Computing image feature pyramid has been a common approach in pedestrian detection for improving detection accuracy. However, building feature pyramid is a time consuming task. In this paper we propose a new multi-scale classifier based method. We approximate the nearby scale classifier instead of extracting features multiple times form the resizing images. These approximated classifiers can be applied to achieve object detection without image resizing. In addition, we introduce a new feature, BPG (Binary Pattern of Gradient), to further accelerate the feature extraction speed. The experimental result demonstrates that the new feature is efficient in pedestrian detection. It is also proved that the proposed method not only reduces the detection speed, but also has performance comparable to some state-of-the-art pedestrian detection approaches.
计算图像特征金字塔是行人检测中提高检测精度的常用方法。然而，构建特征金字塔是一项耗时的任务。本文提出了一种新的基于多尺度分类器的分类方法。我们近似邻近尺度分类器，而不是从调整大小的图像中多次提取特征。这些近似分类器可以在不调整图像大小的情况下实现目标检测。此外，我们引入了新的特征BPG (Binary Pattern of Gradient)，进一步加快了特征提取的速度。实验结果表明，该特征在行人检测中是有效的。实验还证明，该方法不仅降低了检测速度，而且性能与目前一些最先进的行人检测方法相当。
{"title":"Fast Pedestrian Detection with Multi-scale Classifiers","authors":"Baoyin Yu, Yingdong Ma, Jun Li","doi":"10.1109/CIIS.2017.41","DOIUrl":"https://doi.org/10.1109/CIIS.2017.41","url":null,"abstract":"Computing image feature pyramid has been a common approach in pedestrian detection for improving detection accuracy. However, building feature pyramid is a time consuming task. In this paper we propose a new multi-scale classifier based method. We approximate the nearby scale classifier instead of extracting features multiple times form the resizing images. These approximated classifiers can be applied to achieve object detection without image resizing. In addition, we introduce a new feature, BPG (Binary Pattern of Gradient), to further accelerate the feature extraction speed. The experimental result demonstrates that the new feature is efficient in pedestrian detection. It is also proved that the proposed method not only reduces the detection speed, but also has performance comparable to some state-of-the-art pedestrian detection approaches.","PeriodicalId":254342,"journal":{"name":"2017 International Conference on Computing Intelligence and Information System (CIIS)","volume":"31 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128648403","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}
After studying the working principle of feed-forward neural network and analyzing network structure and the learning mechanism of BP neural network and the extreme learning machine (ELM), a prediction model, GA-ELM, is proposed based on genetic algorithm to optimize the learning machine limit. The genetic algorithm is used to select the weights and thresholds of ELM neural network, and the optimal weights and thresholds are used to determine the connection weights between the hidden layer and the output layer. Further, this model is combined with the grey system model to correct the residual of GM, and then GM-GA-ELM combination forecasting model is established. Compared with BP model, GA-BP model and standard ELM model, it is further verified that the predicting accuracy and running time of the proposed model are better.
在研究了前馈神经网络工作原理的基础上，分析了BP神经网络和极限学习机(ELM)的网络结构和学习机理，提出了一种基于遗传算法优化学习机极限的GA-ELM预测模型。采用遗传算法选择ELM神经网络的权值和阈值，并利用最优权值和阈值确定隐含层与输出层之间的连接权值。进一步，将该模型与灰色系统模型相结合，对GM残差进行校正，建立GM- ga - elm组合预测模型。通过与BP模型、GA-BP模型和标准ELM模型的比较，进一步验证了该模型具有更好的预测精度和运行时间。
{"title":"A Combination Forecasting Model of Extreme Learning Machine Based on Genetic Algorithm Optimization","authors":"Zhiheng Yu, Chengli Zhao","doi":"10.1109/CIIS.2017.14","DOIUrl":"https://doi.org/10.1109/CIIS.2017.14","url":null,"abstract":"After studying the working principle of feed-forward neural network and analyzing network structure and the learning mechanism of BP neural network and the extreme learning machine (ELM), a prediction model, GA-ELM, is proposed based on genetic algorithm to optimize the learning machine limit. The genetic algorithm is used to select the weights and thresholds of ELM neural network, and the optimal weights and thresholds are used to determine the connection weights between the hidden layer and the output layer. Further, this model is combined with the grey system model to correct the residual of GM, and then GM-GA-ELM combination forecasting model is established. Compared with BP model, GA-BP model and standard ELM model, it is further verified that the predicting accuracy and running time of the proposed model are better.","PeriodicalId":254342,"journal":{"name":"2017 International Conference on Computing Intelligence and Information System (CIIS)","volume":"183 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131468623","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}
With the development of information technology and the improvement on the quality of network services, it is possible to achieve the internet-based project process information management. To reduce the degree of coupling between the model, view and control in the system, a management information system for research project process based on the front-end and back-end separation was proposed in this study. By guaranteeing the correct and orderly processing of research project, the system could realize the hierarchical management and monitoring of management information system for research project process. The system architecture was based on JavaEE technology, which could provide the full and convenient data supports in the process of system operation and ensure the precise and standard operation of research project. The system was mainly designed for the managerial personnel of research project, project evaluation expert, department of research project management and undertaker of research project, covering the all-round management of establishment, schedule, conclusion and evaluation. Meanwhile, relying on the integrated management of research project, the science authorities could reasonably allocate the resources for the research project, guarantee the proper implementation of research project and track the achievements, and thus promote the scientific research efficiency and management effectiveness of research institutions.
