Pub Date : 2021-10-12DOI: 10.5302/j.icros.2021.21.0143
Choi Keonghun, J. Ha
In the case of the unmanned surveillance system field, even if it is the same object, the detection result will be different depending on the state of the object and the configuration of the surrounding environment. Therefore, artificial intelligence for unmanned surveillance needs to understand the environment on the image, understand the state of the object within the image, and understand the relationship between them. For this purpose, in this study, a transformed transformer structure that can receive a single image, which is 2D data, as an input, unlike splitting one image into a certain size and using it as an input, is presented, and the effect between neighboring pixels is considered by using a segmentation model to which it is applied. A possible background classification model was constructed.
{"title":"Visual surveillance transformer","authors":"Choi Keonghun, J. Ha","doi":"10.5302/j.icros.2021.21.0143","DOIUrl":"https://doi.org/10.5302/j.icros.2021.21.0143","url":null,"abstract":"In the case of the unmanned surveillance system field, even if it is the same object, the detection result will be different depending on the state of the object and the configuration of the surrounding environment. Therefore, artificial intelligence for unmanned surveillance needs to understand the environment on the image, understand the state of the object within the image, and understand the relationship between them. For this purpose, in this study, a transformed transformer structure that can receive a single image, which is 2D data, as an input, unlike splitting one image into a certain size and using it as an input, is presented, and the effect between neighboring pixels is considered by using a segmentation model to which it is applied. A possible background classification model was constructed.","PeriodicalId":411064,"journal":{"name":"2021 21st International Conference on Control, Automation and Systems (ICCAS)","volume":"289 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"120913060","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 : 2021-10-12DOI: 10.23919/ICCAS52745.2021.9649997
Pavlo Vlastos, A. Hunter, R. Curry, Carlos Isaac Espinosa Ramirez, G. Elkaim
Gaussian process regression and ordinary kriging are effective methods for spatial estimation, but are generally not used in online trajectory-planning applications for autonomous vehicles. A common use for kriging is spatial estimation for exploration. Kriging is limited by the necessary covariance matrix inversion and its computational complexity of O(n3), where $n$ represents the number of measurements taken in a sparsely-sampled field. Using the Sherman-Morison matrix inversion lemma, the complexity can be reduced to O(n2). This work focuses on further improving the computational time required to conduct spatial estimation with partitioned ordinary kriging (POK) for online trajectory-planning. A recursive algorithm is introduced to quickly subdivide a field for local kriging, reducing the computation time. We show computational time decreases between ordinary kriging with a regular inverse (OK), the iterative inverse ordinary kriging (IIOK), and POK with the iterative inverse method. Computation times are also compared between OK, IIOK, and POK methods for trajectory planning using a highest variance criterion and linear trajectory segments.
{"title":"Partitioned Gaussian Process Regression for Online Trajectory Planning for Autonomous Vehicles","authors":"Pavlo Vlastos, A. Hunter, R. Curry, Carlos Isaac Espinosa Ramirez, G. Elkaim","doi":"10.23919/ICCAS52745.2021.9649997","DOIUrl":"https://doi.org/10.23919/ICCAS52745.2021.9649997","url":null,"abstract":"Gaussian process regression and ordinary kriging are effective methods for spatial estimation, but are generally not used in online trajectory-planning applications for autonomous vehicles. A common use for kriging is spatial estimation for exploration. Kriging is limited by the necessary covariance matrix inversion and its computational complexity of O(n3), where $n$ represents the number of measurements taken in a sparsely-sampled field. Using the Sherman-Morison matrix inversion lemma, the complexity can be reduced to O(n2). This work focuses on further improving the computational time required to conduct spatial estimation with partitioned ordinary kriging (POK) for online trajectory-planning. A recursive algorithm is introduced to quickly subdivide a field for local kriging, reducing the computation time. We show computational time decreases between ordinary kriging with a regular inverse (OK), the iterative inverse ordinary kriging (IIOK), and POK with the iterative inverse method. Computation times are also compared between OK, IIOK, and POK methods for trajectory planning using a highest variance criterion and linear trajectory segments.","PeriodicalId":411064,"journal":{"name":"2021 21st International Conference on Control, Automation and Systems (ICCAS)","volume":"30 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127177030","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 : 2021-10-12DOI: 10.23919/ICCAS52745.2021.9649854
Y. Quan, Jin Sung Kim, C. Chung
In this paper, we propose a Robust Model Predictive Control combined with Control Barrier Function (RMPC-CBF) for a nonholonomic robot with obstacle avoidance subject to additive input disturbances. Both Input-to-State Stability (ISS) and Input-to-State Safety (ISSf) are provided to theoretically guarantee the system's stability and safety. CBF-based safety conditions are formulated as constraints inside a robust MPC strategy. Robust satisfaction of the constraints is ensured by tightening the state constraint set. With admissible disturbances under a certain bound, ISS and robust recursive feasibility are guaranteed by computing the terminal region and state constraint set. For obstacle avoidance, Input-to-State Safe Control Barrier Function (ISSf-CBF) is chosen to provide robust set safety for the dynamic systems under input disturbances, which always guarantees that states stay inside or close to the safe set. With the proposed method, the future state prediction is taken into consideration and optimal performance is accomplished via MPC, and the system's safety is ensured via CBF. Numerical simulation results confirm the effectiveness and validity of the proposed approach.
