Phooi Yee Lau

A Comparative Study for Texture Classification Techniques on Wood Species Recognition Problem

Wood species recognition is a texture classification problem that has yet to be well studied. The textures observed on the cross section surface of the wood samples can be used to identify the species of the wood. In this paper, we tested various texture classification techniques, i.e. grey level co-occurrence matrices (GLCM), Gabor filters, combined GLCM and Gabor filters as well as covariance matrix. The experiments are conducted on 512 × 512 images of the six wood species from the CAIRO wood dataset. The experimental results show that the covariance matrix produced using the feature images generated by the Gabor filters is 85% compared to 78.33% for the raw GLCM, 73.33% for the Gabor filters and 76.67% for the combined GLCM and Gabor filters. The experimental results show that the covariance matrix has the best recognition rate.

Pay-as-you-use on-demand cloud service: An IPTV case

Cloud computing, the latest buzzwords in the technology sector, describes a new era of Internet usage when services are delivered over the Web. The present work proposes a new strategy to integrate cloud solutions to curb the rising IT costs, the complexity of network management, and infrastructure inefficiencies for IPTV. This new approach adopts two virtualization techniques: 1) server virtualization – virtual servers co-reside on a single dedicated server, and 2) application virtualization – the customer select to play a channel or videos from program menu. In these cases, all videos behave as if they are being streamed by a dedicated server. An architectural framework, named the Video-on-demand as a Service (VoDaaS), which formalizes the contextualization of the on-demand cloud service for IPTV, enabling subscribers to receive television programs or video stream from anywhere, is proposed. We highlighted the components of the framework and discussed preliminary implementation in the paper.

Rotational Invariant Wood Species Recognition through Wood Species Verification

An automated wood species recognition system using computer vision techniques is not widely used today, it is highly needed in various industries, but a wood identification expert is not easily trained to meet the market demand. This paper proposes a rotational invariant method using the grey level co-occurrence matrices (GLCM) as the features, an energy value representing the similarity between the test sample and the template is computed to decide whether the test sample is the same species as the template. A template is accepted when the energy is lower than the threshold value. The species with the highest number of accepted templates will be regarded as the recognition result. The experiment is conducted on six wood species of the CAIRO dataset with a total of 450 training samples and 60 testing samples and achieved a result of 80.00%.

Gabor Filters as Feature Images for Covariance Matrix on Texture Classification Problem

The two groups of popularly used texture analysis techniques for classification problems are the statistical and signal processing methods. In this paper, we propose to use a signal processing method, the Gabor filters to produce the feature images, and a statistical method, the covariance matrix to produce a set of features which show the statistical information of frequency domain. The experiments are conducted on 32 textures from the Brodatz texture dataset. The result that is obtained for the use of 24 Gabor filters to generate a 24 × 24 covariance matrix is 91.86%. The experiment results show that the use of Gabor filters as the feature image is better than the use of edge information and co-occurrence matrices.

One-dimensional Grey-level Co-occurrence Matrices for texture classification

The grey-level co-occurrence matrices (GLCM) has been widely used for various texture analysis implementations and has provided satisfying results. The conventional GLCM method is two dimensional as it focus on the co-occurrence of the specific pixel pairs. The one-dimensional GLCM reduces the matrices to a single dimension by focusing only on the differences of the grey level between pixel pairs. The experiment results on 32 Brodatz textures shows that in a same setting, the one-dimensional GLCM achieved a recognition rate of 83.01% while the conventional GLCM achieved a recognition rate of 81.35%. The results show that the one-dimensional GLCM can perform as good as the conventional GLCM but with fewer computations involved.

Mobile mBus System Using Near Field Communication

Near field communication (NFC) is a form of short range contactless and wireless communication between devices, a subset of RFID, with a much shorter communication range for security purposes. Any objects can become a passive device by tagging a NFC tag on them. In the past, RFID and smart card was a feasible choice for bus fare/token ticketing as their tag can cost as low as 10 cents/piece, being also the average price for a NFC tag. However, their tag readers are expensive, such as passive RFID and smart card reader, and not universal. Thus, this project proposes an economical and portable ticketing system, named mBus, designed to (1) simplify payments for bus rides and (2) low-cost fare/token tracking using NFC. The proposed mBus consist of (1) bTag, (2) bReader, and (3) bData. The bTag are able to (1) store information, and (2) communicate with an active reader device, bReader. User can purchase a bTag, being a NFC tag, and a bus pass. bTag contain crucial information such as user information and remaining token/fare. A mobile application, named as bReader, allows the bus driver to (1) read bTag, (2) act as a payment gateway, and (3) create new users. The system record user data in bData, being a remote SQL database using PHP web pages. The system is being designed as mobile application since mobile devices has huge market in the portable devices industry and recent mobile system has NFC chips installed in them, i.e. anyone who desires to use this bus fare/token ticketing system can find and obtain inexpensive NFC-embedded mobile device easily.

