• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.8
• /
• pp.3567-3582
• /
• 2020

Human activities are often affected by weather conditions. Automatic weather recognition is meaningful to traffic alerting, driving assistance, and intelligent traffic. With the boost of deep learning and AI, deep convolutional neural networks (CNN) are utilized to identify weather situations. In this paper, a three-channel convolutional neural network (3C-CNN) model is proposed on the basis of ResNet50.The model extracts global weather features from the whole image through the ResNet50 branch, and extracts the sky and ground features from the top and bottom regions by two CNN5 branches. Then the global features and the local features are merged by the Concat function. Finally, the weather image is classified by Softmax classifier and the identification result is output. In addition, a medium-scale dataset containing 6,185 outdoor weather images named WeatherDataset-6 is established. 3C-CNN is used to train and test both on the Two-class Weather Images and WeatherDataset-6. The experimental results show that 3C-CNN achieves best on both datasets, with the average recognition accuracy up to 94.35% and 95.81% respectively, which is superior to other classic convolutional neural networks such as AlexNet, VGG16, and ResNet50. It is prospected that our method can also work well for images taken at night with further improvement.

• Journal of Korea Multimedia Society
• /
• v.5 no.2
• /
• pp.176-190
• /
• 2002

According as vehicles increases, system such as Advanced Drivers Assistance System(ADAS ) to inform forward situation to driver is required. In this paper, we proposes method to detect forward vehicles and lane from sequential color images by basis process to inform forward situation to driver. We detect a front vehicle using that shadow area exists on part under vehicles and that road area occupies many parts even if road traffic is confused. We detect lane information using that lane part is white order by reverse characteristic of shadow area. This method shows good result in case road is confused or there is direction indication to road. HSV color space is selected for color modeling. This method uses saturation component and value component in HSV color model to detect vehicles and lane. It uses statistics features of HSV component and position to know whether detected vehicles area is vehicles such as vehicles previous frame. To verify the effects of the proposed method, we capture the road images with notebook and CCD camera for PC and Present the results such as processing time, accuracy and vehicles detection against the images.

• Investigative Magnetic Resonance Imaging
• /
• v.4 no.2
• /
• pp.120-127
• /
• 2000

Adaptive template filtering has been proposed recently for enhancement of signal-to-noise ratio without loss of resolution. In the adaptive template filtering, an optimal template among multiple templates is selected, then linear least square error filtering based on the template is applied in vowel by vowel basis. In some magnetic resonance imaging, where the distribution of gray level has relatively small dynamic range, e.g., $T_1$ imaging, however, artificial stair-like artifact is observed at near edges. This is partially due to the edge enhancement effect in such yokels that contain multiple compounds at the boundaries of tissues. The gray levels of these yokels become similar gray levels of near dominant vowels that contain single compound by the adaptive filtering, which enlarges edge discontinuities. In this paper, we propose a technique to eliminate such artifact by identifying those yokels that contain multiple compounds and assigning the largest template for them. Filtered images with the proposed technique show substantial visual enhancement at the edges without degradation of peak signal-to-noise ratio compared to the original adaptive template filtering for both magnetic resonance images and phantom images.

• Proceedings of the KSRS Conference
• /
• 1998.09a
• /
• pp.76-82
• /
• 1998

Increase of greenhouse gas due to $CO_2$ and CH$_4$ gases would cause the global warming in the atmosphere. According to the global circulation model, it is pointed out in the Okhotsk Sea that the large increase of atmospheric temperature might be occurredin this region by global warming due to the doubling of greenhouse effectgases. Therefore, it is very important to monitor the sea ice extents in the Okhotsk Sea. To improve the sea ice extents and concentration with more highly accuracy, the field experiments have begun to comparewith Airborne Microwave Radiometer (AMR) and video images installed on the aircraft (Beach-200). The sea ice concentration is generally proportional to the brightness temperature and accurate retrieval of sea ice concentration from the brightness temperature is important because of the sensitivity of multi-channel data with the amount of open water in the sea ice pack. During the field experiments of airborned AMR the multi-frequency data suggest that the sea ice concentration is slightly dependending on the sea ice types since the brightness temperature is different between the thin and small piece of sea ice floes, and a large ice flow with different surface signatures. On the basis of classification of two sea ice types, it is cleary distinguished between the thin ice and the large ice floe in the scatter plot of 36.5 and 89.0GHz, but it does not become to make clear of the scatter plot of 18.7 and 36.5GHz Two algorithms that have been used for deriving sea ice concentrations from airbomed multi-channel data are compared. One is the NASA Team Algorithm and the other is the Bootstrap Algorithm. Intrercomparison on both algorithms with the airborned data and sea ice concentration derived from video images bas shown that the Bootstrap Algorithm is more consistent with the binary maps of video images.

• Journal of Fashion Business
• /
• v.7 no.6
• /
• pp.57-66
• /
• 2003

With the occurrence of interest in image-presentation that helps express one's individuality in several place in our society, it has affected make-up field. Now, make-up is getting to be recognized as a key role player that reflects our social and cultural concerns not only by presenting a appropriate one that is proper when to use and what for, but also presenting individualistic images that are able to express one's attractiveness free from a last plain concept that meant merely applying cosmetics to one's face is makeup. Like above, the reason that a change for consciousness of the function of make-up was possible is a noticeable outcome that results from the difference before and after make-up. In other word, with the help of Corrective Make-up that makes use of optical illusion, the analytic research for features could have been carried out systematically, therefore, women get to present their beauty of innate facial shape and features at their pleasure. The basic principles of the Corrective Make-up, which was introduced to satisfy the purpose of the existing make-up take advantage of lines and colors that are part of the optical illusion. The nature of direction and movement was expressed by the direction and angle of lines and the colors are also able to induce the variation of images depend on the whole color around them or contrastive color around. This thesis is mainly about optical illusion through lines that are part of elements that induce the optical illusion and deals with phenomena that occur when we apply lines that are part of design into the make-up. And this is written for the purpose of establishing the last basis of make-up more scientifically and systematically through the research for the variation when we apply the nature of lines into our features and deriving facial image variation from it. This is also for study into the role of make-up as for psychological effect that is able to make facial images look different when the optical illusion with possible visual errors is applied into it. In conclusion, we can say "Make-up is optical illusion through visual elements such as lines, surfaces, books, touch and so on."

