target features adaptively, leading to low detection accuracy and weak generalization ability. To tackle the aforementioned problems, this study proposes a novel approach by optimizing the detection head of the model, invoking decoupled detection heads, dividing the classification and localization tasks, and optimizing for small targets. This approach effectively reduces the false detection rate and the missed detection rate, enabling the network to better discriminate and recognize distant objects as well as objects of relatively small size. Moreover, the proposed approach outperforms the yolov7 model by 0.8% in the COCO dataset, not only improving the accuracy of detecting distant and small targets but also exhibiting better performance in conventional target detection. Additionally, this approach is capable of handling small targets, overlapping targets, or relatively dense targets, significantly expanding its practical use range.","PeriodicalId":151364,"journal":{"name":"2023 IEEE International Conference on Mechatronics and Automation (ICMA)","volume":"50 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-08-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121073590","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：