Pub Date : 2018-06-01DOI: 10.1109/ICIVC.2018.8492751
Xiaolin Zhao, Shilin Zhou, Lin Lei, Zhipeng Deng
In Unmanned Aerial Vehicle (UAV) videos, object tracking remains a challenge, due to its low spatial resolution and poor real-time performance. Recently, methods of deep learning have made great progress in object tracking in computer vision, especially fully-convolutional siamese neural networks (SiamFC). Inspired by it, this paper aims to investigate the use of SiamFC for object tracking in UAV videos. The network is trained on part of a UAV123 dataset and Stanford Drone dataset. First, exemplar image is extracted from the first frame and search regions are extracted in the following frames. Then, a Siamese network is used for tracking objects by calculating the similarity between exemplar image and search region. To evaluate our method, we test on a challenge VIVID dataset. The experiment shows that the proposed method has improvements in accuracy and speed in low spatial resolution UAV videos compared to existing methods.
{"title":"Siamese Network for Object Tracking in Aerial Video","authors":"Xiaolin Zhao, Shilin Zhou, Lin Lei, Zhipeng Deng","doi":"10.1109/ICIVC.2018.8492751","DOIUrl":"https://doi.org/10.1109/ICIVC.2018.8492751","url":null,"abstract":"In Unmanned Aerial Vehicle (UAV) videos, object tracking remains a challenge, due to its low spatial resolution and poor real-time performance. Recently, methods of deep learning have made great progress in object tracking in computer vision, especially fully-convolutional siamese neural networks (SiamFC). Inspired by it, this paper aims to investigate the use of SiamFC for object tracking in UAV videos. The network is trained on part of a UAV123 dataset and Stanford Drone dataset. First, exemplar image is extracted from the first frame and search regions are extracted in the following frames. Then, a Siamese network is used for tracking objects by calculating the similarity between exemplar image and search region. To evaluate our method, we test on a challenge VIVID dataset. The experiment shows that the proposed method has improvements in accuracy and speed in low spatial resolution UAV videos compared to existing methods.","PeriodicalId":173981,"journal":{"name":"2018 IEEE 3rd International Conference on Image, Vision and Computing (ICIVC)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128910828","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 : 2018-06-01DOI: 10.1109/ICIVC.2018.8492874
Feng Huahui, Zhang Geng, Zhang Xin, Hu Bingliang
Focusing on the problem existing in stereo matching that low-SNR image, such as images collected at night, we propose a novel matching framework based on semi-global matching algorithm and AD-Census. This algorithm extends the original algorithms in two ways. First, image segmentation information as an additional constraint is added that solve the problem of incomplete path and improve the accuracy of cost calculation. Second, the matching cost volume is calculated with AD-SoftCensus measure that minimizes the impact of noise on the quality of matching by changing the pattern of census descriptor from binary to trinary. Results of Middlebury standard test data show that the algorithm significantly improves the precision of matching. In addition, a low-light binocular platform is built to test our method in night environment. Results show the disparity maps are more accurate compared to previous methods.
{"title":"A Noise-Resistant Stereo Matching Algorithm Integrating Regional Information","authors":"Feng Huahui, Zhang Geng, Zhang Xin, Hu Bingliang","doi":"10.1109/ICIVC.2018.8492874","DOIUrl":"https://doi.org/10.1109/ICIVC.2018.8492874","url":null,"abstract":"Focusing on the problem existing in stereo matching that low-SNR image, such as images collected at night, we propose a novel matching framework based on semi-global matching algorithm and AD-Census. This algorithm extends the original algorithms in two ways. First, image segmentation information as an additional constraint is added that solve the problem of incomplete path and improve the accuracy of cost calculation. Second, the matching cost volume is calculated with AD-SoftCensus measure that minimizes the impact of noise on the quality of matching by changing the pattern of census descriptor from binary to trinary. Results of Middlebury standard test data show that the algorithm significantly improves the precision of matching. In addition, a low-light binocular platform is built to test our method in night environment. Results show the disparity maps are more accurate compared to previous methods.","PeriodicalId":173981,"journal":{"name":"2018 IEEE 3rd International Conference on Image, Vision and Computing (ICIVC)","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129893415","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 : 2018-06-01DOI: 10.1109/ICIVC.2018.8492728
Xuebo Zhang, Cheng Tan, Wenwei Ying
The parameter initializations play an important role in the iteration of parameter estimation. Based on characteristic function, a parameter estimation method for Class B noise considering the parameter initialization is presented in this paper. The noise is firstly considered as the symmetric alpha stable (SαS) distribution. With the log method, we get the estimated parameters, which are further used as the parameter initial values of iteration. It improves the convergence speed. The processing results of simulated data indicate that the parameters of Class B noise can be efficiently estimated with the presented method.
