• Journal of the Korea Society of Computer and Information
• /
• v.27 no.5
• /
• pp.55-67
• /
• 2022

Sensor data can provide fault diagnosis for equipment. However, the cause analysis for fault results of equipment is not often provided. In this study, we propose an explainable convolutional neural network framework for the sensor-based time series classification model. We used sensor-based time series dataset, acquired from vehicles equipped with sensors, and the Wafer dataset, acquired from manufacturing process. Moreover, we used Cycle Signal dataset, acquired from real world mechanical equipment, and for Data augmentation methods, scaling and jittering were used to train our deep learning models. In addition, our proposed classification models are convolutional neural network based models, FCN, 1D-CNN, and ResNet, to compare evaluations for each model. Our experimental results show that the ResNet provides promising results in the context of time series classification with accuracy and F1 Score reaching 95%, improved by 3% compared to the previous study. Furthermore, we propose XAI methods, Class Activation Map and Layer Visualization, to interpret the experiment result. XAI methods can visualize the time series interval that shows important factors for sensor data classification.

• Journal of Drive and Control
• /
• v.20 no.4
• /
• pp.1-8
• /
• 2023

In recent years, robots have been utilized in various industries to reduce workload and enhance work efficiency. The following mobility offers users convenience by autonomously tracking specific locations and targets without the need for additional equipment such as forklifts or carts. In this paper, deep learning techniques were employed to recognize individuals and assign each of them a unique identifier to enable the recognition of a specific person even among multiple individuals. To achieve this, the distance and angle between the robot and the targeted individual are transmitted to respective controllers. Furthermore, this study explored the control methodology for mobility that tracks a specific person, utilizing Simultaneous Localization and Mapping (SLAM) and Proportional-Integral-Derivative (PID) control techniques. In the PID control method, a genetic algorithm is employed to extract the optimal gain value, subsequently evaluating PID performance through simulation. The SLAM method involves generating a map by synchronizing data from a 2D LiDAR and a depth camera using Real-Time Appearance-Based Mapping (RTAB-MAP). Experiments are conducted to compare and analyze the performance of the two control methods, visualizing the paths of both the human and the following mobility.

• Journal of IKEEE
• /
• v.28 no.1
• /
• pp.65-71
• /
• 2024

This paper proposes a neural network that can reduce the number of parameters while reducing matching errors in occluded regions to increase the accuracy of depth maps in stereo matching. Stereo matching-based object recognition is utilized in many fields to more accurately recognize situations using images. When there are many objects in a complex image, an occluded area is generated due to overlap between objects and occlusion by background, thereby lowering the accuracy of the depth map. To solve this problem, existing research methods that create context information and combine it with the cost volume or RoIselect in the occluded area increase the complexity of neural networks, making it difficult to learn and expensive to implement. In this paper, we create a depthwise seperable neural network that enhances regional feature extraction before cost volume generation, reducing the number of parameters and proposing a neural network that is robust to occlusion errors. Compared to PSMNet, the proposed neural network reduced the number of parameters by 30%, improving 5.3% in color error and 3.6% in test loss.

• Proceedings of the KOSOMBE Conference
• /
• v.1991 no.05
• /
• pp.65-70
• /
• 1991

A hybrid model which uses a probabilistic model and a MLP(multi layer perceptron) model for pattern recognition of EMG(electromyogram) signals is proposed in this paper. MLP model has problems which do not guarantee global minima of error due to learning method and have different approximation grade to bayesian probabilities due to different amounts and quality of training data, the number of hidden layers and hidden nodes, etc. Especially in the case of new test data which exclude design samples, the latter problem produces quite different results. The error probability of probabilistic model is closely related to the estimation error of the parameters used in the model and fidelity of assumtion. Generally, it is impossible to introduce the bayesian classifier to the probabilistic model of EMG signals because of unknown priori probabilities and is estimated by MLE(maximum likelihood estimate). In this paper we propose the method which get the MAP(maximum a posteriori probability) in the probabilistic model by estimating the priori probability distribution which minimize the error probability using the MLP. This method minimize the error probability of the probabilistic model as long as the realization of the MLP is optimal and approximate the minimum of error probability of each class of both models selectively. Alocating the reference coordinate of EMG signal to the outside of the body make it easy to suit to the applications which it is difficult to define and seperate using internal body coordinate. Simulation results show the benefit of the proposed model compared to use the MLP and the probabilistic model seperately.