{"title":"Design and Development of Management Information System for Research Project Process Based on Front-End and Back-End Separation","authors":"Kun Liu, Jinmin Jiang, Xiaohan Ding, Hui Sun","doi":"10.1109/CIIS.2017.55","DOIUrl":"https://doi.org/10.1109/CIIS.2017.55","url":null,"abstract":"With the development of information technology and the improvement on the quality of network services, it is possible to achieve the internet-based project process information management. To reduce the degree of coupling between the model, view and control in the system, a management information system for research project process based on the front-end and back-end separation was proposed in this study. By guaranteeing the correct and orderly processing of research project, the system could realize the hierarchical management and monitoring of management information system for research project process. The system architecture was based on JavaEE technology, which could provide the full and convenient data supports in the process of system operation and ensure the precise and standard operation of research project. The system was mainly designed for the managerial personnel of research project, project evaluation expert, department of research project management and undertaker of research project, covering the all-round management of establishment, schedule, conclusion and evaluation. Meanwhile, relying on the integrated management of research project, the science authorities could reasonably allocate the resources for the research project, guarantee the proper implementation of research project and track the achievements, and thus promote the scientific research efficiency and management effectiveness of research institutions.","PeriodicalId":254342,"journal":{"name":"2017 International Conference on Computing Intelligence and Information System (CIIS)","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127863399","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}
With the development of artificial intelligence algorithm, the combination of intelligent algorithm and directed graph has become an important tool of current path planning. The application of the intelligent algorithm affects the optimal path planning in the directed graph, including the length of the optimal path and the time of the operation of the algorithm. The following studies are carried out through this paper based on the application of intelligent algorithm in directed graph. In the experimental environment of MATLAB, the coordinates of 30 target points are generated by random numbers. The ant colony algorithm and genetic algorithm are used to make the optimal path planning for the 30 target points, and the starting point and the end point are fixed to form a directed and closed graph. The parameters of the two algorithms are adjusted accordingly. Comparison of the two algorithms of the optimal path diagram and the algorithm running time, so as to draw the conclusion of the optimal algorithm.
{"title":"Research on Application of Artificial Intelligence Algorithm in Directed Graph","authors":"Yuanbo Zhou, Fangqin Xu","doi":"10.1109/CIIS.2017.26","DOIUrl":"https://doi.org/10.1109/CIIS.2017.26","url":null,"abstract":"With the development of artificial intelligence algorithm, the combination of intelligent algorithm and directed graph has become an important tool of current path planning. The application of the intelligent algorithm affects the optimal path planning in the directed graph, including the length of the optimal path and the time of the operation of the algorithm. The following studies are carried out through this paper based on the application of intelligent algorithm in directed graph. In the experimental environment of MATLAB, the coordinates of 30 target points are generated by random numbers. The ant colony algorithm and genetic algorithm are used to make the optimal path planning for the 30 target points, and the starting point and the end point are fixed to form a directed and closed graph. The parameters of the two algorithms are adjusted accordingly. Comparison of the two algorithms of the optimal path diagram and the algorithm running time, so as to draw the conclusion of the optimal algorithm.","PeriodicalId":254342,"journal":{"name":"2017 International Conference on Computing Intelligence and Information System (CIIS)","volume":"21 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122863818","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}
Community detection is of great importance in the study of complex networks, which motivates a body of new work in this domain. However, almost all networks change over time; traditional methods for static networks are not able to track evolutionary behaviors in temporal networks. To address this problem, we present a novel dynamic community detection model ENMF using nonnegative matrix factorization (NMF), which can not only track the temporal evolutions but also maintain the quality of detecting communities. Specifically, we propose gradient descent algorithm to optimize object function and evaluate the performance of the algorithm on one synthetic datasets. The results show that our proposed model outperforms other NMF methods.