{"title":"Robust Model Predictive Control with Control Barrier Function for Nonholonomic Robots with Obstacle Avoidance","authors":"Y. Quan, Jin Sung Kim, C. Chung","doi":"10.23919/ICCAS52745.2021.9649854","DOIUrl":"https://doi.org/10.23919/ICCAS52745.2021.9649854","url":null,"abstract":"In this paper, we propose a Robust Model Predictive Control combined with Control Barrier Function (RMPC-CBF) for a nonholonomic robot with obstacle avoidance subject to additive input disturbances. Both Input-to-State Stability (ISS) and Input-to-State Safety (ISSf) are provided to theoretically guarantee the system's stability and safety. CBF-based safety conditions are formulated as constraints inside a robust MPC strategy. Robust satisfaction of the constraints is ensured by tightening the state constraint set. With admissible disturbances under a certain bound, ISS and robust recursive feasibility are guaranteed by computing the terminal region and state constraint set. For obstacle avoidance, Input-to-State Safe Control Barrier Function (ISSf-CBF) is chosen to provide robust set safety for the dynamic systems under input disturbances, which always guarantees that states stay inside or close to the safe set. With the proposed method, the future state prediction is taken into consideration and optimal performance is accomplished via MPC, and the system's safety is ensured via CBF. Numerical simulation results confirm the effectiveness and validity of the proposed approach.","PeriodicalId":411064,"journal":{"name":"2021 21st International Conference on Control, Automation and Systems (ICCAS)","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126580340","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 : 2021-10-12DOI: 10.23919/ICCAS52745.2021.9650060
Wen Du, Yusheng Wei, Mingjun Du
In this paper, we consider the distributed optimization problem under structurally balanced signed graph. First, we convert the original distributed optimization problem into a conditional minimum problem under the condition that the graph is structurally balanced. Our goal is to find the saddle points of augmented Lagrange function. Inspired by the Lagrange multiplier method, we present our algorithms for both undirected graph and digraph, and show that our algorithms asymptotically converge to the global minimizer. Particularly, our algorithms for digraph can not only handle the weight balanced case but the weight unbalanced case. We show that the unsigned graph is a special case of our signed graph cases. Finally, theoretical results are illustrated by numerical simulations.
{"title":"Distributed Optimization Algorithms on Structurally Balanced Signed Networks","authors":"Wen Du, Yusheng Wei, Mingjun Du","doi":"10.23919/ICCAS52745.2021.9650060","DOIUrl":"https://doi.org/10.23919/ICCAS52745.2021.9650060","url":null,"abstract":"In this paper, we consider the distributed optimization problem under structurally balanced signed graph. First, we convert the original distributed optimization problem into a conditional minimum problem under the condition that the graph is structurally balanced. Our goal is to find the saddle points of augmented Lagrange function. Inspired by the Lagrange multiplier method, we present our algorithms for both undirected graph and digraph, and show that our algorithms asymptotically converge to the global minimizer. Particularly, our algorithms for digraph can not only handle the weight balanced case but the weight unbalanced case. We show that the unsigned graph is a special case of our signed graph cases. Finally, theoretical results are illustrated by numerical simulations.","PeriodicalId":411064,"journal":{"name":"2021 21st International Conference on Control, Automation and Systems (ICCAS)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129193916","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 : 2021-10-12DOI: 10.23919/ICCAS52745.2021.9649943
Changhyeong Lee, Junwoo Jason Son, Hakjun Lee, Soohee Han
This paper analyze the energy consumption of a new concept of tethered quadrotor system, downward tethered quadrotor (DTQ), to provide insight about using DTQ. DTQ is one of a tethered quadrotor system in which station is located above the flight level of the quadrotor. Thanks to its system layout, the tether can play an important role in energy efficient flight of the quadrotor by adjusting the tension properly. In order to extremize the advantage of DTQ, optimization problem for mechanical power consumption of DTQ in hovering states is formulated. The DTQ in this paper considers misalignment between tether link point on the quadrotor and center of mass (CoM) of the quadrotor to simulate real system. Numerical simulation illustrates efficiency of the proposed method.
