Papers - THI THI ZIN
-
An innovative background model based on multiple queuing framework Reviewed
Thi Thi Zin,Pyke Tin, T. Toriu, H. Hama
ICIC Express Letters 6 ( 4 ) 1039 - 1044 2012.4
Language：English Publishing type：Research paper (scientific journal) Publisher：ICIC Express Letters
Many computer vision applications, especially video surveillance systems, highly depend on background model and foreground segmentation. In this paper, we propose an innovative background subtraction method based on a multiple queuing framework. Under this framework, both the static and dynamic background pixels are investigated by using a novel hypothesis method to establish an active real time background modeling in presence of moving foreground objects in the complex scene and adaptation of background model to gradual and sudden "once-off" background changes. Experiments were conducted by using public datasets PETS2006 and our own video sequences taken at an international airport and a university campus.
-
A general framework for knowledge based human behavior understanding Reviewed
Pyke Tin, Thi Thi Zin, T. Toriu, H. Hama
ICIC Express Letters 6 ( 4 ) 899 - 904 2012.4
Language：English Publishing type：Research paper (scientific journal) Publisher：ICIC Express Letters
Human behavior understanding systems today are mainly knowledge based. And the diffusion of human actions and interactions knowledge is imperative for proper and prompt decision making processes in human behavior related crime prevention, care support systems, intelligent surveillance systems, etc. In this paper, we propose a general framework for knowledge based human behavior understanding system comprising three components: (i) knowledge acquisition, (ii) knowledge representation and modeling, and (in) knowledge base and use of knowledge. Specifically, human behavior is modeled as a Stochastic sequence of low level actions. The results of the Low Level Activities (LLA) analysis will subsequently be fed into a Knowledge Base System (KBS) that is used as High Level Activities (HLA) model. As an application, we will focus on the detection of theft and robbery related events. Unlike the traditional approach to just detecting stationary and moving objects in monitored scenes, our KBS approach detects the events based on accumulated knowledge about human and non-human objects from continuous object classification. ICIC International © 2012.
-
A probability-based model for detecting abandoned objects in video surveillance systems
Zin T.
Lecture Notes in Engineering and Computer Science 2198 1246 - 1251 2012
Language：English Publishing type：Research paper (scientific journal) Publisher：Lecture Notes in Engineering and Computer Science
© 2012 Newswood Limited. All rights reserved. Detection of suspicious packages or abandoned objects is one of the most important tasks in video surveillance systems. Some recent terrorist attacks involving explosive packages left behind in many contexts such as airports, rail stations and etc. illustrate the importance of this problem. In this paper, we propose a probability-based model for robustly and efficiently detecting abandoned objects in complex environments. Specifically, we develop a new probability-based background subtraction algorithm based on combination of multiple background models for motion detection. In addition, several improvements are implemented to the background subtraction method for shadow removal and quick lighting change adaptation. We then analyze the extracted objects to classify as static or dynamic objects. After the analysis, we employ the statistical running average of the static foreground masks for event type decision making either abandoned or very still person. Finally, the robustness and efficiency of the method are tested on our video sequences and PETS2006 datasets.
-
Simultaneous visual ranking and clustering using weighted multiple features
Geng T., Zin T., Tin P., Toriu T., Hama H.
ICIC Express Letters 5 ( 10 ) 3773 - 3778 2011.10
Language：Japanese Publishing type：Research paper (scientific journal) Publisher：ICIC Express Letters
In this paper, we present a new approach to visual ranking and clustering system by jointly exploring the multiple visual features information using adaptive visual similarity. We investigate how to effectively incorporate both intra-image and inter-image spatial structure information into Markov Stationary Features derived from the normalized co-occurrence matrix. The convex quadratic programming algorithm is developed to learn the weights for fusing the rankings from multiple features using color and texture. As a result, the multiple visual features based re-ranking can take more reliable information from each other. Experimental results on a real-world datasets collected from various image search engines show that our method outperforms several existing approaches which do not or weakly consider multi-feature interactions. © 2011 ISSN 1881-803X.
