• KIPS Transactions on Software and Data Engineering
• /
• v.4 no.11
• /
• pp.521-528
• /
• 2015

Road-view object classification methods are mostly influenced by weather and illumination conditions, thus the most of the research activities are based on dataset in clean weathers. In this paper, we present a road-view object classification method based on color segmentation that works for all kinds of weathers. The proposed method first classifies the weather and illumination conditions and then applies the weather-specified color models to find the road traffic signs. Using 5 different features of the road-view images, we classify the weather and light conditions as sunny, cloudy, rainy, night, and backlight. Based on the classified weather and illuminations, our model selects the weather-specific color ranges to generate Gaussian Mixture Model for each colors, Green, Yellow, and Blue. The proposed method successfully detects the traffic signs regardless of the weather and illumination conditions.

• Proceedings of the Korean Information Science Society Conference
• /
• 2007.10a
• /
• pp.229-230
• /
• 2007

Lane detection is one of the major components of traffic intelligence. It is impossible to recognize lanes as human do in all kinds of special situations; however, we can try to solve special problems with special methods. In this paper we propose a simple method using color segmentation, the Probabilistic Hough Transform (PHT), and the Least-Square in real-time lane tracking. Vehicles in neighborhood can be eliminated with one simple threshold in segmentation. Meanwhile, broken shape lanes in different road conditions can be successfully detected using the combination of PHT and Least-Square method. Eventually, this method is tested with groups of static images downloaded from internet and video sequences shot randomly on some highways. Satisfactory results are received.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2022.11a
• /
• pp.402-404
• /
• 2022

Recent research has shown that deep learning models are vulnerable to adversarial attacks not only in the digital but also in the physical domain. This becomes very critical for applications that have a very high safety concern, such as self-driving cars. In this study, we propose a physical adversarial attack technique for one of the common tasks in self-driving cars, namely segmentation of the urban scene. Our method can create a texture on a wall so that it can be misclassified as a road. The demonstration of the technique on a state-of-the-art cityscape pretrained model shows a fairly high success rate, which should raise awareness of more potential attacks in self-driving cars.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.23 no.1
• /
• pp.31-38
• /
• 2005

Importance of road that is country base equipment is occupying fair part. Therefore, establishment of road and maintenance expense for road management are increasing continuously. These problem can manage efficiently through data model construction that take advantage of framework data. But, because of difference of method of study in research institution, framework data research was constructed being overlapped until current. This is because framework data research was no access of application side. Therefore, National Geographic Information Institute presented subject framework data model guide through framework data model standardization business. This research constructed road management data model that take advantage of traffic framework data. Therefore, we can check equal data construction and reduce expense accordingly. Also, because there are not data model development instances by framework data model, it is difficult that judge whether is suitable to apply framework data model guide. Hence, in this study, the extended road management data medel and the suitability of framework data is presented.

• Journal of KIISE:Computing Practices and Letters
• /
• v.5 no.6
• /
• pp.748-757
• /
• 1999

본 논문에서는 차량 영역의 추출을 이용한 효율적인 교통 혼잡도 측정 시스템을 설계하고 구현한다. 차량 영역 정보의 추출은 첫째 영역 분할, 둘째 작은 영역의 제거와 영역의 직사각형화, 셋째 영역의 병합 및 삭제의 단계로 나눌 수 있다. 영역 분할 단계에서는 획득한 도로 영상을 영역 기반 영역 분할에 의해 영역으로 분할한다. 그 다음 영역 분할 후의 영역 정보 중 차량 영역을 추출하는데 영향을 미치지 않는 작은 영역들을 제거하고, 남은 영역들을 직사각형화한다. 마지막으로 차선 별로 남은 영역들을 병합, 삭제함으로써 각 차선마다 차량 영역 정보를 추출할 수 있다. 이러한 방법은 배경 영상과 같은 부가적인 정보를 사용하지 않고 도로 자체 영상만으로 교통 혼잡도를 측정할 수 있으며, 그림자의 영향이 없을 경우 적용할 수 있는 기법이다.Abstract In this paper, we designed and implemented an efficient road congestion analysis system using regional information. To extract vehicle regions from a road image, the system process the image in five steps: segmentation, small region elimination, region rectangularization, region merging and region deletion. First, we segment road image by a threshold value. Then, we eliminate useless small regions to extract vehicle region, and perform region rectangularization. Finally, we extract vehicle region of each lane of the road by region merging and deletion. This method has the advantage of measuring road congestion without additional information such as background images. But this method must be applied to road images without shadow.

