Min-Soo Jang

Performance comparison between backpropagation, neuro-fuzzy network, and SVM

In this study, we compare the performance of well-known neural networks, namely, back-propagation (BP) algorithm, Neuro-Fuzzy network and Support Vector Machine (SVM) using the standard three database sets: Wisconsin breast cancer, Iris and wine data. Since such database have been useful for evaluating performance of a group of machine learning algorithms, a series of experiments have been carried out for three algorithms using the cross validation method. Results suggest that SVM outperforms the others and the Neuro-Fuzzy network is better than the BP algorithm for this data set.

Vision based automatic occupant classification and pose recognition for smart airbag deployment

Airbags have been saved thousands of lives and reduced the number of serious injuries from collisions. However, the car occupant can be often hurt, or killed, by the airbag itself. For reducing the risk caused by airbag, designing a smart airbag is an important issue. This paper presents a vision based automatic system that can control triggering and intensity of airbag deployment. The system consists of an occupant classification system and an occupant pose recognition system, by which we aim to control whether the airbag should be triggered or not, and how strongly it should be deployed when it is triggered. Results suggest that the system is feasible as a vision based airbag controller.

Occupant Classification for Smart Airbag Using Stereovision and Support Vector Machines

Airbag in the cars plays an important role for the safety of occupants. However, Highway Traffic Safety report shows that many occupants are actually killed by wrong deployment of the airbags. For reducing risk caused by airbag, designing a smart airbag is an important issue. The present paper describes an occupant classification system, by which triggering of the airbag deployment can be controlled. The system consists of a pair of stereo cameras and a SVM classifier. Performance of the system shows its feasibility as a vision-based airbag controller.

Head detection and tracking for the car occupant’s pose recognition

This paper describes a Vision-based Occupant Pose Recognition (VOPR) system, which can ensure a safe airbag deployment. Head detection and its tracking are necessary for occupant’s pose recognition in the car, since the position of occupant’s head provides valuable information, such as his pose, size, position, and so on. We use the stereo cameras to extract a disparity map. Against variable lighting conditions including the night drive, we adopt infrared illumination as well as normal one. Results suggest that VOPR system is reliable and performs reasonably well.

Pose Classification of Car Occupant Using Stereovision and Support Vector Machines

Airbag in the cars plays an important role for the safety of occupants. However, Highway traffic safety report shows that many occupants are actually killed by wrong deployment of the airbags. For reducing risk caused by airbag, designing a smart airbag is an important issue. The present study describes an occupants’ pose classification system, by which triggering and intensity of the airbag deployment can be controlled. The system consists of a pair of stereo cameras and a SVM classifier. Performance of the system shows its feasibility as a vision-based airbag controller.

Shape recognition of the embryo cell using deformable template for micromanipulation

One of the difficulties of using Artificial Neural Networks (ANNs) to estimate atmospheric temperature is the large number of potential input variables available. In this study, four different feature extraction methods were used to reduce the input vector to train four networks to estimate temperature at different atmospheric levels. The four techniques used were: genetic algorithms (GA), coefficient of determination (CoD), mutual information (MI) and simple neural analysis (SNA). The results demonstrate that of the four methods used for this data set, mutual information and simple neural analysis can generate networks that have a smaller input parameter set, while still maintaining a high degree of accuracy.

Image contrast enhancement based on segmentation and tone curve control for LCD TV

In this paper, we propose an image contrast enhancement algorithm for an LCD TV. The proposed algorithm consists of two processes: the image segmentation process and the tone curve control process.

A stereovision based advanced airbag system

Occupant classification in the car is an essential issue for an advanced airbag system. The present paper describes a stereovision based occupant classification system (OCS) within an embedded system, by which triggering of the airbag deployment can be intelligently controlled. The embedded system consists of dual Digital Signal Processors; one is for stereo matching algorithm and the other is for calculating an SVM algorithm for the OCS. Performance was evaluated using our stereo image database. Results suggest that the system is satisfactory as an embedded OCS system.

An Embedded System for Posture Classification of the Vehicle Occupant

Posture analysis of the vehicle occupant is an essential issue for the smart airbag system. A vision-based system with stereo cameras provides diverse information such as volume of the occupant, distance from dash board to the occupant and so on. This paper describes the occupant posture classification method and its hardware implementation using two DSP chips. The system classifies occupants posture into two classes: in-position and out-of-position, and these outputs determine whether the airbag will be deployed. Performance was evaluated using our stereo image database. Results suggest that the system is satisfactory as an embedded occupant posture classification system

3-D pose tracking of the car occupant

Although airbags in the car play an important role for the safety of occupants, in fact, many peoples have been injured or killed by the deployment of airbags themselves. Such conventional airbags are deployed by the shock sensors. As an alternative approach, a vision-based smart airbag system could be promising. This paper describes a new method by which 3-D pose of the car occupant can be recognized. We combine 2-D head tracking information with a disparity map of the occupant for 3-D pose tracking. Result shows that the system can locate the head position around the passengers seat with a real-time basis.