• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.38 no.3
• /
• pp.269-279
• /
• 2020

Traffic estimation mainly involves surveying equipment such as automatic vehicle classification, vehicle detection system, toll collection system, and personnel surveys through CCTV (Closed Circuit TeleVision), but this requires a lot of manpower and cost. In this study, we proposed a method of estimating traffic volume using deep learning and stereo CCTV to overcome the limitation of not detecting the entire vehicle in case of single CCTV. COCO (Common Objects in Context) dataset was used to train deep learning models to detect vehicles, and each vehicle was detected in left and right CCTV images in real time. Then, the vehicle that could not be detected from each image was additionally detected by using affine transformation to improve the accuracy of traffic volume. Experiments were conducted separately for the normal road environment and the case of weather conditions with fog. In the normal road environment, vehicle detection improved by 6.75% and 5.92% in left and right images, respectively, than in a single CCTV image. In addition, in the foggy road environment, vehicle detection was improved by 10.79% and 12.88% in the left and right images, respectively.

• Smart Media Journal
• /
• v.5 no.3
• /
• pp.54-61
• /
• 2016

According to many smart devices are development, SNS(Social Network Service) users are getting higher that is possible for real-time communicating, information sharing without limitations in distance and space. Nowadays, SNS users that based on communication and relationships, are getting uses SNS for information sharing. In this paper, we used the SNS posts for users to extract the category and information provider, how to following of recommend method. Particularly, this paper focuses on classifying the words in the text of the posts and measures the frequency using Inception-v3 model, which is one of the machine learning technique -CNN(Convolutional Neural Network) we classified image word. By classifying the category of a word in a text and image, that based on DMOZ to build the information provider DB. Comparing user categories classified in categories and posts from information provider DB. If the category is matched by measuring the degree of similarity to the information providers is classified in the category, we suggest that how to recommend method of the most similar information providers account.

• Journal of Korea Multimedia Society
• /
• v.6 no.1
• /
• pp.89-96
• /
• 2003

This paper propose the new method witch is adopted in extracting character features and recognizing numeric characters using fractal dimension and modified Henon Attractor of the Chaos Theory. Firstly, it gets features of mesh feature, projection feature and cross distance feature from numeric character images And their feature hi converted into time series data. Then using the modified Henon system suggested in this paper, it gets last features of numeric character image after calculating Natural Measure and information bit which art meant fractal dimension. Finally, numeric character recognition is performed by statistically finding out the each information bit showing the minimum difference against the normalized pattern database. An Experimental result shows 100% character classification rates for 10 digits and 90% of recognition rates in real situation and the recognition speed was 26 characters per second.

• Journal of the Optical Society of Korea
• /
• v.14 no.4
• /
• pp.376-382
• /
• 2010

We propose a novel method to precisely detect car license plates by locating main characters, which are printed with large font size. The regions of the main characters are directly detected without detecting the plate region boundaries, so that license regions can be detected more precisely than by other existing methods. To generate a binary image, multiple thresholds are applied, and segmented regions are selected from multiple binarized images by a criterion of size and compactness. We do not employ any character matching methods, so that many candidates for main character groups are detected; thus, we use a neural network to reject non-main character groups from the candidates. The relation of the character regions and the intensity statistics are used as the input to the neural network for classification. The detection performance has been investigated on real images captured under various illumination conditions for 1000 vehicles. 980 plates were correctly detected, and almost all non-detected plates were so stained that their characters could not be isolated for character recognition. In addition, the processing time is fast enough for a commercial automatic license plate recognition system. Therefore, the proposed method can be used for recognition systems with high performance and fast processing.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.18 no.7
• /
• pp.350-356
• /
• 2017

With increasing attention to real-time body detection, active research is being conducted on human body detection based on skin color. Despite this, most existing skin detection methods utilize static skin color models and have detection rates in images, in which colors are distorted. This study proposed a method of detecting the skin region using a fuzzy classification of the gradient map, saturation, and Cb and Cr in the YCbCr space. The proposed method, first, creates a gradient map, followed by a saturation map, CbCR map, fuzzy classification, and skin region binarization in that order. The focus of this method is to rigorously detect human skin regardless of the lighting, race, age, and individual differences, using features other than color. On the other hand,the borders between these features and non-skin regions are unclear. To solve this problem, the membership functions were defined by analyzing the relationship between the gradient, saturation, and color features and generate 108 fuzzy rules. The detection accuracy of the proposed method was 86.35%, which is 2~5% better than the conventional method.