{"title":"Characteristic Function Based Parameter Estimation for Ocean Ambient Noise","authors":"Xuebo Zhang, Cheng Tan, Wenwei Ying","doi":"10.1109/ICIVC.2018.8492728","DOIUrl":"https://doi.org/10.1109/ICIVC.2018.8492728","url":null,"abstract":"The parameter initializations play an important role in the iteration of parameter estimation. Based on characteristic function, a parameter estimation method for Class B noise considering the parameter initialization is presented in this paper. The noise is firstly considered as the symmetric alpha stable (SαS) distribution. With the log method, we get the estimated parameters, which are further used as the parameter initial values of iteration. It improves the convergence speed. The processing results of simulated data indicate that the parameters of Class B noise can be efficiently estimated with the presented method.","PeriodicalId":173981,"journal":{"name":"2018 IEEE 3rd International Conference on Image, Vision and Computing (ICIVC)","volume":"65 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122775713","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}
In some high-level secure applications in need of multiple participants input their own secret data to achieve access control, such as, secure cabinet opened by multiple owners together, traditional security technology is not applicable. Although secret sharing may be used in the scenarios, there are some problems when directly applying primary secret sharing methods including visual cryptography (VC) and polynomial-based secret sharing. In this paper, we first describe the application scenario (namely secret data fusion) and its requirements, where secret data fusion is different from secret sharing. Then, we propose a possible method for secret data fusion based on Chinese remainder theorem (CRT). Theoretical analyses and experiments are examined to represent the effectiveness of our method.
{"title":"Secret Data Fusion Based on Chinese Remainder Theorem","authors":"Yuliang Lu, Xuehu Yan, Lintao Liu, Jingju Liu, Guozheng Yang, Qiang Li","doi":"10.1109/ICIVC.2018.8492875","DOIUrl":"https://doi.org/10.1109/ICIVC.2018.8492875","url":null,"abstract":"In some high-level secure applications in need of multiple participants input their own secret data to achieve access control, such as, secure cabinet opened by multiple owners together, traditional security technology is not applicable. Although secret sharing may be used in the scenarios, there are some problems when directly applying primary secret sharing methods including visual cryptography (VC) and polynomial-based secret sharing. In this paper, we first describe the application scenario (namely secret data fusion) and its requirements, where secret data fusion is different from secret sharing. Then, we propose a possible method for secret data fusion based on Chinese remainder theorem (CRT). Theoretical analyses and experiments are examined to represent the effectiveness of our method.","PeriodicalId":173981,"journal":{"name":"2018 IEEE 3rd International Conference on Image, Vision and Computing (ICIVC)","volume":"151 10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123568199","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 : 2018-06-01DOI: 10.1109/ICIVC.2018.8492772
Yiwei Liu, Peirui Bai, Chang Li, Yue Zhao
Level set models are widely used in the image segmentation field. However, the sensitivity of the initial contours and the manual adjustment of the controlling parameters have limited the segmentation performance. To effectively solve this problem, a novel level set model utilizing both intensity and spatial information is proposed in this paper. Firstly, the fuzzy connectedness (FC) algorithm is applied to obtain the appropriate initial contours, and as a result the complexity and computation cost of building initial contours is reduced. Secondly, based on the morphological characteristics of the initial contours and the parameters of fuzzy connectedness, several equations are proposed to automatically estimate the controlling parameters of the level set evolution and avoid human intervention. Finally, the region-scalable fitting (RSF) model is adopted to evolve and obtain the final robust segmentation results. The efficiency and accuracy of the model proposed in this paper is verified by comparing the three quantitative indexes of time, Dice similarity coefficient (DSC) and peak signal to noise ratio (PSNR) with four different initialized level set models.
