• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2022.10a
• /
• pp.256-257
• /
• 2022

In this paper, we introduce a method of extracting three-dimensional feature points through the movement of a single mobile device. Using a monocular camera, a 2D image is acquired according to the camera movement and a baseline is estimated. Perform stereo matching based on feature points. A feature point and a descriptor are acquired, and the feature point is matched. Using the matched feature points, the disparity is calculated and a depth value is generated. The 3D feature point is updated according to the camera movement. Finally, the feature point is reset at the time of scene change by using scene change detection. Through the above process, an average of 73.5% of additional storage space can be secured in the key point database. By applying the algorithm proposed to the depth ground truth value of the TUM Dataset and the RGB image, it was confirmed that the\re was an average distance difference of 26.88mm compared with the 3D feature point result.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.18 no.6
• /
• pp.1307-1312
• /
• 2023

In this paper, we propose a hybrid camera system that combines cameras with different focal lengths and LiDAR (Light Detection and Ranging) sensors to address the core components of autonomous driving perception technology, which include object recognition and distance measurement. We extract objects within the scene and generate precise location and distance information for these objects using the proposed hybrid camera system. Initially, we employ the YOLO7 algorithm, widely utilized in the field of autonomous driving due to its advantages of fast computation, high accuracy, and real-time processing, for object recognition within the scene. Subsequently, we use multi-focal cameras to create depth maps to generate object positions and distance information. To enhance distance accuracy, we integrate the 3D distance information obtained from LiDAR sensors with the generated depth maps. In this paper, we introduce not only an autonomous vehicle platform capable of more accurately perceiving its surroundings during operation based on the proposed hybrid camera system, but also provide precise 3D spatial location and distance information. We anticipate that this will improve the safety and efficiency of autonomous vehicles.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.46 no.1
• /
• pp.14-22
• /
• 2009

By calculating the differences between two images, which are captured with the same scene at different time, we can detect a set of outliers, such as occluding objects due to moving vehicles. To reduce the influence from the different intensity properties of the images, a simple technique that reruns the regression, which is based on the polynomial regression model, is employed. For a robust detection of outliers, the image difference is normalized by the noise variance. Hence, an accurate estimate of the noise variance is very important. In this paper, using an empirically obtained correction constant is proposed. Numerical analysis using both synthetic and real images are also shown in this paper to show the robust performance of the detection algorithm.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.9 no.2
• /
• pp.155-160
• /
• 2014

In this paper, an algorithm using RGB-depth camera is proposed to detect smoke in interrior. RGB-depth camera, the Kinect provides RGB color image and depth information. The Kinect sensor consists of an infra-red laser emitter, infra-red camera and an RGB camera. A specific pattern of speckles radiated from the laser source is projected onto the scene. This pattern is captured by the infra-red camera and is analyzed to get depth information. The distance of each speckle of the specific pattern is measured and the depth of object is estimated. As the depth of object is highly changed, the depth of object plain can not be determined by the Kinect. The depth of smoke can not be determined too because the density of smoke is changed with constant frequency and intensity of infra-red image is varied between each pixels. In this paper, a smoke detection algorithm using characteristics of the Kinect is proposed. The region that the depth information is not determined sets the candidate region of smoke. If the intensity of the candidate region of color image is larger than a threshold, the region is confirmed as smoke region. As results of simulations, it is shown that the proposed method is effective to detect smoke in interior.

• Journal of Internet Computing and Services
• /
• v.15 no.6
• /
• pp.117-124
• /
• 2014

Cluttered background is a major obstacle in developing salient object detection and tracking system for mobile device captured natural scene video frames. In this paper we propose a context aware feature vector selection model to provide an efficient noise filtering by machine learning based classifiers. Since the context awareness for feature selection is achieved by searching nearest neighborhoods, known as NP hard problem, we apply a fast approximation method with complexity analysis in details. Separability enhancement in feature vector space by adding the context aware feature subsets is studied rigorously using principal component analysis (PCA). Overall performance enhancement is quantified by the statistical measures in terms of the various machine learning models including MLP, SVM, Naïve Bayesian, CART. Summary of computational costs and performance enhancement is also presented.

