Supaporn Kiattisin

A Comparison of Traditional and Neural Networks Forecasting Techniques for Container Throughput at Bangkok Port

Abstract Containerization is one of the important factors for Thailands economics. However, forecasts of container throughput growth and development of Bangkok Port, the significant port of Thailand, have been scant and the findings are divergence. Moreover, the existing literature emphasizes only two forecasting methods, namely time series and regression analysis. The aim of this paper is to explore Multilayer Perceptron (MLP) and Linear Regression for predicting future container throughput at Bangkok Port. Factors affecting cargo throughput at Bangkok Port were identified and then collected from Bank of Thailand, Office of the National Economic and Social Development Board, World Bank, Ministry of Interior, and Energy Policy and Planning Office. These factors were entered into MLP and Linear Regression forecasting models that generated a projection of cargo throughput. Subsequently, the results were measured by root mean squared error (RMSE) and mean absolute error (MAE). Based on the results, this research provides the best application of forecasting technique which is Neural Network — Multilayer Perceptron technique for predicting container throughput at Bangkok Port.

A match of X-ray teeth films using image processing based on special features of teeth

This paper presents a match of X-ray teeth films using image processing based on special features of teeth. This method will help the dental doctors to match simply a pair of teeth using the special features of the teeth films. Teethpsilas pictures are scanned and adjusted by a scanner and a computer, respectively, as well as then they are converted into binary code and decoded to the direction code (chain code). The chain code of each picture is compared with the statistical chain code. Therefore, the percentage of the same chain code is approximately 90% (i.e. matching same patterns) for the comparison of one root to one root (7 times) and two roots to two roots (7 times) while the percentage of the same chain code is reduced at relatively below 50% (i.e. matching different patterns) for comparison of one root to two roots (2 times).

Digital biometric facial image encryption using chaotic cellular automata for secure image storages

This paper presents the digital image encryption scheme for biometric facial image using Cellular Automata (CA) for secure image storage. The proposed scheme is relatively simple using a segmentation of CA binary image with embedded secret keys generated by the third class of well-known Wolfram Cellular Automata that exhibits chaotic patterns. Such segmented CA binary image is diffused to the shuffled and bit-plane separated of the original biometric facial image through to XOR operations. Experiments have been performed in MATLAB using a standard digital biometric facial image with the size of 160×160 pixels. Encryption qualitative performances are evaluated through pixel density histograms, 2-dimensional power spectral density, and vertical, horizontal, and diagonal correlation plots. For the encryption quantitative measures, correlation coefficients, entropy, NPCR and UACI are realized. Demonstrations of wrong-key decrypted image are also included. The proposed encryption scheme offers a potential alternative to digital biometric facial image storage in a various applications such as in border security control, payment system, or in crime prevention, detection, and forensics.

Implementation and quality evaluation of video telephony using Session Initiation Protocol

Video telephony is a big change of communication technologies and becomes a standard-based technology available on personal computers and mobile phones for today. Video telephony technology has been developed using several advanced technologies which can be simplified into two fundamental parts, i.e., server and client. The server enables the clients to keep connect to each other while the client captures the audio and video data for transmission and converts the transmission of data back to the form of audio and video signals. For our implemented systems, Session Initiation Protocol (SIP) server is used as video telephony server. Video codec such as VP8, MPEG4, H263, and H264 are used as video compression in the client. The performances of codec are evaluated by objective and subjective quality tests. The result shows that the implemented system can work properly and the video quality of each codec is also evaluated.

A simulation of 6R industrial articulated robot arm using backpropagation neural network

This paper presents a simulation of a 6 degrees-of-freedom (6R) articulated robot arm using backpropagation neural network to solve the problem regarding inverse kinematics for the industrial articulated robot. The Denavit - Hartenberg model is used to analyze the robot arm movement. Next, the forward kinematics is used to identify the relationships for each joint of the robot arm and to determine various parameters for learning system of random neural network for 5,000 data points. The simulation results show that the robot arm can move to target positions with precision, and the average error for the entire 6 joints is at approximately 4.03 degrees.

