Search Results

Now showing 1 - 6 of 6

Some of the metrics are blocked by your
consent settings
Robust tracking algorithm with designed marker for limited-power computer
(2016-03-23)
Siriteerakul, Teera
;
Gullayanon, Rutchanee
Object tracking is an important part of many visual systems. The robot navigation task, for example, does not only require a robust and accurate tracking, but also need the tracking module to perform in real-time. This may not be a problem with hi-end robotic system with powerful computing unit. However, in the case of mobile robot where the computing and the battery power are limited, there are only a few tracking methods available. Thus, we propose an algorithm, coupled with a designed marker, for fast and stable tracking. The proposed algorithm utilizes robustness of Hough Circle Transform to detect candidate locations for the marker. Then, all the detected locations are filtered by known properties of the designed marker until we have the correct location. An experiment done on Raspberry PI shows that the marker can be detected robustly with sufficient speed.
Some of the metrics are blocked by your
consent settings
Character classification framework based on support vector machine and k-nearest neighbour schemes
(2016-02-01)
Siriteerakul, Teera
;
Boonjing, Veera
;
Gullayanon, Rutchanee
The problem of Thai character classification can be difficult because of the large number of characters and the similarity in the shape of many characters. While previous work combined different fonts to build their classifier, this paper proposes a framework based on support vector machine (SVM) and k-NN schemes to exploit characteristics of each font separately. In this framework, each font is used to train an SVM separately. With the trained SVMs, a vector of predicted values can be produced for any input image. Then a class label of the input image can be found by a k-NN based scheme. The proposed framework performs well with familiar fonts while providing an acceptable performance on unfamiliar fonts.
Some of the metrics are blocked by your
consent settings
Fast tracking algorithm for designed marker
(2016-01-01)
Siriteerakul, Teera
;
Gullayanon, Rutchanee
Robust and real-time object tracking is a vital part of many application including the navigation of mobile robot and unmanned aerial vehicle. In such environment, the computational power and battery are limited. Hence, most of the state of the art algorithms will not be able to display their full potential. Thus, we propose an improved tracking algorithm, with a designed marker, to perform robustly in real-time in complex environments. Our algorithm based on a simple binarization and finding candidate connected components. Then, with physical and geometric property of the designed marker, we can filter out all the other components until we only have the marker. This scheme has been tested on Raspberry PI, a limited-power computing unit, and demonstrate sufficient robustness and speed.
Some of the metrics are blocked by your
consent settings
Support Vector Machine accuracy improvement with k-means clustering
(2013-12-01)
Siriteerakul, Teera
;
Boonjing, Veera
Support Vector Machine (SVM) is a classifier tool which, originally, uses a hyperplane as a border for separating two classes of data in hyperspace. However, if data from each class are not clustered together, the two classes might not be linearly separable. Typically, researchers attempted to resolve this issue by replacing the hyperplane with a complex border via kernel tricks. However, these kernel tricks could result in a longer training time or only a minute accuracy improvement (or both). On the other hand, if data from one class are separated into subclasses according to their proximity, then all the subclasses should be easily separated by hyperplanes. Therefore, this paper proposes a method to improve the accuracy of linear SVM by first applying k means clustering to each class of input data. Then, after clustered, a multi-classes linear SVM is trained using each subclass as a separate class. Thus, the trained SVM can identify any new input into a subclass which can be easily mapped to the correct class. To evaluate, the proposed method is experimentally used to classify images of Thai character where multiple fonts of characters can be taken as hidden clusters within classes. Empirically, the proposed method could achieve over 6% improvement from a linear SVM or SVMs with RBF or polynomial kernel. © 2013 IEEE.
Some of the metrics are blocked by your
consent settings
Mixed thai-english character classification based on histogram of oriented gradient feature
(2013-01-01)
Siriteerakul, Teera
The task of classifying mixed Thai-English characters carries considerable challenges due to the number and complexity of the characters. This paper proposes and empirically investigates the performance of a classification system that uses Histogram of Oriented Gradient as an image feature with Support Vector Machine as a classification tool. The experiments were done on the datasets provided by NECTEC which consists of over 600,000 printed images of individual characters from 142 distinct classes. With this proposed method, an accuracy of 97% can be achieved without a look up dictionary or any post-processing system. © 2013 IEEE.
Some of the metrics are blocked by your
consent settings
Estimating change in head pose from low resolution video using LBP-based tracking
(2011-12-01)
Siriteerakul, Teera
;
Sato, Yoichi
;
Boonjing, Veera
Head pose estimation from low resolution video can be a difficult task due to the limited quantity and quality of pixels which can be harvested for information. In this work, we propose a method for tracking change in head pose by solving a texture detection problem in an orientation space. In another words, the current head pose can be tracked by employed a local binary pattern based comparison between its texture and several textures estimated from rotating the previous head pose by several known angles. Thus, the selected previous head pose can be regarded as an estimator. To boost the tracking accuracy, multiple estimators, each with different previous head pose, are deployed to produce a combined result. Empirically, the proposed method outperforms the state of the art methods for head pose tracking in low resolution image sequence. Furthermore, the method also performs well on image sequences with varying light. © 2011 IEEE.

KMITL

Browse

Filters

Settings

Sort By

Results per page

Search Results