Recently, Radio on Free space optics (RoFSO) has gained growing attention as a key technology for complementing optical fiber networks to provide broadband wireless access in areas where the deployment of optical fiber is not feasible. In this paper, we present an analytical and experimental investigation on transmission of terrestrial digital TV broadcasting signals over free-space optical (FSO) link. We introduce an analytical model for transmission of the Orthogonal Frequency Division Multiplexing (OFDM) based signals over RoFSO link and a closed-form symbol error probability expression taking into account the laser diode nonlinear distortion as well as the weak to strong atmospheric turbulence effect of the FSO channel. We then present an experimental evaluation of transmitting terrestrial digital broadcasting signals over 1-km FSO link. The obtained results are used for designing, predicting and evaluating the RoFSO system capable of transmitting multiple wireless services over turbulent FSO link.

View full abstract

One of high level functions of brain is a vision system. The organ is readily visualized and we can approach understanding of brain function most easily among mental activities. Hereupon, we have been shown by simulation on the stand point of evolutionary theory that our vision function might be evolved to fear rather a posteriori than hereditary when we get a sight of long and narrow shape like snake. The other side of imge processing, a depth recognition from images is one of the most important task. Therefore, we discuss whether it is naturally generated in the process of evolution by simulating binocular stereopsis system of artificial life. First, we implement the stereo perception ability utilizing artificial creatures with 2 types of binocular stereopsis with neural network and GA. Since the search space is huge, the evolution processes are separated in two groups; children (training only) and adults (multiplication). We implement 2D recognition of the target and 2D prediction map of the moving object of the creature. We show by the simulation results that the function of binocular stereopsis is possibly generated in the process of natural evolution.

View full abstract

By the image capturing condition, the contrast of the image captured by camera does not become good. Therefore, it is popular to improve the contrast of images with the contrast improving method. However, the conventional methods do not always generate good contrast images, because these methods do not use the subjective factors of image. This paper proposes a method for image contrast improvement based on subjective factors. The method transforms distribution of all gray-level by neural network that has been optimized according to the back propagation learning. The experimental results show that the resultant images with good contrast were generated by the proposed method in comparison with by the conventional methods.

View full abstract

In this paper, we propose the method of three dimensional recovery from a photograph. In general, two photographs are used in the case of three dimensional recovery. Three dimensional recovery cannot be solved from a photograph. In the other hand, human percept the three dimension from a photograph. This is reason why the knowledge exists from experiences. In this paper, we generate three dimensional space from a photograph by showing knowledge. We use the vertical relations between quadrangles and vertical relations between vertices of quadrangles as the knowledge. We show the effectiveness of this method by recovering three dimensional space from a photograph.

View full abstract

In this paper, we propose a method to detect a specular reflection area from the road surface by using a camera moving along it. In this method, the intensity of the same point on the area is tracked. Based on the change of the intensity, the specular reflection region is detected. When the target area is a specular reflection region, the intensity changes due to the change of the incident ray. The experiment is done against the specular reflection region and the result shows the proposed method can detect the specular reflection region.

View full abstract

This paper deals with the precise positioning technique using image matching in conditions that it is impossible to extract the object. We propose a method for image matching using local feature in order to measure the distance of uniformly colored materials. We focus attention on feature of a sub-range texture, using images that converted gray-levels into coefficient of correlation. The results of experiments show proposed method can measure the distance at high-accuracy and high-speed under the practical conditions of its use.

View full abstract

Most of current on-line handwriting recognition for Thai scripts systems is based on natural writing. Those systems allow user to write in unconstrained handwriting, which is convenient but users have found a lot of error in recognition. Our proposed method “Simplified Thai Script” (STC) aims to reduce the error rate by designed an easy-to-write constructed scripts based on a dot line of the scripts. We ignored heading part, cursive parts then use straight line instead of cursive line and divide writing panel into 5 panels to make it more natural to write. We conducted the experiment with 16 Thai natives to compare the accuracy rate and writing speed (CPM) of STC with two existing programs; ThaiWinCE and InPad. The result shows that STC and InPad have 100% accuracy rate while ThaiWinCE has only 23% accuracy rate. The average writing speed of STC is 40.08 CPM which is considerably faster than InPad that has writing speed at 23.50 CPM while the CPM of ThaiWinCE could not be determined because of its error is too high. Moreover, STC is also faster than Multi-tap which is the commonly used text input method for Thai at present which has typing speed at 26.67 CPM.

View full abstract

Pursuit Reinforcement guided Competitive Learning: PRCL, introducing pursuit algorithm on relatively fast reinforcement guided competitive learning method as more fast online-clustering scheme, when the number of data and distribution of data are varying, is proposed. One of applications of the proposed method is image portion retrievals from the relatively large scale of the images such as Earth observation satellite images. It is found that the proposed method shows relatively fast on the retrievals in comparison to the other existing conventional online-clustering such as Vector Quatization: VQ. Moreover, the proposed method shows much faster than the others for the multi-stage retrievals of image portion as well as scale estimation.

View full abstract