• Journal of the Korean Institute of Electrical and Electronic Material Engineers
• /
• v.25 no.1
• /
• pp.76-83
• /
• 2012

In a real-time indoor place recognition system using image features detection, specific markers included in input image should be detected exactly and quickly. However because the same markers in image are shown up differently depending to movement, direction and angle of camera, it is required a method to solve such problems. This paper proposes a technique to extract the features of object without regard to change of the object scale. To support real-time operation, it adopts SURF(Speeded up Robust Features) which enables fast feature detection. Another feature of this system is the user mark designation which makes possible for user to designate marks from input image for location detection in advance. Unlike to use hardware marks, the feature above has an advantage that the designated marks can be used without any manipulation to recognize location in input image.

• Asia pacific journal of information systems
• /
• v.31 no.3
• /
• pp.335-357
• /
• 2021

As reproduction of images can be done with ease, copy detection has increasingly become important. In the duplication process, image modifications are likely to occur and some alterations are deliberate and can be viewed as attacks. A wide range of copy detection techniques has been proposed. In our study, content-based copy detection, which basically applies DCT-based features for images, namely, pixel values, edges, texture information and frequency-domain component distribution, is employed. Experiments are carried out to evaluate robustness and sensitivity of DCT-based features from attacks. As different types of DCT-based features hold different pieces of information, how features and attacks are related can be shown in their robustness and sensitivity. Rather than searching for proper features, use of robustness and sensitivity is proposed here to realize how the attacked features have changed when an image attack occurs. The experiments show that, out of ten attacks, the neural networks are able to detect seven attacks namely, Gaussian noise, S&P noise, Gamma correction (high), blurring, resizing (big), compression and rotation with mostly related to their sensitive features.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.1
• /
• pp.245-265
• /
• 2022

In response to problems such as insufficient extraction information, low detection accuracy, and frequent misdetection in the field of Thangka image defects, this paper proposes a YOLOv5 prediction algorithm fused with the attention mechanism. Firstly, the Backbone network is used for feature extraction, and the attention mechanism is fused to represent different features, so that the network can fully extract the texture and semantic features of the defect area. The extracted features are then weighted and fused, so as to reduce the loss of information. Next, the weighted fused features are transferred to the Neck network, the semantic features and texture features of different layers are fused by FPN, and the defect target is located more accurately by PAN. In the detection network, the CIOU loss function is used to replace the GIOU loss function to locate the image defect area quickly and accurately, generate the bounding box, and predict the defect category. The results show that compared with the original network, YOLOv5-SE and YOLOv5-CBAM achieve an improvement of 8.95% and 12.87% in detection accuracy respectively. The improved networks can identify the location and category of defects more accurately, and greatly improve the accuracy of defect detection of Thangka images.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.13 no.4
• /
• pp.284-290
• /
• 2013

Feature detection is very important to image processing area. In this paper we compare and analyze some characteristics of image processing algorithms for corner and blob feature detection. We also analyze the simulation results through image matching process. We show that how these algorithms work and how fast they execute. The simulation results are shown for helping us to select an algorithm or several algorithms extracting corner and blob feature.

• Journal of Information Processing Systems
• /
• v.13 no.3
• /
• pp.590-598
• /
• 2017

Recently, there has been an increase in the number of hazardous events, such as fire accidents. Monitoring systems that rely on human resources depend on people; hence, the performance of the system can be degraded when human operators are fatigued or tensed. It is easy to use fire alarm boxes; however, these are frequently activated by external factors such as temperature and humidity. We propose an approach to fire detection using an image processing technique. In this paper, we propose a fire detection method using multichannel information and gray level co-occurrence matrix (GLCM) image features. Multi-channels consist of RGB, YCbCr, and HSV color spaces. The flame color and smoke texture information are used to detect the flames and smoke, respectively. The experimental results show that the proposed method performs better than the previous method in terms of accuracy of fire detection.