Video forensics on block detection with deblocking filter in H.264/AVC

Video forensics is an emerging research area that is still in need of studies, especially with regards to aspect of the advancement in video compression. De-blocking filters, firstly introduced in H.264/AVC, provide a means to reduce the blocking artifacts in compressed videos. These filters, in the forensic perspective, could induce a number of forensics threat, especially in the area of authentication, as it reduces the traces of block boundaries, which, in the past, often used in forensic studies. This paper proposed a spatial structural analysis (SSA) method to extract the de-blocking filter traces, by examining pixel intensity changes. At first, the SSA computes the intensity differences along the horizontal and vertical direction on the U channel of the compressed video. The SSA then locates all perpendicular turning points that were formed by a pair of intersected and perpendicular edges that appear to continuously share the same intensity difference in the same direction. Finally, a common spatial distance will be determined based on the highest occurrence of spatial distance between the horizontal and vertical lines formed along the perpendicular turning points detected. The distance will be used to determine if compression has been conducted for the video, where the common spatial distance computed is identical to the size of the macroblocks on both horizontal and vertical direction. The experimental results showed that the determined common spatial distance is 8 in the videos compressed with H.264/AVC that matches the size of the 8×8 macroblocks. Videos compressed without de-blocking filter showed stronger peak in the accumulated spatial distance count compared to those with de-blocking filters applied.

A distributed framework for mining financial data

Financial institutions are struggling with the challenges of managing and analyzing large dataset, especially if the dataset are accessed and shared among various subsidiaries from different location. These existing organizations, in the past, often employed a centralized control over their dataset as a way to curb data integrity and data solidarity. In this instance, a central processing unit, often a single machine, could consume all resources, else not affording to process the data altogether. To solve these incumbent problems, distributed computing techniques are being introduced to improve the efficiency of data processing. In this project, we present a framework to distribute processing resources for data mining tool, i.e. to provide the ability to distribute the processes, over a cloud software stack using open source tools and platforms. As a conclusion to this framework, we discussed the open source tools and platforms along with our preliminary investigation results.

Automatic analysis of deep-water remotely operated vehicle footage for estimation of Norway lobster abundance

Underwater imaging is being used increasingly by marine biologists as a means to assess the abundance of marine resources and their biodiversity. Previously, we developed the first automatic approach for estimating the abundance of Norway lobsters and counting their burrows in video sequences captured using a monochrome camera mounted on trawling gear. In this paper, an alternative framework is proposed and tested using deep-water video sequences acquired via a remotely operated vehicle. The proposed framework consists of four modules: (1) preprocessing, (2) object detection and classification, (3) object-tracking, and (4) quantification. Encouraging results were obtained from available test videos for the automatic video-based abundance estimation in comparison with manual counts by human experts (ground truth). For the available test set, the proposed system achieved 100% precision and recall for lobster counting, and around 83% precision and recall for burrow detection.

GuARD: A Real-Time System for Detecting Aggressive Human Behavior in Cage Environment

The relative closeness in a cage environment, such as lock-up or elevator, will become a place that is conducive to conduct criminal activities such as fighting. Monitoring the activities, in the cage environment, therefore, became a necessity. However, placing security guards could be inefficient and ineffective, as it is impossible to monitor the scene 24 by 7. A vision-based system, employing video analysis technology, to detect abnormalities such as aggressive behavior, becomes a challenging and emerging problem. In order to monitor suspicious activities in a cage environment, the system should be able track individuals from the scene, to identify their action, and to keep a record of how often these aggressive behaviors happen. On top of the previous consideration, the system should be implemented in real-time, whereby, the following conditions were taken into consideration, being: (1) wide angle (fish-eye) (2) resolution (low) (3) number of people (4) lighting (low). This paper proposes to develop a vision-based system that is able to monitor aggressive activities of individuals in a cage environment. This work focuses on analyzing the temporal feature of aggressive movement, taking consideration of the acquisition limitations discusses previously. Experimental results show that the proposed system is easily realized and achieved real-time performance, even in low performance computer.