• Journal of Digital Convergence
• /
• v.15 no.1
• /
• pp.239-245
• /
• 2017

For privacy protection, we propose a vehicle license plate region detection method in road view image served from portal site. Because vehicle license plate regions in road view images have different feature depending on distance, long distance vehicle license plate regions are not detected by feature of low resolution. Therefore, we suggest a method to detect short distance vehicle license plate regions by edge feature and long distance vehicle license plate regions using MSER feature. And then, we select candidate region of vehicle license plate region from detected region of each method, because the number of the vehicle license plate has a structural feature, we used it to detect the final vehicle license plate region. As the experiment result, we got a recall rate of 93%, precision rate of 75%, and F-Score rate of 80% in various road view images.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.48 no.1
• /
• pp.157-165
• /
• 2011

It is challenging to detect foreground objects when background includes an illumination variation, shadow or structural variation due to its motion. Basically pixel-based background models including codebook-based modeling suffer from statistical randomness of each pixel. This paper proposes an algorithm that incorporates Markov random field model into pixel-based background modeling to achieve more accurate foreground detection. Under the assumptions the distance between the pixel on the input imaging and the corresponding background model and the difference between the scene estimates of the spatio-temporally neighboring pixels are exponentially distributed, a recursive approach for estimating the MRF regularizing parameters is proposed. The proposed method alternates between estimating the parameters with the intermediate foreground detection and estimating the foreground detection with the estimated parameters, after computing it with random initial parameters. Extensive experiment is conducted with several videos recorded both indoors and outdoors to compare the proposed method with the standard codebook-based algorithm.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.49 no.2
• /
• pp.83-89
• /
• 2012

Natural Vegetation cover, which is very important earth resource, has been significantly altered by humans in some manner. Since this has currently resulted in a significant effect on global climate, various studies on vegetation environment including forest have been performed and the results are utilized in policy decision making. Remotely sensed data can detect, identify and map vegetation cover change based on the analysis of spectral characteristics and thus are vigorously utilized for monitoring vegetation resources. Among various vegetation indices extracted from spectral reponses of remotely sensed data, NDVI is the most popular index which provides a measure of how much photosynthetically active vegetation is present in the scene. In this study, for change detection in vegetation cover, a Multi-layer Perceptron Network (MLPN) as a nonparametric approach has been designed and applied to MODIS/Aqua vegetation indices 16-day L3 global 250m SIN Grid(v005) (MYD13Q1) data. The feature vector for change detection is constructed with the direct NDVI diffenrence at a pixel as well as the differences in some subset of NDVI series data. The research covered 5 years (2006-20110) over Korean peninsular.

• Korean Journal of Optics and Photonics
• /
• v.25 no.5
• /
• pp.245-253
• /
• 2014

We report a realistic field-performance simulation for a new MWIR camera. It is designed for early detection of missile plumes over a distance range of a few hundred kilometers. Both imaging and radiometric performance of the camera are studied by using real-scale integrated ray tracing, including targets, atmosphere, and background scene models. The simulation results demonstrate that the camera would satisfy the imaging and radiometric performance requirements for field operation.

• Journal of KIISE
• /
• v.42 no.1
• /
• pp.138-143
• /
• 2015

In this paper we develop a novel human action recognition system based on communication between an ear-mounted Bluetooth camera and an action recognition server to aid scene recognition for the blind. First, if the blind capture an image of a specific location using the ear-mounted camera, the captured image is transmitted to the recognition server using a smartphone that is synchronized with the camera. The recognition server sequentially performs human detection, object detection and action recognition by analyzing human poses. The recognized action information is retransmitted to the smartphone and the user can hear the action information through the text-to-speech (TTS). Experimental results using the proposed system showed a 60.7% action recognition performance on the test data captured in indoor and outdoor environments.