Detailed Research

Anomaly Detection


Surface Anomaly Detection of Wood Grain Image Using Fourier Transform : A Preliminary Study

2022. 12

Jongwook Si and Sungyoung Kim

Download


In this paper, we introduce anomaly detection method of reconstruction method for wood grain images. Fourier transform is performed on the result of a deep learning model with an autoencoder structure. Then, certain frequency domain is removed and restored to detect normal and defect using the sum of errors.

Defect Detection of Carpet PatternUsing Average Filtering and Morphology

2022. 10

Jongwook Si and Sungyoung Kim

Download


The overall distribution is calculated by blurring the image using mean value filtering and the poor position is corrected through expansion and erosion operations of the morphology. In addition, the results of analyzing the optimal parameters are shown through average value filtering and experiments according to each kernel size of morphology.

Multi-Classification of Solar Cell Infrared Images based on CNN Model

2022. 05

JongwookSi and Sungyoung Kim

Download


This paper introduces a method of classifying thermal image of solar cell into normal and abnormal patterns, that is, 8 classes, using the CNN model. The proposed network uses the structure of Residual Block to preserve the features of abnormal images and uses Dropout and Batch Normalization on each layer to prevent overfitting.

Defect Detection of Injection Molding Machine Results for Smart Factory

2021. 11

Moonnyeon Kim, Minsu Jeong, Younghyung Kim, Kwanghyun Cho and Sungyoung Kim

Download


There are many advantages to automating the detection of defects in the injection molding machine output by attaching sensors to the inside of the injection machine rather than relying on the efficiency of the operator and using the sensor information. Therefore, in this paper, we introduce sensor information in the injection molding machine for factory automation, and an injection molding machine defect detection using AutoEncoder and machine learning techniques.

Development of Automatic Classification of Defects in Solar Panels

2021. 11

Hyeoncheol Son and Sungyoung Kim

Download


The defect of the solar panel has a high temperature by dissipating heat due to the characteristics of the defect. Therefore, photographing heat of a panel with thermal imaging camera and a network that classifying solar defects into thermal imaging images is proposed. Since thermal images themselves represent characteristics, they include the precess of preserving the identity of input data using Residual block.

Traffic Accident Detection in First-Person Videos Based on Depth and Background Motion Estimation

2021. 03

Jongwook Si and Sungyoung Kim

Download


In this paper, we propose a method to determine if traffic accidents have occurred in the first-person videos. The detection of the accidents is based on future frame prediction and the frame predictions are performed based on GAN. The generator of the GAN generates a predicted frame from several previous frames. we use depth and background motion estimation to improve the performance of future frame prediction.

Fall Detection using Skeletal Coordinate Vector and LSTM Model

2020. 12

Jongwook Si , Hyeoncheol Son , Daseul Kim , Moonnyeon Kim , Jiyeon Jeong , Gyuree Kim , Younghyung Kim and Sungyoung Kim

Download


In this paper, the method of detecting falls was proposed with a focus on the aspect of safety among the methods of improving the quality of life of single-person households. First, we extract key points of human skeletons based on the existing method, vectorize them to represent correlation between them, and input them into LSTM to determine whether falls have occurred. We try to enhance the performance by using only representative feature points, not all feature points.

Vehicle-Level Traffic Accident Detection on Vehicle-Mounted Camera Based on Cascade Bi-LSTM

2020. 12

Hyeoncheol Son, Daseul Kim and Sungyoung Kim

Download


In this paper, we propose a traffic accident detection on vehicle-mounted camera. In the proposed method, the minimum bounding box coordinates the central coordinates on the bird's eye view and motion vectors of each vehicle object, and ego-motions of the vehicle equipped with dash-cam are extracted from the dash-cam video. By using extracted 4 kinds features as the input of Bi-LSTM (bidirectional LSTM), the accident probability (score) is predicted.

Traffic Accident Detection Based on Ego Motion and Object Tracking

2020. 08

Daseul Kim, Hyeoncheol Son, Jongwook Si and Sungyoung Kim

Download


By the way, in a crowded road environment, it is so difficult to decide an accident has occurred because of parked vehicles at the edge of the road. It is not easy to discriminate against accidents from non-accidents because a moving vehicle and a stopped vehicle are mixed on a regular downtown road. In this paper, we try to increase the accuracy of the vehicle accident detection by using not only the motion of the surrounding vehicle but also ego-motion as the input of the Recurrent Neural Network (RNN).

Traffic Accident Detection Using Bird’s-Eye View and Vehicle Motion Vector

2020. 07

Hyeoncheol Son, Jongwook Si, Daseul Kim, Younghwan Lee and Sungyoung Kim

Download


In this paper, we'll show you how to determine if a car accident occurs in a video shot using a car black box. We propose that the proposed method first uses the bird's-eye coordinates obtained during object tracking to build on the distance between each vehicle. However, if an accident is determined by using only the distance, the number of cars is concentrated. To do this, we calculate the motion vectors for each vehicle, and we're going to calculate the motion vectors between them Use information (e.g., angle and size) to determine whether the vehicle is parked or stopped and to exclude it from the accident detection list.