{"title":"Energy Consumption Analysis of Downward-Tethered Quadcopter","authors":"Changhyeong Lee, Junwoo Jason Son, Hakjun Lee, Soohee Han","doi":"10.23919/ICCAS52745.2021.9649943","DOIUrl":"https://doi.org/10.23919/ICCAS52745.2021.9649943","url":null,"abstract":"This paper analyze the energy consumption of a new concept of tethered quadrotor system, downward tethered quadrotor (DTQ), to provide insight about using DTQ. DTQ is one of a tethered quadrotor system in which station is located above the flight level of the quadrotor. Thanks to its system layout, the tether can play an important role in energy efficient flight of the quadrotor by adjusting the tension properly. In order to extremize the advantage of DTQ, optimization problem for mechanical power consumption of DTQ in hovering states is formulated. The DTQ in this paper considers misalignment between tether link point on the quadrotor and center of mass (CoM) of the quadrotor to simulate real system. Numerical simulation illustrates efficiency of the proposed method.","PeriodicalId":411064,"journal":{"name":"2021 21st International Conference on Control, Automation and Systems (ICCAS)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133784197","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 : 2021-10-12DOI: 10.23919/ICCAS52745.2021.9649973
Leonid Kostrykin, Claus Rohr, K. Rohr
Large-scale image stitching of temporal video data acquired by a small robot can facilitate the inspection process of electric generators, reduce the inspection time, and improve the reliability. However, the image data poses a number of challenges due to the small field of view, lack of distinct texture, specular highlights, and other image artifacts. We introduce a novel image stitching method, which generates composite images of generator wedges from temporal videos using intensity-based registration and non-linear blending. In contrast to previous intensity-based registration approaches, our global method simultaneously exploits the information of all image frames of a video and directly determines the global image translations. We propose a suitable energy function and employ a graph-based method for globally optimal minimization in linear runtime. Regularization is used to exploit physical knowledge about the application domain which improves the robustness. We have applied our approach to temporal video data of rotor wedges and performed a comparison with previous methods. We found that our method yields superior results.
{"title":"Globally Optimal and Scalable Video Image Stitching for Robotic Visual Inspection of Electric Generators","authors":"Leonid Kostrykin, Claus Rohr, K. Rohr","doi":"10.23919/ICCAS52745.2021.9649973","DOIUrl":"https://doi.org/10.23919/ICCAS52745.2021.9649973","url":null,"abstract":"Large-scale image stitching of temporal video data acquired by a small robot can facilitate the inspection process of electric generators, reduce the inspection time, and improve the reliability. However, the image data poses a number of challenges due to the small field of view, lack of distinct texture, specular highlights, and other image artifacts. We introduce a novel image stitching method, which generates composite images of generator wedges from temporal videos using intensity-based registration and non-linear blending. In contrast to previous intensity-based registration approaches, our global method simultaneously exploits the information of all image frames of a video and directly determines the global image translations. We propose a suitable energy function and employ a graph-based method for globally optimal minimization in linear runtime. Regularization is used to exploit physical knowledge about the application domain which improves the robustness. We have applied our approach to temporal video data of rotor wedges and performed a comparison with previous methods. We found that our method yields superior results.","PeriodicalId":411064,"journal":{"name":"2021 21st International Conference on Control, Automation and Systems (ICCAS)","volume":"19 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133589572","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 : 2021-10-12DOI: 10.23919/ICCAS52745.2021.9649876
J. Yoo, Donghwi Kim, D. Shim
This study designed an intelligent control system for autonomous air-to-air combat and verified it in a realtime flight simulation. Previous studies of aerial combat have required significant effort to design agile control actions for different engagement conditions. In this work, optimal flight control under random engagement conditions was performed by using reinforcement learning and recurrent neural networks. A target trajectory was predicted using Sequence-to-Sequence model with LSTM, for occupying an advantageous location from an enemy aircraft in a close engagement. In addition, this study proposed an algorithm with improved performance compared to the existing algorithm. The result of the study confirmed that the maneuvers of trained agent were similar to the performance of human pilots and the future position of the enemy was tracked by own ship aircraft.