Investigation of chest x-ray images based on medical knowledge and balanced histograms

The primary checking for our health at hospital needs to include a chest x-ray as routine diagnosis because it effectively illustrates the lung diseases especially tuberculosis or lung cancer which are asymptomatic earlier. It is a convenient and quick process with a low cost in comparison with other studies. This paper presents an investigation of the radiographs of lung from the chest x-ray using on medical knowledge and balanced histogram. Selected images of lungs are depicted by the use of an active contour (e.g. snake algorithm) to find two regions of lungs (left and right). Then, such two regions of lungs are represented for two histograms which are profiles of two lung patterns. Such two histograms are compared for normal and abnormal lungs using a method of center of gravity (COG) to demonstrate the difference of both lung radiographs. If two histograms are balance, then the result is a normal case. However, if they are not balance, then it is an abnormal case. For the experimental results, the overall accuracy is at approximately 95% which there are 100 samples of patients for testing their lung images. All samples are previously checked from the medical doctors.

A method of detecting tonsillitis images based on medical knowledge and neural network

Tonsillitis is a disease occurring mostly in child and adults as this disease may take to the other effects. Nowadays, a detection of tonsil grand exploits medical doctors diagnosis to check on oral cavity. Therefore, this paper presents a method of detecting tonsillitis images based on medical knowledge and neural network (NN); as well as, the paper considers three important factors which can be indicated in swelling by the pictures in terms of a) the ratio of tonsil grand dimension, b) average of tonsil grand color and c) surface of tonsil grand as it is purulent (yes/no) using two dimensional Fast Fourier Transform (2D FFT). Finally, the three factors are inputted into NN, and samples of 30 pictures are used for training into the NN which is divided by tonsillitis patience 15 pictures and usual tonsil grand 15 pictures. In the experimental results, 20 pictures are tested to compare with the result of the medical doctors demonstration as the result of correction approximately at 90%.

A partial encryption scheme using absolute-value chaotic map for secure electronic health records

This paper presents a partial encryption scheme using absolute-value chaotic map for secure electronic health records (EHR). The HER system has been an emerging technology that allows medical personals to create, manage, and control medical data electronically through specific database or even web browsers. The proposed encryption scheme realizes XOR operations between separated planes of binary gray-scale image and a binaty imgae generated by an absolute-value chaotic map. The proposed is relatively simple containning a single absolute-value function with two constants and offers complex and robust dynamical behaviors in terms of random output values. Experiments have been performed in MATLAB using a magnetic resosnace image which is divided into 64 sub-blocks and 13th itterations were proceeded for encryption. Encryption qualitative performances are evaluated through pixel density histograms, 2-dimensional power spectral density, and vertical, horizontal, and diagonal correlation plots. For the encryption quantitative measures, correlation coefficients, entropy, NPCR and UACI are realized. Demonstrations of wrong-key decrypted image are also included. The proposed encryption scheme offers a potential alternative to a secure medical data records and web browsing through clound computing systems.

A detection of defect in diamond images using 2-D haar wavelet transform

This paper presents a detection of defect in diamond images using 2-D wavelet transform for helping to solve problems in diamond quality examination by photographs. The wavelet is able to examine images in a form of texture, graph and histogram which this method is cable to take these values to analyze the differences between defective and flawless diamond because the image of each diamond have different textures. There are more or less the texture depending on many factors such as square cut and flaw. With this approach, the examining process of diamond detection is reduced then it can be done faster. As a result of 2-D wavelet transform, there are 30 diamond images for testing which all images can be verified for the flaw in diamond images.

An Appraisal Model of Real Estate in Thailand Using Fuzzy Lattice Reasoning

This paper presents an appraisal model of real estate in Thailand using fuzzy logic which is a method for a capable of solving any value in blank land submarkets relative to clustering methods based on classic (or crisp) set theory. The valuer shall record the inspection result regarding condition and location of such land. Systematic analysis shall be applied and the fact of the land’s condition will also be recorded. Generally, such valuation data will be taken to compare with other comparable data and then classified by the valuer according to its significance, namely, A, B, C and D. This classification criteria is still not clear, especially weighted-factor of such property (Effective Factor of Property). Therefore, this problem has drawn attention from the researcher to explore the causes of such problem and at the same time propose models for clustering using fuzzy as a tool for classifying the components of the property (Effective Factor of Property) in order to determine upon such vagueness regarding weighted-factor. Issues of choosing algorithm parameters are discussed on the basis of applying fuzzy clustering to 101 metropolitan areas in the Thailand. The result from the experiment shows that the components of the property are weighed more appropriately and closely to the real value which can give the percentage of reliance to be at approximately 97 %. This enables the valuer to determine and make a comment on property value that its evaluated value becomes closer to the real one to the greatest extent.