Vooi Voon Yap

Implementation and optimization of human tracking system using ARM embedded platform

Computer vision is a field that includes methods for acquiring, processing, analyzing and understanding images. In the embedded world, computer vision applications have to fight with limited processing power and limited resources to achieve optimized algorithms and high performance. This paper presents work on implementing a human tracking system on both Intel based PC platform and embedded systems to optimize the algorithms for high performance. The algorithms are benchmarked on the Intel platform processor and BeagleBoard xM baed on low-power Texas Instruments (TI) DM3730 ARM processor. Functions and library in OpenCV which developed by Intel Corporation was utilized for building the human tracking algorithms.

Local adaptive approach toward segmentation of microscopic images of activated sludge flocs

Abstract. Activated sludge process is a widely used method to treat domestic and industrial effluents. The conditions of activated sludge wastewater treatment plant (AS-WWTP) are related to the morphological properties of flocs (microbial aggregates) and filaments, and are required to be monitored for normal operation of the plant. Image processing and analysis is a potential time-efficient monitoring tool for AS-WWTPs. Local adaptive segmentation algorithms are proposed for bright-field microscopic images of activated sludge flocs. Two basic modules are suggested for Otsu thresholding-based local adaptive algorithms with irregular illumination compensation. The performance of the algorithms has been compared with state-of-the-art local adaptive algorithms of Sauvola, Bradley, Feng, and c-mean. The comparisons are done using a number of region- and nonregion-based metrics at different microscopic magnifications and quantification of flocs. The performance metrics show that the proposed algorithms performed better and, in some cases, were comparable to the state-of the-art algorithms. The performance metrics were also assessed subjectively for their suitability for segmentations of activated sludge images. The region-based metrics such as false negative ratio, sensitivity, and negative predictive value gave inconsistent results as compared to other segmentation assessment metrics.

Measuring learning outcomes of Bachelor degree program in outcome-based education

Learning outcomes are skill and ability acquired shortly after the conclusion of a learning session, which is typically a 14-week semester in undergraduate engineering degree program. In this report, a 4 year programme offering 130 credit hours is chosen as an example. One of the criteria in Outcome Based Education is the assessment of learning outcomes as achievement indicator for continuous performance quality management. This criterion is driving educators to look for ways to deliver continuous improvement. This paper reports on practical method of assessing the learning outcomes on unit by unit and the contributions to the degree programme learning outcomes.

Segmentation and grading of eczema skin lesions

In this paper Eczema skin lesions are segmented and graded using image processing and analysis. For preprocessing adaptive light compensation and gamma correction has been used. The effect of color space normalization has been studied for lesion segmentation. K-means algorithm is used for segmentation. We have used RGB and CIELab color models; and their normalized I and II versions. The experimental results show that normalized color spaces give better segmentation results than the normal color spaces. Our proposed algorithm gives best segmentation accuracy of 84.6% for RGB normalized II color space for the G channel for adaptive light compensation. The grading accuracy for erythema is around 70%.

Human behavioral analytics system for video surveillance

Video surveillance system is an important system to monitor safety and security for public place. Unfortunately, most of them do not have additional. In this paper, we have proposed an algorithm for detect human aggressive behavior by using Microsoft Kinect sensor. In our proposed algorithm, Cartesian coordinate formula is used for calculate position of body joints detected by Microsoft Kinect sensor. By analysis position of body joints, we can differentiate whether it is a normal or aggressive behavior. Experiment results show that our proposed method has an average accuracy of 95.83% for punching pose, and almost perfectly detected for kicking and normal behavior.

Generalized classification modeling of activated sludge process based on microscopic image analysis

ABSTRACT The state of activated sludge wastewater treatment process (AS WWTP) is conventionally identified by physico-chemical measurements which are costly, time-consuming and have associated environmental hazards. Image processing and analysis-based linear regression modeling has been used to monitor the AS WWTP. But it is plant- and state-specific in the sense that it cannot be generalized to multiple plants and states. Generalized classification modeling for state identification is the main objective of this work. By generalized classification, we mean that the identification model does not require any prior information about the state of the plant, and the resultant identification is valid for any plant in any state. In this paper, the generalized classification model for the AS process is proposed based on features extracted using morphological parameters of flocs. The images of the AS samples, collected from aeration tanks of nine plants, are acquired through bright-field microscopy. Feature-selection is performed in context of classification using sequential feature selection and least absolute shrinkage and selection operator. A support vector machine (SVM)-based state identification strategy was proposed with a new agreement solver module for imbalanced data of the states of AS plants. The classification results were compared with state-of-the-art multiclass SVMs (one-vs.-one and one-vs.-all), and ensemble classifiers using the performance metrics: accuracy, recall, specificity, precision, F measure and kappa coefficient (κ). The proposed strategy exhibits better results by identification of different states of different plants with accuracy 0.9423, and κ 0.6681 for the minority class data of bulking.