{"title":"Deep Reinforcement Learning based Autonomous Air-to-Air Combat using Target Trajectory Prediction","authors":"J. Yoo, Donghwi Kim, D. Shim","doi":"10.23919/ICCAS52745.2021.9649876","DOIUrl":"https://doi.org/10.23919/ICCAS52745.2021.9649876","url":null,"abstract":"This study designed an intelligent control system for autonomous air-to-air combat and verified it in a realtime flight simulation. Previous studies of aerial combat have required significant effort to design agile control actions for different engagement conditions. In this work, optimal flight control under random engagement conditions was performed by using reinforcement learning and recurrent neural networks. A target trajectory was predicted using Sequence-to-Sequence model with LSTM, for occupying an advantageous location from an enemy aircraft in a close engagement. In addition, this study proposed an algorithm with improved performance compared to the existing algorithm. The result of the study confirmed that the maneuvers of trained agent were similar to the performance of human pilots and the future position of the enemy was tracked by own ship aircraft.","PeriodicalId":411064,"journal":{"name":"2021 21st International Conference on Control, Automation and Systems (ICCAS)","volume":"59 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128375946","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 : 2021-10-12DOI: 10.23919/ICCAS52745.2021.9649797
Nozomi Shime, Tohru Kamiya, T. Ishida
One of the difficulties of cancer is the metastasis of cancer through the lymph nodes. Therefore, early detection and treatment of cancerous is important. Visual screening is one of useful tool for diagnosing of the cancer. However, the number of images obtained at a time is large, and the burden on the reading physician is increasing. Furthermore, reading of image is based on the subjective judgment of the physician, which may lead to different diagnostic results. To solve these problems, computer aided diagnosis (CAD) systems have been attracting attention in recent years. One of the CAD systems is the temporal subtraction image technology. In this paper, we propose an image alignment method for generating temporal subtraction images from images taken before and after contrast agent was administered to the cervical lymph nodes of the same subject. In addition, we propose an image analysis method that suppresses overextraction of lymph node candidate regions on the temporal subtraction image based on the features of lymph nodes. We applied the proposed method to the CT images of three sets and compared the temporal subtraction images with the final lymph node extraction images, and confirmed that the proposed method can suppress the overextraction of lymph node regions.
{"title":"Extraction of Cervical Lymph Nodes Based on Three-Dimensional Image Registration","authors":"Nozomi Shime, Tohru Kamiya, T. Ishida","doi":"10.23919/ICCAS52745.2021.9649797","DOIUrl":"https://doi.org/10.23919/ICCAS52745.2021.9649797","url":null,"abstract":"One of the difficulties of cancer is the metastasis of cancer through the lymph nodes. Therefore, early detection and treatment of cancerous is important. Visual screening is one of useful tool for diagnosing of the cancer. However, the number of images obtained at a time is large, and the burden on the reading physician is increasing. Furthermore, reading of image is based on the subjective judgment of the physician, which may lead to different diagnostic results. To solve these problems, computer aided diagnosis (CAD) systems have been attracting attention in recent years. One of the CAD systems is the temporal subtraction image technology. In this paper, we propose an image alignment method for generating temporal subtraction images from images taken before and after contrast agent was administered to the cervical lymph nodes of the same subject. In addition, we propose an image analysis method that suppresses overextraction of lymph node candidate regions on the temporal subtraction image based on the features of lymph nodes. We applied the proposed method to the CT images of three sets and compared the temporal subtraction images with the final lymph node extraction images, and confirmed that the proposed method can suppress the overextraction of lymph node regions.","PeriodicalId":411064,"journal":{"name":"2021 21st International Conference on Control, Automation and Systems (ICCAS)","volume":"163 4","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132227631","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 : 2021-10-12DOI: 10.23919/ICCAS52745.2021.9649770
Jee-Soo Ha, Soo-Ri Im, W. Lee, Dong-Hoon Kim, Jaekwan Ryu
This paper proposes a dynamic obstacle detection system for USV based on marine radar and Electronic Navigational chart (ENC), the most common navigation sensors on ships. This system has the advantage of enabling simple obstacle recognition without the need to additionally mount expensive equipment. In this system, we generated two types of grid maps: one is plan position indicator (PPI) images from marine radar, the other is a hull information-based grid map extracted from ENC. By accumulating the two grid map images, obstacles that appear repeatedly are classified as fixed obstacles, and obstacles that move as the grid map is updated are classified as dynamic obstacles. The proposed obstacle detection system was installed in the Sea Sword USV developed by LIGNex1 and tested in a marine environment. The system proposed in the experiment recognized the small rubber boat as a dynamic obstacle and the surrounding environment as a static obstacle. Along with our proposed obstacle detection system, it is possible to recognize obstacles through ENC and radar, which are essential equipment for ships, without video equipment.