• International Journal of Internet, Broadcasting and Communication
• /
• v.10 no.4
• /
• pp.50-64
• /
• 2018

This paper describes a visual object detection approach utilizing ensemble based machine learning. Object detection methods employing 1D features have the benefit of fast calculation speed. However, for real image with complex background, detection accuracy and performance are degraded. In this paper, we propose an ensemble learning algorithm that combines a 1D feature classifier and 2D DNF (Disjunctive Normal Form) classifier to improve the object detection performance in a single input image. Also, to improve the computing efficiency and accuracy, we propose a feature selecting method to reduce the computing time and ensemble algorithm by combining the 1D features and 2D DNF features. In the verification experiments, we selected the Haar-like feature as the 1D image descriptor, and demonstrated the performance of the algorithm on a few datasets such as face and vehicle.

• The Transactions of the Korean Institute of Electrical Engineers D
• /
• v.54 no.3
• /
• pp.166-173
• /
• 2005

Preserving image features, edges and details in the process of impulsive noise filtering is an important problem. To avoid image blurring, only corrupted pixels must be filtered. In this paper, we propose an effective impulse noise detection method using Self-Organizing Neural Network(SONN) which applies median filter selectively for removing random-valued impulse noises while preserving image features, edges and details. Using a $3\times3$ window, we obtain useful local features with which impulse noise patterns are classified. SONN is trained with sample image patterns and each pixel pattern is classified by its local information in the image. The results of the experiments with various images which are the noise range of $5-15\%$ show that our method performs better than other methods which use multiple threshold values for impulse noise detection.

• Journal of Broadcast Engineering
• /
• v.25 no.7
• /
• pp.1081-1094
• /
• 2020

With the recent development of deep learning, most computer vision-related tasks are being solved with deep learning-based network technologies such as CNN and RNN. Computer vision tasks such as object detection or object segmentation use intermediate features extracted from the same backbone such as Resnet or FPN for training and inference for object detection and segmentation. In this paper, an experiment was conducted to find out the compression efficiency and the effect of encoding on task inference performance when the features extracted in the intermediate stage of CNN are encoded. The feature map that combines the features of 256 channels into one image and the original image were encoded in HEVC to compare and analyze the inference performance for object detection and segmentation. Since the intermediate feature map encodes the five levels of feature maps (P2 to P6), the image size and resolution are increased compared to the original image. However, when the degree of compression is weakened, the use of feature maps yields similar or better inference results to the inference performance of the original image.

• Journal of the Korea Society of Computer and Information
• /
• v.20 no.12
• /
• pp.29-36
• /
• 2015

In this paper, we propose the horizontal line detection using the Haar-like features and linear regression in infrared images. In the marine environment horizon image is very useful information on a variety of systems. In the proposed method Haar-like features it was noted that the standard deviation be calculated in real time on a static area. Based on the pixel position, calculating the standard deviation of the around area in real time and, if the reaction is to filter out the largest pixel can get the energy map of the area containing the straight horizontal line. In order to select a horizontal line of pixels from the energy map, we applied the linear regression, calculating a linear fit to the transverse horizontal line across the image to select the candidate optimal horizontal. The proposed method was carried out in a horizontal line detecting real infrared image experiment for day and night, it was confirmed the excellent detection results than the legacy methods.

• Proceedings of the IEEK Conference
• /
• 2009.05a
• /
• pp.31-33
• /
• 2009

Face detection using AdaBoost algorithm is capable of processing images rapidly while having high detection rates. It seemed to be the fastest and the most robust and it is still today. Many improvements or extensions of this method have been proposed. However, previous approaches only deal with upright faces. They suffer from limited discriminant capability for rotated faces as these methods apply the same features for both upright and rotated faces. To solve this problem, it is necessary that we rotate input images or make independently trained detectors. However, this can be slow and can require a lot of training data, since each classifier requires the computation of many different image features. This paper proposes a robust algorithm for finding rotated faces within an image. It reduces the computational and sample complexity, by finding common features that can be shared across the classes. And it will be able to apply with multi-class object detection.