• Journal of the Korea Society of Computer and Information
• /
• v.27 no.2
• /
• pp.33-42
• /
• 2022

In this paper, we proposed a deep learning model based on multi-scale dilated convolution feature fusion for the segmentation of black ice region in road image to send black ice warning to drivers in real time. In the proposed multi-scale dilated convolution feature fusion network, different dilated ratio convolutions are connected in parallel in the encoder blocks, and different dilated ratios are used in different resolution feature maps, and multi-layer feature information are fused together. The multi-scale dilated convolution feature fusion improves the performance by diversifying and expending the receptive field of the network and by preserving detailed space information and enhancing the effectiveness of diated convolutions. The performance of the proposed network model was gradually improved with the increase of the number of dilated convolution branch. The mIoU value of the proposed method is 96.46%, which was higher than the existing networks such as U-Net, FCN, PSPNet, ENet, LinkNet. The parameter was 1,858K, which was 6 times smaller than the existing LinkNet model. From the experimental results of Jetson Nano, the FPS of the proposed method was 3.63, which can realize segmentation of black ice field in real time.

• 제어로봇시스템학회:학술대회논문집
• /
• 2005.06a
• /
• pp.2393-2396
• /
• 2005

We present a method to detect on-road vehicle using geometric invariant of feature points on side planes of the vehicle. The vehicles are assumed into a set of planes and the invariant from motion information of features on the plane segments the plane from the theory that a geometric invariant value defined by five points on a plane is preserved under a projective transform. Harris corners as a salient image point are used to give motion information with the normalized correlation centered at these points. We define a probabilistic criterion to test the similarity of invariant values between sequential frames. Experimental results using images of real road scenes are presented.

• Korean Journal of Remote Sensing
• /
• v.22 no.2
• /
• pp.123-129
• /
• 2006

High-resolution aerial color image offers great possibilities for geometric and semantic information for spatial data generation. However, shadow casts by buildings and trees in high-density urban areas obscure much of the information in the image giving rise to potentially inaccurate classification and inexact feature extraction. Though many researches have been implemented for solving shadow casts, few studies have been carried out about the extraction of features hindered by shadows from aerial color images in urban areas. This paper presents a asphalt road boundary extraction technique that combines information from aerial color image and LIDAR (LIght Detection And Ranging) data. The following steps have been performed to remove shadow effects and to extract road boundary from the image. First, the shadow regions of the aerial color image are precisely located using LEAR DSM (Digital Surface Model) and solar positions. Second, shadow regions assumed as road are corrected by shadow path reconstruction algorithms. After that, asphalt road boundary extraction is implemented by segmentation and edge detection. Finally, asphalt road boundary lines are extracted as vector data by vectorization technique. The experimental results showed that this approach was effective and great potential advantages.

• Spatial Information Research
• /
• v.14 no.2 s.37
• /
• pp.245-260
• /
• 2006

A practical use of the location information and user requirement are increased in a mobile environment which supports the portability. And Various service which GIS is related with a Spatial DB have been processed. Generally, logical relation of a traffic network which organizes the Road DB uses a basic node-link structure. In this way, Designed structure can not be flexible at various model apply and are not efficient with a database retrieval in a maintenance management side. In this research, We supplement with the problem of a existing network model and the limitation of the building through the design of a network model which uses dynamic segmentation. And we tried to implement efficient hierarchy model at the retrieval of the network and presentation. Designed model supports a stage presentation of various level and a hierarchy entity relation and We are expected to supplement a network spatial modelling function which the GIS has.

• Journal of the Korea Convergence Society
• /
• v.12 no.1
• /
• pp.29-35
• /
• 2021

License plate recognition plays a key role in intelligent transportation systems. Therefore, it is a very important process to efficiently detect the number and character areas. In this paper, we propose a method to effectively detect license plate number area by applying deep learning and semantic image segmentation algorithm. The proposed method is an algorithm that detects number and text areas directly from the license plate without preprocessing such as pixel projection. The license plate image was acquired from a fixed camera installed on the road, and was used in various real situations taking into account both weather and lighting changes. The input images was normalized to reduce the color change, and the deep learning neural networks used in the experiment were Vgg16, Vgg19, ResNet18, and ResNet50. To examine the performance of the proposed method, we experimented with 500 license plate images. 300 sheets were used for learning and 200 sheets were used for testing. As a result of computer simulation, it was the best when using ResNet50, and 95.77% accuracy was obtained.