{"title":"Radar based Obstacle Detection System for Autonomous Unmanned Surface Vehicles","authors":"Jee-Soo Ha, Soo-Ri Im, W. Lee, Dong-Hoon Kim, Jaekwan Ryu","doi":"10.23919/ICCAS52745.2021.9649770","DOIUrl":"https://doi.org/10.23919/ICCAS52745.2021.9649770","url":null,"abstract":"This paper proposes a dynamic obstacle detection system for USV based on marine radar and Electronic Navigational chart (ENC), the most common navigation sensors on ships. This system has the advantage of enabling simple obstacle recognition without the need to additionally mount expensive equipment. In this system, we generated two types of grid maps: one is plan position indicator (PPI) images from marine radar, the other is a hull information-based grid map extracted from ENC. By accumulating the two grid map images, obstacles that appear repeatedly are classified as fixed obstacles, and obstacles that move as the grid map is updated are classified as dynamic obstacles. The proposed obstacle detection system was installed in the Sea Sword USV developed by LIGNex1 and tested in a marine environment. The system proposed in the experiment recognized the small rubber boat as a dynamic obstacle and the surrounding environment as a static obstacle. Along with our proposed obstacle detection system, it is possible to recognize obstacles through ENC and radar, which are essential equipment for ships, without video equipment.","PeriodicalId":411064,"journal":{"name":"2021 21st International Conference on Control, Automation and Systems (ICCAS)","volume":"20 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131703759","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 : 2021-10-12DOI: 10.23919/ICCAS52745.2021.9650057
Jongchan Kim, In-Deok Park, Sungho Kim
Recently, research related to Advanced Driver Assistance Systems is active. In this paper, EfficientDet Fusion Framework for multi-spectral pedestrian detection is constructed through Sum, Max, and Concatenation at the feature level. In the experiment, it was confirmed that the performance improvement of the convergence multispectral network was quantitatively improved by about 10% compared to the single spectral network. In addition, it shows that the shortcomings of a single spectral can be actually compensated through the resulting image. In the future, various fusion studies will be conducted based on the EfficientDet Fusion Framework.
{"title":"A Fusion Framework for Multi-Spectral Pedestrian Detection using EfficientDet","authors":"Jongchan Kim, In-Deok Park, Sungho Kim","doi":"10.23919/ICCAS52745.2021.9650057","DOIUrl":"https://doi.org/10.23919/ICCAS52745.2021.9650057","url":null,"abstract":"Recently, research related to Advanced Driver Assistance Systems is active. In this paper, EfficientDet Fusion Framework for multi-spectral pedestrian detection is constructed through Sum, Max, and Concatenation at the feature level. In the experiment, it was confirmed that the performance improvement of the convergence multispectral network was quantitatively improved by about 10% compared to the single spectral network. In addition, it shows that the shortcomings of a single spectral can be actually compensated through the resulting image. In the future, various fusion studies will be conducted based on the EfficientDet Fusion Framework.","PeriodicalId":411064,"journal":{"name":"2021 21st International Conference on Control, Automation and Systems (ICCAS)","volume":"54 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132189451","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：