Browsing by Author "Gu, Hui-Zhen"

Now showing 1 - 7 of 7

A view-invariant and anti-reflection algorithm for car body extraction and color classification
(2013-08-01)
Gu, Hui-Zhen
;
Lee, Suh-Yin
;
資訊工程學系
;
Department of Computer Science
This study proposes an intelligent algorithm with tri-state architecture for real-time car body extraction and color classification. The algorithm is capable of managing both the difficulties of viewpoint and light reflection. Because the influence of light reflection is significantly different on bright, dark, and colored cars, three different strategies are designed for various color categories to acquire a more intact car body. A SARM (Separating and Re-Merging) algorithm is proposed to separate the car body and the background, and recover the entire car body more completely. A robust selection algorithm is also performed to determine the correct color category and car body. Then, the color type of the vehicle is decided only by the pixels in the extracted car body. The experimental results show that the tri-state method can extract almost 90% of car body pixels from a car image. Over 98% of car images are distinguished correctly in their categories, and the average accuracy of the 10-color-type classification is higher than 93%. Furthermore, the computation load of the proposed method is light; therefore it is applicable for real-time systems.
Baseball Event Semantic Exploring System Using HMM
(2011)
Tsai, Wei-Chin
;
Chen, Hua-Tsung
;
Gu, Hui-Zhen
;
Lee, Suh-Yin
;
Yu, Jen-Yu
Despite a lot of research efforts in baseball video processing in recent years, little work has been done in analyzing the detailed semantic baseball event detection. This paper presents an effective and efficient baseball event classification system for broadcast baseball videos. Utilizing the specifications of the baseball field and the regularity of shot transition, the system recognizes highlight in video clips and identifies what semantic baseball event of the baseball clips is currently proceeding. First, a video is segmented into several highlights starting with a PC (Pitcher and Catcher) shot and ending up with some specific shots. Before every baseball event classifier is designed, several novel schemes including some specific features such as soil percentage and objects extraction such as first base are applied. The extracted mid-level cues are used to develop baseball event classifiers based on an HMM (Hidden Markov model). Due to specific features detection, more hitting baseball events are detected and the simulation results show that the classification of twelve significant baseball events is very promising.
Car model recognition by utilizing symmetric property to overcome severe pose variation
(2013-02-01)
Gu, Hui-Zhen
;
Lee, Suh-Yin
;
資訊工程學系
;
Department of Computer Science
This paper presents a mirror morphing scheme to deal with the challenging pose variation problem in car model recognition. Conventionally, researchers adopt pose estimation techniques to overcome the pose problem, whereas it is difficult to obtain very accurate pose estimation. Moreover, slight deviation in pose estimation degrades the recognition performance dramatically. The mirror morphing technique utilizes the symmetric property of cars to normalize car images of any orientation into a typical view. Therefore, the pose error and center bias can be eliminated and satisfactory recognition performance can be obtained. To support mirror morphing, active shape model (ASM) is used to acquire car shape information. An effective pose and center estimation approach is also proposed to provide a good initialization for ASM. In experiments, our proposed car model recognition system can achieve very high recognition rate (> 95%) with very low probability of false alarm even when it is dealing with the severe pose problem in the cases of cars with similar shape and color.
HVPN: The Combination of Horizontal and Vertical Pose Normalization for Face Recognition
(2009)
Gu, Hui-Zhen
;
Kao, Yung-Wei
;
Lee, Suh-Yin
;
Yuan, Shyan-Ming
;
資訊科學與工程研究所
Face recognition has received much attention with numerous applications in various fields. Although many face recognition algorithms have been proposed, usually they are not highly accurate enough when the poses of faces vary considerably. In order to solve this problem, some researches have proposed pose normalization algorithm to eliminate the negative effect cause by poses. However, only horizontal normalization has been considered in these researches. In this paper, the HVPN (Horizontal and Vertical Pose Normalization) system is proposed to accommodate the pose problem effectively. A pose invariant reference model is re-rendered after the horizontal and vertical pose normalization sequentially. The proposed face recognition system is evaluated based on the face database constructed by our self. ne experimental results demonstrate that pose normalization can improve the recognition performance using conventional principal component analysis (PCA) and linear discriminant analysis (LDA) approaches under varying pose. Moreover, we show that the combination of horizontal and vertical pose normalization can be evaluated with higher performance than mere the horizontal pose normalization.
Integration of face and hand gesture recognition
(2008)
Kao, Yung-Wei
;
Gu, Hui-Zhen
;
Yuan, Shyan-Ming
;
資訊科學與工程研究所
;
Institute of Computer Science and Engineering
Face recognition and hand gesture recognition technologies have been developed separately for many years. Usually they are treated as independent systems. In this paper, we integrate the face and hand gesture recognition. We claim that the face recognition rate can be improved by hand gesture recognition. Also, we propose a security elevator scenario. Finally, we simulate this security elevator scenario by PCA method, based on the ORL database, and show that the face recognition rate and overall accuracy is improved after integration. We believe that this is a general method to integrate two recognition engines, not only for face and hand gesture recognition.
Traffic Congestion Classification for Nighttime Surveillance Videos
(2012-01-01)
Chen, Hua-Tsung
;
Tsai, Li-Wu
;
Gu, Hui-Zhen
;
Lee, Suh-Yin
;
Lin, Bao-Shuh P.
Traffic surveillance systems have been widely used for traffic monitoring. If the degree of traffic congestion can be evaluated from the surveillance videos immediately, the drivers can choose alternate routes to avoid traffic jam when traffic congestion arises. Compared to daytime surveillance, some tough factors such as poor visibility and higher noise increase the difficulty in video understanding under nighttime environments. In this paper, we propose a framework of traffic congestion classification for nighttime surveillance videos. The framework consists of three steps: the first one is to detect headlights based on three salient headlight features. Second, headlights are grouped into individual vehicles by evaluating their correlations. Third, a virtual detection line is adopted to gather the traffic information for traffic congestion evaluation. Then the traffic congestion is classified into five levels: jam, heavy, medium, mild and low in real-time. We use freeway nighttime surveillance videos to demonstrate the performances on accuracy and computation. Satisfactory experimental results validate the effectiveness of the proposed framework.
uCloud: a user-centric key management scheme for cloud data protection
(2013)
Kao, Yung-Wei
;
Huang, Kuan-Ying
;
Gu, Hui-Zhen
;
Yuan, Shyan-Ming
;
資訊工程學系
One of the most challenging problems of cloud service solicitation is to persuade users to trust the security of cloud service and upload their sensitive data. Although cloud service providers can claim that their services are well-protected by elaborate encryption mechanisms, traditional cloud systems still cannot persuade the users that even if the cloud servers are compromised, the data are still securely protected. This study proposes uCloud, a user-centric key management scheme for cloud data protection, to solve this problem. uCloud utilises RSA and indirectly encrypts users' data by users' public keys, but stores the users' private keys on neither servers nor users' PCs; instead, the private keys are stored on users' mobile devices and presented via two-dimensional (2D) barcode images when they are utilised to decrypt users' sensitive data. In this manner, users' data are safely protected even if the cloud servers are compromised. Also, uCloud provides users with the experience of managing visible private keys by storing the keys into mobile phones and displaying them via 2D barcodes. Moreover, three scenarios: personal storage, home surveillance and enterprise storage scenarios are proposed to present the practicability of uCloud. In addition, a hierarchical structure is designed for basic key backup and data sharing in the proposed scheme.