{"title":"Dynamic Community Detection Using Nonnegative Matrix Factorization","authors":"Feng Gao, Limengzi Yuan, Wenjun Wang, Huandong Chang","doi":"10.1109/CIIS.2017.56","DOIUrl":"https://doi.org/10.1109/CIIS.2017.56","url":null,"abstract":"Community detection is of great importance in the study of complex networks, which motivates a body of new work in this domain. However, almost all networks change over time; traditional methods for static networks are not able to track evolutionary behaviors in temporal networks. To address this problem, we present a novel dynamic community detection model ENMF using nonnegative matrix factorization (NMF), which can not only track the temporal evolutions but also maintain the quality of detecting communities. Specifically, we propose gradient descent algorithm to optimize object function and evaluate the performance of the algorithm on one synthetic datasets. The results show that our proposed model outperforms other NMF methods.","PeriodicalId":254342,"journal":{"name":"2017 International Conference on Computing Intelligence and Information System (CIIS)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129211345","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}
Xiaoying Wang, Chengshui Niu, Yu-an Zhang, Lei Zhang
Focusing on the higher ratio of processor utilization and lower execution cost of a scientific workflow in the cloud environment, an improved list scheduling algorithm was proposed in this paper. This algorithm combines the ideas of both list scheduling and task duplication. According to the priority of the tasks, choosing reasonable parent task to replicate can help reduce the overhead between tasks. To properly insert tasks during processor idling time can help to increase the processor utilization. Based on these, we proposed an improved strategy to generate the workflow execution plan, called EPGILS. Experiment results show that the algorithm is feasible and efficient in reducing the task completion time and improving the utilization ratio of the processor.
{"title":"Workflow Execution Plan Generation in the Cloud Computing Environment Based on an Improved List Scheduling Algorithm","authors":"Xiaoying Wang, Chengshui Niu, Yu-an Zhang, Lei Zhang","doi":"10.1109/CIIS.2017.59","DOIUrl":"https://doi.org/10.1109/CIIS.2017.59","url":null,"abstract":"Focusing on the higher ratio of processor utilization and lower execution cost of a scientific workflow in the cloud environment, an improved list scheduling algorithm was proposed in this paper. This algorithm combines the ideas of both list scheduling and task duplication. According to the priority of the tasks, choosing reasonable parent task to replicate can help reduce the overhead between tasks. To properly insert tasks during processor idling time can help to increase the processor utilization. Based on these, we proposed an improved strategy to generate the workflow execution plan, called EPGILS. Experiment results show that the algorithm is feasible and efficient in reducing the task completion time and improving the utilization ratio of the processor.","PeriodicalId":254342,"journal":{"name":"2017 International Conference on Computing Intelligence and Information System (CIIS)","volume":"59 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129337780","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}
Infrared ship detection aiming at remote-sensing image is important in image processing to get priority in current war today. A background suppression method based on improved Top-Hat filtering and saliency map is presented for the detection. Firstly, Top-Hat filtering with linear combination of the open and close operators is utilized on the infrared remote-sensing image input. The filtering employs different structure operators with the same shape as the object image. Then the architecture of the Itti-Koch saliency-map model is utilized on the grayscale infrared image to intensify the interesting objects by visual effect principle. The results show that the improved background suppression method proposed can project the salient domain and detect more possible ship targets cleary.
{"title":"Background Suppression Based on Improved Top-Hat and Saliency Map Filtering for Infrared Ship Detection","authors":"Baorong Xie, Lingna Hu, Wentao Mu","doi":"10.1109/CIIS.2017.71","DOIUrl":"https://doi.org/10.1109/CIIS.2017.71","url":null,"abstract":"Infrared ship detection aiming at remote-sensing image is important in image processing to get priority in current war today. A background suppression method based on improved Top-Hat filtering and saliency map is presented for the detection. Firstly, Top-Hat filtering with linear combination of the open and close operators is utilized on the infrared remote-sensing image input. The filtering employs different structure operators with the same shape as the object image. Then the architecture of the Itti-Koch saliency-map model is utilized on the grayscale infrared image to intensify the interesting objects by visual effect principle. The results show that the improved background suppression method proposed can project the salient domain and detect more possible ship targets cleary.","PeriodicalId":254342,"journal":{"name":"2017 International Conference on Computing Intelligence and Information System (CIIS)","volume":"39 3 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128668991","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}
Population is the major strategy of long-term development, and it is great challenge as well. Population data is an important part of the national resources informatization. After analyzing in present situation of population information system and existing system impefection in collaboration, data exchange and so on, a new integrated population information sharing platform were built to realize resources sharing, complementary advantages and integrated application from aspects of establishing data standard, synchronization of heterogeneous database access, distributed business data integration, and performance of running system etc.