-
Zin T., Tin P., Hama H., Nakajima S., Toriu T.
ICIC Express Letters 5 ( 10 ) 3767 - 3772 2011.10
Language：Japanese Publishing type：Research paper (scientific journal) Publisher：ICIC Express Letters
In this paper, we present a robust and effective multiple layers stochastic background modeling and novel stationary object detection method, which comprise of probability based filtering operations to detect stationary objects in a monitoring scene. Conventionally, a statistical background model is extracted by using a training sequence without foreground objects. Our method does not require starting with a period of empty scenes to facilitate the original background. In our proposed algorithm, three layers background model is constructed by using periodic Markov chain concepts. We then apply background subtractions to the current frame for objects detection and classification. Extensive experimental work has been done, results of which show that the present approach provides a better solution compared with the conventional approach, including the problem of re-active objects in real world complex environments. © 2011 ISSN 1881-803X.
-
Challenges and promises in human behavior understanding research
Tin P., Zin T., Hama H., Toriu T.
ICIC Express Letters 5 ( 10 ) 3761 - 3766 2011.10
Language：Japanese Publishing type：Research paper (scientific journal) Publisher：ICIC Express Letters
Visual surveillance for human behavior understanding is an active research topic in image processing. Behavior analysis in dynamic scenes is a complex task, especially when it concerns human populated environments. Study to emulate the astonishing performances of such a perfect system as the natural and computer vision system represents, without any doubt, a real challenge. It has a wide spectrum of promising applications, including access control in special areas, human identification at a distance, crowd flux statistics and congestion analysis, detection of anomalous behaviors, and interactive surveillance using multiple cameras, etc. In this paper, the current state-of-the-art image processing methods for automatic behavior recognition techniques are described, with a focus on the surveillance of human activities in the context of transit applications. The main purpose of this paper is to provide researchers in the field with a summary of progress achieved to date and to help identify areas where further research is needed. This paper presents a comprehensive study of the research on relevant human behavior understanding methods for public safety and security surveillance. A classification table of research papers on relevant behavior analysis is presented, including behaviors, datasets, implementation details, and results. © 2011 ISSN 1881-803X.
-
The method of area segmentation of cancer using multispectral image
Takahashi H., Yamada K., Tareda M., Yoshida S., Zin T., Aida T.
ICIC Express Letters 5 ( 10 ) 3859 - 3864 2011.10
Language：Japanese Publishing type：Research paper (scientific journal) Publisher：ICIC Express Letters
Transnasal endoscopy has recently become a widely accepted screening method in practical application areas of medical endoscopy. Pain tolerance and safety seem to be the greatest advantage of transnasal endoscopy when compared with those of a conventional peroral endoscopy. Although there are some disadvantages such as poor endoscopic images and less illumination caused by downsizing of scopes, as a whole, slim endoscopy is a safe and less invasive tool for screening purposes. On the other hand, flexible spectral imaging color enhancement (FICE) is one of the diagnostic methods using specific light spectra based on spectral image processing technology. FICE provides comparison of spectral images of diseased and surrounding normal areas for enhancement of the contrast by combining wavelengths with greater differences in signals. It is thus in this paper, that we propose a novel method of segmenting tumor region for medical endoscope surgery and investigation by using the narrowband images on several wavelengths. © 2011 ISSN 1881-803X.
-
Small sample learning for motion estimation based on self-organized internal representation
Toriu T., Zin T., Hama H.
ICIC Express Letters 5 ( 10 ) 3921 - 3926 2011.10
Language：Japanese Publishing type：Research paper (scientific journal) Publisher：ICIC Express Letters
In the previous paper, we proposed an unsupervised learning algorithm for self-organizing an internal representation of ego-motion. We showed that motion param-eters could be topographically mapped onto a robot's internal parameter space sponta-neously. In this paper, first, we provide the theoretical rationale why the previous method can self-organize the internal representation of ego-motion. Then, on the basis of this theoretical foundation, we propose a novel learning method to estimate real motion parameters such as translation and rotation parameters. Only small samples of input and output data are needed to complete this learning. We show that this method works well by experiments using randomly synthesized image. © 2011 ISSN 1881-803X.