Tracking of electroencephalography signals across brain lobes using motion estimation and cross-correlation

Abstract. Electroencephalography (EEG) is the signal generated by electrical activity in the human brain. EEG topographic maps (topo-maps) give an idea of brain activation. Functional connectivity helps to find functionally integrated relationship between spatially separated brain regions. Brain connectivity can be measured by several methods. The classical methods calculate the coherence and correlation of the signal. We have developed an algorithm to map functional neural connectivity in the brain by using a full search block matching motion estimation algorithm. We have used oddball paradigm to examine the flow of activation across brain lobes for a specific activity. In the first step, the EEG signal is converted into topo-maps. The flow of activation between consecutive frames is tracked using full search block motion estimation, which appears in the form of motion vectors. In the second step, vector median filtering is used to obtain a smooth motion field by removing the unwanted noise. For each topo-map, several activation paths are tracked across various brain lobes. We have also developed correlation activity maps by following the correlation coefficient paths between electrodes. These paths are selected when the correlation coefficient between electrodes is >70%. We have compared the motion estimation path with the correlation coefficient activation maps. The tracked paths obtained by using motion estimation and correlation give very similar results. The inter-subject comparison shows that four out of five subjects tracked path involves all four (occipital, temporal, parietal, frontal) brain lobes for the same stimuli. The intra-subject analysis shows that three out of five subjects show different tracked lobes for different stimuli.

Fast directional search algorithm for fractional pixel motion estimation for H.264 AVC

Abstract. Fractional pixel motion estimation (ME) is required to achieve more accurate motion vectors and higher compression efficiency. This results in an increase in the computational complexity of the ME process because of additional computational overheads such as interpolation and fractional pixel search. Fast algorithms for fractional ME in H.264/AVC are presented. To reduce the complexity of fractional pixel ME, unimodal error surface assumption is used to check only some points in the fractional pixel search window. The proposed algorithm employs motion prediction, directional quadrant and point-based search pattern and early termination to speed up the process. Hence, the proposed algorithm efficiently explores the neighborhood of integer pixel based on high correlation that exists between the neighboring fractional pixels and unimodal property of error surface. The proposed search pattern and early termination reduce computational time by almost 8% to 18% as compared to the hierarchical fractional pixel algorithm employed in the reference software with a negligible degradation in video quality and negligible increase in bit rate.

A robust face recognition algorithm under varying illumination using adaptive retina modeling

Variation in illumination has a drastic effect on the appearance of a face image. This may hinder the automatic face recognition process. This paper presents a novel approach for face recognition under varying lighting conditions. The proposed algorithm uses adaptive retina modeling based illumination normalization. In the proposed approach, retina modeling is employed along with histogram remapping following normal distribution. Retina modeling is an approach that combines two adaptive nonlinear equations and a difference of Gaussians filter. Two databases: extended Yale B database and CMU PIE database are used to verify the proposed algorithm. For face recognition Gabor Kernel Fisher Analysis method is used. Experimental results show that the recognition rate for the face images with different illumination conditions has improved by the proposed approach. Average recognition rate for Extended Yale B database is 99.16%. Whereas, the recognition rate for CMU-PIE database is 99.64%.

Embedded based tailgating/piggybacking detection security system

Access control is a system which enables the authority to control and restrict access to a target sensitive or secured area. However, its effectiveness is highly dependent on the proper usage of the system by those who are granted access. These authorized personnel have total control of the door from the time it unlocks until it relocks again. One of the biggest weaknesses of automated access control is the lacked of a system to prevent a practice known as “tailgating” or “piggybacking”. This paper presents a low cost solution to this problem by using a single internet protocol camera and an embedded based control unit combining with video analytics technology. Results showed that the developed system is able to detect various tailgating/piggybacking violations successfully.