• Journal of the Korea Society of Computer and Information
• /
• v.19 no.10
• /
• pp.35-42
• /
• 2014

It is possible to improve the obstacle avoidance capability by training and recognizing the obstacles which is in certain indoor environment. We propose the technique that use underlying intensity value along with intensity map from RGB-D image which is derived from stereo vision Kinect sensor and recognize an obstacle within constant distance. We test and experiment the accuracy and execution time of the pattern recognition algorithms like PCA, ICA, LDA, SVM to show the recognition possibility of it. From the comparison experiment between RGB-D data and intensity data, RGB-D data got 4.2% better accuracy rate than intensity data but intensity data got 29% and 31% faster than RGB-D in terms of training time and intensity data got 70% and 33% faster than RGB-D in terms of testing time for LDA and SVM, respectively. So, LDA, SVM have good accuracy and better training/testing time to use for obstacle avoidance based on intensity dataset of mobile robot.

• Journal of Navigation and Port Research
• /
• v.32 no.5
• /
• pp.355-361
• /
• 2008

Stereo matching is one of the most active research areas in computer vision. In this paper, we propose a stereo matching scheme using genetic algorithm for object depth recognition. The proposed approach considers the matching environment as an optimization problem and finds the optimal solution by using an evolutionary strategy. Accordingly, genetic operators are adapted for the circumstances of stereo matching. An individual is a disparity set. Horizontal pixel line of image is considered as a chromosome. A cost function is composed of certain constraints which are commonly used in stereo matching. Since the cost function consists of intensity, similarity and disparity smoothness, the matching process is considered at the same time in each generation. The LoG(Laplacian of Gaussian) edge is extracted and used in the determination of the chromosome. We validate our approach with experimental results on stereo images.

• The Transactions of the Korea Information Processing Society
• /
• v.6 no.4
• /
• pp.1136-1142
• /
• 1999

A difficulty of the visual inspection for translated, magnified and rotated objects exists owing to the limitation of recognition rate. In this paper, we perform to define Integral Logarithm Transform(ILT), to consider its characteristic for implementation of Translation-, Magnification- and Rotation-invariant inspection system, and to compare with other methods in inspection error rate. By using magnification and rotation invariance properties of ILT, it makes easier than other methods to extract the rotation degree. The new method employs the ILT for the good/bad inspection of translated, magnified and rotated objects and experiment is performed to achieve translation, magnification and rotation invariance. In other methods both magnification and rotation invariance can't be available. As the result of he experiment, it is not better than the self-organizing map in the improvement of recognition rate, but it shows us the possibility to be used as a tool for the good/bad inspection system.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.19 no.6
• /
• pp.109-116
• /
• 2019

This study proposed a robust detection algorithm. It detects face more stably with respect to changes in light and rotation for the identification of a face shape. Also it satisfies both efficiency of calculation and the function of detection. The algorithm proposed segmented the face area through pre-processing using a face shape as input information in an environment with a single camera and then identified the shape using a Self Organized Feature Map(SOFM). However, as it is not easy to exactly recognize a face area which is sensitive to light, it has a large degree of freedom, and there is a large error bound, to enhance the identification rate, rotation information on the face shape was made into a database and then a principal component analysis was conducted. Also, as there were fewer calculations due to the fewer dimensions, the time for real-time identification could be decreased.

• Journal of Institute of Control, Robotics and Systems
• /
• v.19 no.11
• /
• pp.1029-1035
• /
• 2013

This paper proposes a real time lane detection algorithm using LRF (Laser Range Finder) for autonomous navigation of a mobile robot. There are many technologies for safety of the vehicles such as airbags, ABS, EPS etc. The real time lane detection is a fundamental requirement for an automobile system that utilizes outside information of automobiles. Representative methods of lane recognition are vision-based and LRF-based systems. By the vision-based system, recognition of environment for three dimensional space becomes excellent only in good conditions for capturing images. However there are so many unexpected barriers such as bad illumination, occlusions, and vibrations that the vision cannot be used for satisfying the fundamental requirement. In this paper, we introduce a three dimensional lane detection algorithm using LRF, which is very robust against the illumination. For the three dimensional lane detections, the laser reflection difference between the asphalt and lane according to the color and distance has been utilized with the extraction of feature points. Also a stable tracking algorithm is introduced empirically in this research. The performance of the proposed algorithm of lane detection and tracking has been verified through the real experiments.

• Smart Media Journal
• /
• v.5 no.4
• /
• pp.90-95
• /
• 2016

Recently, the quality of cultural experience can be improved through a stereoscopic information service provided by the latest mobile-based Information Telecommunication technology without the human cultural commentators, which was used in order to enhance the understanding of our cultural heritage. The purpose of this paper is to produce contents that introduce cultural heritage using the Android-based GPS and augmented reality. In this paper we propose a culture content creation method that is based on location information such as user/cultural anomalies using GPS and augmented reality based on Markerless Tracking. Marker Detection Technology and Markerless Tracking Technology are used for smart phone's rapid recognition of augmented real world and accurate recognition according to the state of the cultural heritage. Also, the Google Map of Android is used to locate the user. The strength of this method lies in that it can be used for a variety of subjects while the existing methods are limited to certain kinds of augmented reality contents.