-
Pedestrian detection based on hybrid features using near infrared images
Zin T., Tin P., Hama H.
International Journal of Innovative Computing, Information and Control 7 ( 8 ) 5015 - 5025 2011.8
Language：Japanese Publishing type：Research paper (scientific journal) Publisher：International Journal of Innovative Computing, Information and Control
This paper explores a hybrid-based method to fuse multi-slit features and Histograms of Oriented Gradients (HOG) features for pedestrian detection from Near Infrared (NIR) images. The fused feature set utilizes both the multi-slit method's capability of accurately capturing the local spatial layout of body parts (head, torso and legs) in individual frames and the HOG's capability in region information relevant to higher frequency components. The hybrid feature vector describing various types of poses is then constructed and used for detecting the pedestrians. The part based pattern matching analysis indicates that the fused features have much higher feature space separation than the pure features. Experiments with a database of NIR images show that the proposed method achieves a substantial improvement in tackling some difficult cases such as side view, back view which the conventional HOG method cannot handle. Detection and recognition performance is less computationally expensive than existing approaches. © 2011 ICIC International.
-
Background modeling using special type of Markov Chain
Zin T., Tin P., Toriu T., Hama H.
IEICE Electronics Express 8 ( 13 ) 1082 - 1088 2011.8
Language：Japanese Publishing type：Research paper (scientific journal) Publisher：IEICE Electronics Express
Background modeling is important in video surveillance for extracting foreground regions from a complex environment. In this paper, we present a novel background modeling technique based on a special type of Markov Chain. The method is a substantial extension to the existing background subtraction techniques. First, a background pixel is statistically modeled by a linear regressive Gamma Markov distribution. Then, these statistical estimates are used as important parameters in background update schemes. The experimental results show that the proposed model is less sensitive to movements of the texture background and more robust for real time segmenting the foreground object accurately. © IEICE 2011.
DOI： 10.1587/elex.8.1082
-
A novel concept of morphology pivotal elements for object image retrieval
Hama H., Zin T., Tin P.
International Journal of Innovative Computing, Information and Control 7 ( 7 A ) 3891 - 3901 2011.7
Language：Japanese Publishing type：Research paper (scientific journal) Publisher：International Journal of Innovative Computing, Information and Control
In this paper, we introduce a novel and simple Pivotal Element (PE) concept in mathematical morphological operations for object image retrieval schemes based on combinations of empirical and statistical analyses. Mathematical morphology is very attractive for this purpose because it efficiently deals with geometrical features like size, shape, contrast or connectivity that can be considered as image retrieval oriented features. With an optimized structure, morphological dilation is more effective to detect object spot target in image sequences. Based on the real convex figure, morphological operation with circular structure is designed in this paper. The PE is introduced to optimize the noisy background elements. The empirical threshold is decided approximately based on the statistical characters. In this aspect, two approaches for solving morphological applications to image data distributed on the unit circle are presented. In the first approach, a framework for analyzing images, called pivotal role, has been developed based on a set of concentric circles with adjustable radii, with exactly one circle centered at each pivotal image pixel. The second approach is based on Markov decision processes which operate only on grouped data. The retrieval quality is improved by dynamically changing the combinatorial coefficients that are used in equations of optimality principles. by using it as a priori knowledge of the morphology operation, it does favor to improve the algorithm's accuracy and adaptability. The experiment shows that the new concept of PE has made the morphological operations to achieve a higher retrieval efficiency and accuracy. © 2011.
-
Fast multi-slit composition method using attitude measurement unit for pedestrian detection
Hashimoto Y., Zin T., Toriu T., Hama H.