{"title":"A Novel Level Set Model Originated from Fuzzy Connectedness Guided Initial Contours","authors":"Yiwei Liu, Peirui Bai, Chang Li, Yue Zhao","doi":"10.1109/ICIVC.2018.8492772","DOIUrl":"https://doi.org/10.1109/ICIVC.2018.8492772","url":null,"abstract":"Level set models are widely used in the image segmentation field. However, the sensitivity of the initial contours and the manual adjustment of the controlling parameters have limited the segmentation performance. To effectively solve this problem, a novel level set model utilizing both intensity and spatial information is proposed in this paper. Firstly, the fuzzy connectedness (FC) algorithm is applied to obtain the appropriate initial contours, and as a result the complexity and computation cost of building initial contours is reduced. Secondly, based on the morphological characteristics of the initial contours and the parameters of fuzzy connectedness, several equations are proposed to automatically estimate the controlling parameters of the level set evolution and avoid human intervention. Finally, the region-scalable fitting (RSF) model is adopted to evolve and obtain the final robust segmentation results. The efficiency and accuracy of the model proposed in this paper is verified by comparing the three quantitative indexes of time, Dice similarity coefficient (DSC) and peak signal to noise ratio (PSNR) with four different initialized level set models.","PeriodicalId":173981,"journal":{"name":"2018 IEEE 3rd International Conference on Image, Vision and Computing (ICIVC)","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133130180","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 : 2018-06-01DOI: 10.1109/ICIVC.2018.8492846
Xianguang Lu, Xuehui Du, Wenjuan Wang
Intrusion detection system is an effective defense tool for finding security events. However, it will produce a large number of false positive alerts, which greatly increases the difficulty of real-time security analysis for the security managers, in actual applications. The periodic alarm produced by the wrong configuration of network devices and services, and the approximately duplicate alarm generated by different IDS for the same attack are important components of false alarm. In this paper, we improved the SNM algorithm and cleaned up the duplicate alarm in the original alarm database, which reduced the scale of the database; On the other hand, we have made statistics on the number of duplicate alarms, so that we can further find periodic alerts and remove false alarms.
{"title":"Network IDS Duplicate Alarm Reduction Using Improved SNM Algorithm","authors":"Xianguang Lu, Xuehui Du, Wenjuan Wang","doi":"10.1109/ICIVC.2018.8492846","DOIUrl":"https://doi.org/10.1109/ICIVC.2018.8492846","url":null,"abstract":"Intrusion detection system is an effective defense tool for finding security events. However, it will produce a large number of false positive alerts, which greatly increases the difficulty of real-time security analysis for the security managers, in actual applications. The periodic alarm produced by the wrong configuration of network devices and services, and the approximately duplicate alarm generated by different IDS for the same attack are important components of false alarm. In this paper, we improved the SNM algorithm and cleaned up the duplicate alarm in the original alarm database, which reduced the scale of the database; On the other hand, we have made statistics on the number of duplicate alarms, so that we can further find periodic alerts and remove false alarms.","PeriodicalId":173981,"journal":{"name":"2018 IEEE 3rd International Conference on Image, Vision and Computing (ICIVC)","volume":"44 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122421971","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 : 2018-06-01DOI: 10.1109/ICIVC.2018.8492863
Xiaofeng Yan, Jie Zhao
Prediction is a common method in data mining. In the prediction method, it can be divided into linear prediction and nonlinear prediction. The multiple linear regression method belongs to the linear regression method, and the neural network algorithm belongs to nonlinear prediction. The neural network algorithm belongs to the computational intelligence algorithm. It depends on the complexity of the system and connects the relations between the internal nodes of the neural network through the weights to process the data information. Based on multiple linear regression and neural network algorithms, this paper proposes a predictive model based on multiple linear regression and neural network, and uses this model to study national economic data. The prediction model proposed in this paper is realized by using the linear prediction result as the input neuron of the neural network. The neural network used in this paper is a radial basis function neural network, hereinafter referred to as RBF neural network.
{"title":"Application of Neural Network in National Economic Forecast","authors":"Xiaofeng Yan, Jie Zhao","doi":"10.1109/ICIVC.2018.8492863","DOIUrl":"https://doi.org/10.1109/ICIVC.2018.8492863","url":null,"abstract":"Prediction is a common method in data mining. In the prediction method, it can be divided into linear prediction and nonlinear prediction. The multiple linear regression method belongs to the linear regression method, and the neural network algorithm belongs to nonlinear prediction. The neural network algorithm belongs to the computational intelligence algorithm. It depends on the complexity of the system and connects the relations between the internal nodes of the neural network through the weights to process the data information. Based on multiple linear regression and neural network algorithms, this paper proposes a predictive model based on multiple linear regression and neural network, and uses this model to study national economic data. The prediction model proposed in this paper is realized by using the linear prediction result as the input neuron of the neural network. The neural network used in this paper is a radial basis function neural network, hereinafter referred to as RBF neural network.","PeriodicalId":173981,"journal":{"name":"2018 IEEE 3rd International Conference on Image, Vision and Computing (ICIVC)","volume":"41 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115872818","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 : 2018-06-01DOI: 10.1109/ICIVC.2018.8492818
Yubing Li, Jinbo Zhang, Peng Gao, Liangcheng Jiang, Ming Chen
Grab Cut algorithm is one of the most popular method in the field of image segmentation. It uses texture information and boundary information of image, and achieves good segmentation results with a small number of user interaction. But there are two significant drawbacks about this algorithm. Firstly, If the background is complex or the background and the object are very similar, the segmentation will not be very good. On the other hand, the relatively slow speed and Complex iterative process of the algorithm are greatly limited its application. In this paper, to develop these aspects, we proposed an improved grab cut algorithm. This algorithm is the combination of grab cut and graph-based image segmentation [1]. After the experiment, the improved algorithm is applied to more complex situation.