{"title":"Study on Population Comprehensive Information Sharing Platform","authors":"J. Zeng, X. Liang","doi":"10.1109/CIIS.2017.58","DOIUrl":"https://doi.org/10.1109/CIIS.2017.58","url":null,"abstract":"Population is the major strategy of long-term development, and it is great challenge as well. Population data is an important part of the national resources informatization. After analyzing in present situation of population information system and existing system impefection in collaboration, data exchange and so on, a new integrated population information sharing platform were built to realize resources sharing, complementary advantages and integrated application from aspects of establishing data standard, synchronization of heterogeneous database access, distributed business data integration, and performance of running system etc.","PeriodicalId":254342,"journal":{"name":"2017 International Conference on Computing Intelligence and Information System (CIIS)","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127213623","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}
To solve the binary classification transfer learning problem with similar data distributions and class imbalance between positive and negative samples in the target and source domains, we present an integrated transfer learning algorithm for multi-source unbalanced samples classification. We try to avoid the negative transfer problem by utilizing multiple source domains, and propose the new sample weights initialization and weights updating strategies to solve the class imbalance problem. Moreover, we propose a new elimination mechanism to eliminate the redundant samples in the multiple source domains, and then the time and memory costs of the classifier could be significantly reduced. Experimental results on standard UCI datasets show that the proposed algorithm outperforms the state-of-the-arts transfer learning algorithms in terms of F1-measure and AUC evaluations metrics.
{"title":"Integrated Transfer Learning Algorithm Using Multi-source TrAdaBoost for Unbalanced Samples Classification","authors":"Zhixiang Yuan, Damang Bao, Zekai Chen, Ming Liu","doi":"10.1109/CIIS.2017.37","DOIUrl":"https://doi.org/10.1109/CIIS.2017.37","url":null,"abstract":"To solve the binary classification transfer learning problem with similar data distributions and class imbalance between positive and negative samples in the target and source domains, we present an integrated transfer learning algorithm for multi-source unbalanced samples classification. We try to avoid the negative transfer problem by utilizing multiple source domains, and propose the new sample weights initialization and weights updating strategies to solve the class imbalance problem. Moreover, we propose a new elimination mechanism to eliminate the redundant samples in the multiple source domains, and then the time and memory costs of the classifier could be significantly reduced. Experimental results on standard UCI datasets show that the proposed algorithm outperforms the state-of-the-arts transfer learning algorithms in terms of F1-measure and AUC evaluations metrics.","PeriodicalId":254342,"journal":{"name":"2017 International Conference on Computing Intelligence and Information System (CIIS)","volume":"34 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130817701","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}
A novel image cryptosystem based on public image and chaotic systems is proposed in this paper. In proposed system, with the help of piecewise linear map and Chen's chaotic system, a public key is used to generate a public image, and a secret key is used to generate three key streams for image encryption. Then, the public image and one key stream are used for covering image. The other two key streams are used for image diffusion. The image encryption system includes two covering modules, two plaintext-unrelated diffusion modules and one plaintext-related confusion module. With the use of public image and plain image related confusion operation, the proposed system can resist the chosen/known plaintext attacks. Each encryption process uses a new public key, then the public key and cipher image are transmitted to the receiver through the public information channel. Even for the same plain image, each encryption process will produce totally different cipher images. When the public key is authenticated, the image cryptosystem can prevent active attacks. The simulation results show that the proposed scheme possesses the merits of fast encryption/decryption speed and high information security, and can be used to protect the image information on the internet.
{"title":"A Fast Image Encryption Scheme Based on Public Image and Chaos","authors":"Yong Zhang, Qiong Zhang, Hancheng Liao, Wenhua Wu, Xueqian Li, Hui-Chong Niu","doi":"10.1109/CIIS.2017.69","DOIUrl":"https://doi.org/10.1109/CIIS.2017.69","url":null,"abstract":"A novel image cryptosystem based on public image and chaotic systems is proposed in this paper. In proposed system, with the help of piecewise linear map and Chen's chaotic system, a public key is used to generate a public image, and a secret key is used to generate three key streams for image encryption. Then, the public image and one key stream are used for covering image. The other two key streams are used for image diffusion. The image encryption system includes two covering modules, two plaintext-unrelated diffusion modules and one plaintext-related confusion module. With the use of public image and plain image related confusion operation, the proposed system can resist the chosen/known plaintext attacks. Each encryption process uses a new public key, then the public key and cipher image are transmitted to the receiver through the public information channel. Even for the same plain image, each encryption process will produce totally different cipher images. When the public key is authenticated, the image cryptosystem can prevent active attacks. The simulation results show that the proposed scheme possesses the merits of fast encryption/decryption speed and high information security, and can be used to protect the image information on the internet.","PeriodicalId":254342,"journal":{"name":"2017 International Conference on Computing Intelligence and Information System (CIIS)","volume":"5 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128312005","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：