ICIC Express Letters, Part B: Applications 2 ( 3 ) 541 - 546 2011.6
Language：Japanese Publishing type：Research paper (scientific journal) Publisher：ICIC Express Letters, Part B: Applications
During the next decade, on-board pedestrian detection systems will play a key role in the challenge of increasing traffic safety. The main goal of these systems, to detect pedestrians in urban scenarios, involves overcoming difficulties like processing outdoor scenes from a mobile platform and searching for aspect-changing objects in cluttered environments. Such systems thus combine various techniques in state-of-the-art computer vision. In this paper, we present a novel method that uses an Attitude Measurement Unit (AMU) in conjunction with a camera for tracking and which performs well in dynamic and cluttered outdoor environments as long as the target occlusions and losses are temporary. The main purpose is to demonstrate the advantages of using an AMU as a means of on-road pedestrian detection for automobiles over more conventional means. Specifically, we present a two module systems based on both 2D and 3D sensor cues. The first module uses fipitch and roll angles for calculating the vanishing line on the image to select a coherent set of regions of interest (ROIs) to be further analyzed. The second module develops a modified multi-slit method to classify the incoming ROIs into pedestrian and non-pedestrian in order to refine the final results. Our results indicate the integration of the proposed techniques gives rise to a promising system. © 2011 ICIC International.
-
Unsupervised learning algorithm for self-organizing internal representation of ego-motion Reviewed
T. Toriu, Thi Thi Zin, H. Hama
ICIC Express Letters, Part B: Applications 2 ( 3 ) 559 - 564 2011.6
Language：English Publishing type：Research paper (scientific journal) Publisher：ICIC Express Letters, Part B: Applications
In this paper, we propose an unsupervised learning algorithm for self-organi-zing an internal representation of ego-motion. In this method, the system self-organizes the internal space for representing ego-motion without employing a supervisor. We es-tablish in our experiments that objective ego-motion parameters can be topographically mapped onto a robot's internal parameter space spontaneously using this learning. Once the learning is complete, the system can recall the representation of ego-motion from the pair of input sensation and its time derivative. One important aspect of this method is that the system does not use any knowledge of geometrical nature during image gen-eration; therefore, it is not affected by any image distortion such as that induced in omnidirectional or fish eye cameras.
-
Manipulation-Killer web information ranking System
Tin P., Zin T., Toriu T., Hama H.
ICIC Express Letters, Part B: Applications 2 ( 3 ) 523 - 528 2011.6
Language：Japanese Publishing type：Research paper (scientific journal) Publisher：ICIC Express Letters, Part B: Applications
Link analysis has been widely used to evaluate the importance of web pages. Page-Rank, the most famous link analysis algorithm, offers an effective way to rank the pages. However, the algorithm ignores four facts. First, nowadays the way that users retrieve information is different from the previous way when web search engine was not extensively used. Second, inter-site links and intra-site links should not be treated equally. A link from a different site is more important for a page than that within the same site. Third, most users start their browsing from a homepage, which should be given more weight than other pages. Finally, most of existing ranking systems for linked networks can be manipulated by spammers who strategically place links. In this paper, we pro-pose a novel ranking system called manipulation-killer rank (MK-Rank) as a solution to these problems. Specifically, we show that expected passage time to reach a web page in a random walk measures essentially the same quantity as Page-Rank, whereas it does not depend on the manipulated in-links. We also show that it resists tampering by individuals or groups who strategically place manipulated links. In addition, we present an algorithm to efficiently compute passage time for all nodes in a massive graph; conventional algo-rithms do not scale adequately. Experimental results show that our MK-Rank algorithm outperforms other famous ranking algorithms, including Page-Rank and Traffic Rank, especially on sites recommendation and web spam avoidance. © 2011 ICIC International.
-
Unattended object intelligent analyzer for consumer video surveillance
Zin T., Tin P., Hama H., Toriu T.