{"title":"Grab Cut Image Segmentation Based on Image Region","authors":"Yubing Li, Jinbo Zhang, Peng Gao, Liangcheng Jiang, Ming Chen","doi":"10.1109/ICIVC.2018.8492818","DOIUrl":"https://doi.org/10.1109/ICIVC.2018.8492818","url":null,"abstract":"Grab Cut algorithm is one of the most popular method in the field of image segmentation. It uses texture information and boundary information of image, and achieves good segmentation results with a small number of user interaction. But there are two significant drawbacks about this algorithm. Firstly, If the background is complex or the background and the object are very similar, the segmentation will not be very good. On the other hand, the relatively slow speed and Complex iterative process of the algorithm are greatly limited its application. In this paper, to develop these aspects, we proposed an improved grab cut algorithm. This algorithm is the combination of grab cut and graph-based image segmentation [1]. After the experiment, the improved algorithm is applied to more complex situation.","PeriodicalId":173981,"journal":{"name":"2018 IEEE 3rd International Conference on Image, Vision and Computing (ICIVC)","volume":"112 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117281269","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 : 2018-06-01DOI: 10.1109/ICIVC.2018.8492865
Zhiqiang Hu, Meiqi Hu
There are some shortcomings for the cache sensitivity of the index in main memory database, so a new index structure is proposed. T -tree index is studied individually ever before, so as Hash index. Combined with the analysis of the two index structure, a new index structure called the T-Hash tree is introduced. Through analyzing the times of the T -Hash tree cache sensitive and testing the performance of the query, insert, delete operation, the results show that the T -Hash tree can effectively reduce the times of cache sensitive, and as the amount of the data is large, the query, insert, delete efficiency of the T -Hash tree is higher than the T tree.
{"title":"Design and Implementation of T-Hash Tree in Main Memory Data Base","authors":"Zhiqiang Hu, Meiqi Hu","doi":"10.1109/ICIVC.2018.8492865","DOIUrl":"https://doi.org/10.1109/ICIVC.2018.8492865","url":null,"abstract":"There are some shortcomings for the cache sensitivity of the index in main memory database, so a new index structure is proposed. T -tree index is studied individually ever before, so as Hash index. Combined with the analysis of the two index structure, a new index structure called the T-Hash tree is introduced. Through analyzing the times of the T -Hash tree cache sensitive and testing the performance of the query, insert, delete operation, the results show that the T -Hash tree can effectively reduce the times of cache sensitive, and as the amount of the data is large, the query, insert, delete efficiency of the T -Hash tree is higher than the T tree.","PeriodicalId":173981,"journal":{"name":"2018 IEEE 3rd International Conference on Image, Vision and Computing (ICIVC)","volume":"32 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114142718","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 : 2018-06-01DOI: 10.1109/ICIVC.2018.8492830
Qirong Bo, Jun Feng, P. Li, Zhaohui Lv, Jing Zhang
According to gestalt psychology theory, the human brain merges and simplifies unrelated units by some relations through eyes for subsequent cognition. We introduce a new segmentation framework based on gestalt psychology in this paper. An input image is first divided into visual patches using two gestalt principles, similarity and proximity, by a clustering method, and then the visual patches are grouped to form soft tissues by a classification step using the spatial relationship and texture features. We evaluated the proposed method using TCIA database at both sectional level and volumetric level. The experimental results demonstrated the efficiency and robustness of the presented method and indicated its promising applications in the field of medical image processing.
{"title":"Towards Better Soft-Tissue Segmentation Based on Gestalt Psychology","authors":"Qirong Bo, Jun Feng, P. Li, Zhaohui Lv, Jing Zhang","doi":"10.1109/ICIVC.2018.8492830","DOIUrl":"https://doi.org/10.1109/ICIVC.2018.8492830","url":null,"abstract":"According to gestalt psychology theory, the human brain merges and simplifies unrelated units by some relations through eyes for subsequent cognition. We introduce a new segmentation framework based on gestalt psychology in this paper. An input image is first divided into visual patches using two gestalt principles, similarity and proximity, by a clustering method, and then the visual patches are grouped to form soft tissues by a classification step using the spatial relationship and texture features. We evaluated the proposed method using TCIA database at both sectional level and volumetric level. The experimental results demonstrated the efficiency and robustness of the presented method and indicated its promising applications in the field of medical image processing.","PeriodicalId":173981,"journal":{"name":"2018 IEEE 3rd International Conference on Image, Vision and Computing (ICIVC)","volume":"26 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115294874","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：