• Journal of the Computational Structural Engineering Institute of Korea
• /
• v.31 no.4
• /
• pp.207-213
• /
• 2018

This paper investigates applicability of Microsoft $Kinect^{(R)}$, RGB-depth camera, to implement a 3D image and spatial information for sensing a target. The relationship between the image of the Kinect camera and the pixel coordinate system is formulated. The calibration of the camera provides the depth and RGB information of the target. The intrinsic parameters are calculated through a checker board experiment and focal length, principal point, and distortion coefficient are obtained. The extrinsic parameters regarding the relationship between the two Kinect cameras consist of rotational matrix and translational vector. The spatial images of 2D projection space are converted to a 3D images, resulting on spatial information on the basis of the depth and RGB information. The measurement is verified through comparison with the length and location of the 2D images of the target structure.

• Proceedings of the Korean Institute of Navigation and Port Research Conference
• /
• 2009.10a
• /
• pp.73-74
• /
• 2009

This study attempts to establish a system extracting and monitoring cultural grounds of seaweeds (lavers, brown seaweeds and seaweed fulvescens) and abalone on the basis of both KOMPSAT-2 and Terrasar-X data. The study areas are located in the northwest and southwest coast of South Korea, famous for coastal cultural grounds. The northwest site is in a high tidal range area (on the average, 6.1 m in Asan Bay) and has laver cultural grounds for the most. An semi-automatic detection system of laver facilities is described and assessed for spaceborne optic images. On the other hand, the southwest cost is most famous for seaweeds. Aquaculture facilities, which cover extensive portions of this area, can be subdivided into three major groups: brown seaweeds, capsosiphon fulvescens and abalone farms. The study is based on interpretation of optic and SAR satellite data and a detailed image analysis procedure is described here. On May 25 and June 2, 2008 the TerraSAR-X radar satellite took some images of the area. SAR data are unique for mapping those farms. In case of abalone farms, the backscatters from surrounding dykes allows for recognition and separation of abalone ponds from all other water-covered surfaces. But identification of seaweeds such as laver, brown seaweeds and seaweed fulvescens depends on the dampening effect due to the presence of the facilities and is a complex task because objects that resemble seaweeds frequently occur, particularly in low wind or tidal conditions. Lastly, fusion of SAR and optic spatial images is tested to enhance the detection of aquaculture facilities by using the panchromatic image with spatial resolution 1 meter and the corresponding multi-spectral, with spatial resolution 4 meters and 4 spectrum bands, from KOMPSAT-2. The mapping accuracy achieved for farms will be estimated and discussed after field verification of preliminary results.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.11 no.10
• /
• pp.1937-1943
• /
• 2007

Given the mountainous nature of Korea, cutting slopes are bound to happen. Every you, slope failures lead to enormous human and material losses. More recently, reckless development and subsequent degradation of forests has brought about soil erosion, which coupled with heavy rainfall, contaminates rivers, raises the level of river bottoms and thereby deteriorates their discharge capacity. In Korea, environmental impact assessments and disaster impact assessments have been conducted to come up with proper countermeasures. In order to perform quantitative analysis for this purpose, reliable slope information is quintessential. This study aims to obtain slope-related digital images using an RC model helicopter with a non-metric camera embedded, and to process these images to gain more accurate slope information. To this end, digital images obtained regarding cutting slopes were processed to gain numerical information of slopes and, on the basis of slope information gained here, reliable soil erosion factors were estimated.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.22 no.11
• /
• pp.1428-1434
• /
• 2018

Canine cataract is developed with aging and can cause the blindness or surgical treatment if not treated timely. In this paper, we propose a method for extracting cataract suspicious areas automatically with FCM(Fuzzy C_Means) algorithm to overcome the weakness of previously attempted ART2 based method. The proposed method applies the fuzzy stretching technique and the Max-Min based average binarization technique to the dog eye images photographed by simple devices such as mobile phones. After applying the FCM algorithm in quantization, we apply the brightness average binarization method in the quantized region. The two binarization images - Max-Min basis and brightness average binarization - are ANDed, and small noises are removed to extract the final cataract suspicious areas. In the experiment with 45 dog eye images with canine cataract, the proposed method shows better performance in correct extraction rate than the ART2 based method.

• Journal of the Korea Computer Graphics Society
• /
• v.28 no.3
• /
• pp.45-54
• /
• 2022

Image depth estimation is a technology that is the basis of various image analysis. As analysis methods using deep learning models emerge, studies using deep learning in image depth estimation are being actively conducted. Currently, most deep learning-based depth estimation models are being trained with clean and ideal images. However, due to the lack of data on adverse conditions such as haze or fog, the depth estimation may not work well in such an environment. It is hard to sufficiently secure an image in these environments, and in particular, obtaining non-homogeneous haze data is a very difficult problem. In order to solve this problem, in this study, we propose a method of synthesizing non-homogeneous haze images and a learning method for a monocular depth estimation deep learning model using this method. Considering that haze mainly occurs outdoors, datasets mainly containing outdoor images are constructed. Experiment results show that the model with the proposed method is good at estimating depth in both synthesized and real haze data.