IEEE Transactions on Consumer Electronics 57 ( 2 ) 549 - 557 2011.5
Language：Japanese Publishing type：Research paper (scientific journal) Publisher：IEEE Transactions on Consumer Electronics
Consumer video camera surveillance with the continuous advancements of image processing technologies is emerging for consumer world of applications. Technology for detecting objects left unattended in consumer world such as shopping malls, airports, railways stations has resulted in successful commercialization, worldwide sales and the winning of international awards. However, as a consumer video application the need is now greater than ever for a surveillance system that is robustly and effectively automated. In this paper, we propose an intelligent vision based analyzer for semantic analysis of objects left unattended relation with human behaviors from a monocular surveillance video, captured by a consumer camera through cluttered environments. Our analyzer employs visual cues to robustly and efficiently detect unattended objects which are usually considered as potential security breach in public safety from terrorist explosive attacks. The proposed system consists of three processing steps: (i) object extraction, involving a new background subtraction algorithm based on combination of periodic background models with shadow removal and quick lighting change adaptation,(ii) extracted objects classification as stationary or dynamic objects, and (iii) classified objects investigation by using running average about the static foreground masks to calculate a confidence score for the decision making about event (either unattended or very still person). We show attractive experimental results, highlighting the system efficiency and classification capability by using our real-time consumer video surveillance system for public safety application in big cities. © 2011 IEEE.
-
A Markov random walk model for loitering people detection
Zin T., Tin P., Toriu T., Hama H.
Proceedings - 2010 6th International Conference on Intelligent Information Hiding and Multimedia Signal Processing, IIHMSP 2010 680 - 683 2010.12
Language：Japanese Publishing type：Research paper (scientific journal) Publisher：Proceedings - 2010 6th International Conference on Intelligent Information Hiding and Multimedia Signal Processing, IIHMSP 2010
Today video surveillance systems are widely used in public spaces, such as train stations or airports, to enhance security. In order to observe large and complex facilities a huge amount of cameras is required. These create a massive amount of data to be analyzed. It is therefore crucial to support human security staff with automatic surveillance applications, which will create an alert if security relevant events are detected. This way video surveillance could be used to prevent potentially dangerous situations, instead of just being used as forensic instrument, to analyze an event after it happened. In this treatise we present a surveillance system which supports human operators, by automatically detecting loitering people. Usually, loitering human behavior often leads to abnormal situations, like suspected drug-dealing activity, bank robbery, and pickpocket, etc. Thus, the problem of loitering detection in image sequences involving situations with multiple objects is studied based two dimensional Markov random walks in which both motion and appearance features describing the movements of a varying number of objects as well as their entries and exits are used. To obtain efficient and compact representations we encode the spatiotemporal information of intra-inter trajectory contexts into the transition matrix of a Markov Random Walk, and then extract its stationary distribution and boundary crossing probabilities as final detection criteria. The model is also made less sensitive to uninteresting objects occluding the region of interest by integration out their effect on the observation probabilities. The resulting system is tested on the real life dataset scenarios giving 95% performance results. © 2010 IEEE.
-
A stochastic model for popularity measures in web dynamics Reviewed
H. Hama H., Pyke Tin, Thi Thi Zin, T. Toriu
Proceedings - 2010 6th International Conference on Intelligent Information Hiding and Multimedia Signal Processing, IIHMSP 2010 676 - 679 2010.12
Language：English Publishing type：Research paper (international conference proceedings) Publisher：Proceedings - 2010 6th International Conference on Intelligent Information Hiding and Multimedia Signal Processing, IIHMSP 2010
In this paper, we propose a stochastic web dynamic model based on the concept of queuing theory to measure popularity of websites in the World Wide Web. We assume that the characteristics of a website such as novelty, popularity, reliability, and relevancy are governed by two major forces: internal or self-growth of each website and external functions acting on the website. In stochastic language, these two forces can be considered as random variables and able to investigate the important characteristics of websites. These characteristics include a probability distribution of the number of visitors to websites, the visiting times distribution, users' attention, the functional growth and decay of individual websites, the relationship between the dynamics of the web and its structure. We then define and investigate measuring process of popularity of websites using stochastic difference equations based on the structure of the web taken users' attention into account. For validation, we present some simulation results with the respect to parameter variations in the model. It shows that the proposed model can efficiently and adequately analyze the characteristics and behaviors of websites in web dynamical system. © 2010 IEEE.