• Journal of the Korean Society of Environmental Restoration Technology
• /
• v.23 no.4
• /
• pp.13-30
• /
• 2020

This study aims to analyze the qualitative and quantitative effects of environmental policies in terms of green space management using logistic regression model(LRM). Landsat satellite imageries in 1985, 1992, 2000, 2008, and 2015 are classified using a hybrid-classification method. Based on these classified maps, logistic regression model having a deforestation tendency of the past is built. Binary green space change map is used for the dependent variable and four explanatory variables are used: distance from green space, distance from settlements, elevation, and slope. The green space map of 2008 and 2015 is predicted using the constructed model. The conservation effect of Ulsan's environmental policies is quantified through the numerical comparison of green area between the predicted and real data. Time-series analysis of green space showed that restoration and destruction of green space are highly related to human activities rather than natural land transition. The effect of green space management policy was spatially-explicit and brought a significant increase in green space. Furthermore, as a result of quantitative analysis, Ulsan's environmental policy had effects of conserving and restoring 111.75㎢ and 175.45㎢ respectively for the periods of eight and fifteen years. Among four variables, slope was the most determinant factor that accounts for the destruction of green space in the city. This study presents logistic regression model as a way of evaluating the effect of environmental policies that have been practiced in the city. It has its significance in that it allows us a comprehensive understanding of the effect by considering every direct and indirect effect from other domains, such as air and water, on green space. We conclude discussing practicability of implementing environmental policy in terms of green space management with the focus on a non-statutory plan.

• Journal of the Korea Convergence Society
• /
• v.12 no.9
• /
• pp.21-30
• /
• 2021

Unlike behaviors that are generally known, information regarding unspecific behaviors is insufficient. For an education or guidance regarding the unspecific behaviors, collection and management of data regarding the unspecific behaviors of special-needs students are needed. In this paper, a consultation management model based on behavior classification of special-needs students using machine learning is proposed. It collects data by photographing the behavior of special students in real time, analyzes the behavior pattern, composes a data set, and trains it in the suggestion system. It is possible to improve the accuracy by comparing the behavior of special students photographed later into the suggestion system and analyzing the results by comparing it with the existing data again. The test has been performed by arbitrarily applying unspecific behaviors that are not stored in the database, and the forecast model has accurately classified and grouped the input data. Also, it has been verified that it is possible to accurately distinguish and classify the behaviors through the feature data of the behaviors even if there are some errors in the input process.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.13 no.2
• /
• pp.751-770
• /
• 2019

Action recognition is an essential task in computer vision due to the variety of prospective applications, such as security surveillance, machine learning, and human-computer interaction. The availability of more video data than ever before and the lofty performance of deep convolutional neural networks also make it essential for action recognition in video. Unfortunately, limited crafted video features and the scarcity of benchmark datasets make it challenging to address the multi-person action recognition task in video data. In this work, we propose a deep convolutional neural network-based Effective Hybrid Learning (EHL) framework for two-person interaction classification in video data. Our approach exploits a pre-trained network model (the VGG16 from the University of Oxford Visual Geometry Group) and extends the Faster R-CNN (region-based convolutional neural network a state-of-the-art detector for image classification). We broaden a semi-supervised learning method combined with an active learning method to improve overall performance. Numerous types of two-person interactions exist in the real world, which makes this a challenging task. In our experiment, we consider a limited number of actions, such as hugging, fighting, linking arms, talking, and kidnapping in two environment such simple and complex. We show that our trained model with an active semi-supervised learning architecture gradually improves the performance. In a simple environment using an Intelligent Technology Laboratory (ITLab) dataset from Inha University, performance increased to 95.6% accuracy, and in a complex environment, performance reached 81% accuracy. Our method reduces data-labeling time, compared to supervised learning methods, for the ITLab dataset. We also conduct extensive experiment on Human Action Recognition benchmarks such as UT-Interaction dataset, HMDB51 dataset and obtain better performance than state-of-the-art approaches.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.16 no.6
• /
• pp.1329-1338
• /
• 2021

There are rarely effective methods to help visually impaired people when they want to know the expiration date of products excepted to only Braille. In this study, we developed an expiration date notification system based on YOLO and OCR for visually impaired people. The handicapped people can automatically know the expiration date of a specific product by using our system without the help of a caregiver, fast and accurately. The proposed system is worked by four different steps: (1) identification of a target product by scanning its barcode; (2) segmentation of an image area with the expiration date using YOLO; (3) classification of the expiration date by OCR: (4) notification of the expiration date by TTS. Our system showed an average classification accuracy of about 86.00% when blindfolded subjects used the proposed system in real-time. This result validates that the proposed system can be potentially used for visually impaired people.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.13 no.5
• /
• pp.323-330
• /
• 2020

This paper proposes the road roughness based Braking Pressure Calculation System(BPCS) for an Autonomous Vehicle Stability. The system consists of an image normalization module that processes the front image of a vehicle to fit the input of the random forest, a Random Forest based Road Roughness Classification Module that distinguish the roughness of the road on which the vehicle is travelling by using the weather information and the front image of a vehicle as an input, and a brake pressure control module that modifies a friction coefficient applied to the vehicle according to the road roughness and determines the braking strength to maintain optimal driving according to a vehicle ahead. To verify the efficiency of the BPCS experiment was conducted with a random forest model. The result of the experiment shows that the accuracy of the random forest model was about 2% higher than that of the SVM, and that 7 features should be bagged to make an accurate random forest model. Therefore, the BPCS satisfies both real-time and accuracy in situations where the vehicle needs to brake.