-
HOG embedded Markov chain model for pedestrian detection
Zin T., Hama H., Tin P., Toriu T.
ICIC Express Letters 4 ( 6 B ) 2463 - 2468 2010.12
Language：Japanese Publishing type：Research paper (scientific journal) Publisher：ICIC Express Letters
This paper presents a new method for pedestrian detection by establishing Histograms of Oriented Gradients (HOG) embedded Markov chain model based on the cooccurrence of dominant orientations of gradients. In this model, HOG are used to obtain dominant orientations and their co-occurrences are computed by using various positional Structuring Elements (SEs). We then embed the HOG pair co-occurrences into a Markov chain. Having defined the embedded Markov chain, the corresponding joint probability density functions (pdf) are derived and used as Markov similarity features for pedestrian detection process. Due to the use of various positional SEs, the derived features can express complex shapes of objects with local and global distributions of gradient orientations. Experimental results on common datasets and comparison with some previous methods are given. The results show that the performance of our method is significantly well and it outperforms some existing methods such as conventional co-occurrence and HOGs. ICIC International © 2010 ISSN 1881-803X.
-
Multivariate web information reliability search engine
Hama H., Tin P., Zin T., Toriu T.
ICIC Express Letters 4 ( 6 B ) 2457 - 2462 2010.12
Language：Japanese Publishing type：Research paper (scientific journal) Publisher：ICIC Express Letters
This paper proposes a novel multivariate analysis approach to web information reliability search systems. In this analysis, we introduce a new representation of the web as a directed stochastic hyper-graph, instead of a simple graph, where links can connect not only pairs of web pages, but also pairs of disjoint sets of pages. The stochastic hyper-graph is constructed by regrouping the set of pages into non-overlapping page-sets (subsets) and using the links between pages of distinct page-sets to create weighted hyperarcs with the goal of providing more reliable information. We then embed the hyper-graph structure into a Layered Markov Model in which transitions among web sites, page-sets and web pages are distinguished to compute reliability of web information. In addition, personalized rankings which are keys to next generation search engine will be produced by adapting the computation at local and global layers. Finally, we present some illustrative simulation results showing that the ranking system generated by the proposed approach is qualitatively comparable to or even better and more reliable than the ranking produced by some famous search engines such as Google. ICIC International (5)2010 ISSN 1881-803X.
-
A novel way of computing multimedia information similarities
Zin T., Tin P., Toriu T., Hama H.
ICIC Express Letters, Part B: Applications 1 ( 1 ) 27 - 32 2010.9
Language：Japanese Publishing type：Research paper (scientific journal) Publisher：ICIC Express Letters, Part B: Applications
This paper presents a new perspective on characterizing the similarity be- tween elements in multimedia information systems concerning with images, text docu- ments or, more generally, nodes of a weighted and undirected graph. It is based on a sequence of Embedded Markov-chains and its rich properties. More precisely, we com- pute quantities such as the steady state probability distributions and the average commute time, that provide similarities between any pair of nodes or information. This approach is not limited to only images but it is applicable to any type of data such as range sensor data, communication frequency data and so on. Besides, it is adaptable for detecting and retrieving information from the multimedia database by using our newly developed Markov-based similarity measures. Specifically, the proposed system contributes a new method of foreground/background modeling in visible or non-visible multi-sensors infor- mation processing, detection, tracking and analyzing human-machine related surveillance systems. Experimental results on real life data sets show that the Markov Chain-based similarities perform well in comparison with other methods. © 2010 ICIC International.