• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.19 no.5
• /
• pp.697-704
• /
• 2018

Since the face in image content corresponds to individual information that can distinguish a specific person from other people, it is important to accurately detect faces not hidden in an image. In this paper, we propose a method to accurately detect a face from input images using a deep learning algorithm, which is one of the machine learning methods. In the proposed method, image input via the red-green-blue (RGB) color model is first changed to the luminance-chroma: blue-chroma: red-chroma ($YC_bC_r$) color model; then, other regions are removed using the learned skin color model, and only the skin regions are segmented. A CNN model-based deep learning algorithm is then applied to robustly detect only the face region from the input image. Experimental results show that the proposed method more efficiently segments facial regions from input images. The proposed face area-detection method is expected to be useful in practical applications related to multimedia and shape recognition.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.26 no.11A
• /
• pp.1946-1956
• /
• 2001

In this paper, we propose a novel 2-D partial matching algorithm based on model-based stochastic analysis of feature correspondences in a relation vector space, which is quite robust to shape variations as well as invariant to geometric transformations. We represent an object using the ARG (Attributed Relational Graph) model with features of a set of relation vectors. In addition, we statistically model the partial occlusion or noise as the distortion of the relation vector distribution in the relation vector space. Our partial matching algorithm consists of two-phases. First, a finite number of candidate sets areselected by using logical constraint embedding local and structural consistency Second, the feature loss detection is done iteratively by error detection and voting scheme thorough the error analysis of relation vector space. Experimental results on real images demonstrate that the proposed algorithm is quite robust to noise and localize target objects correctly even inseverely noisy and occluded scenes.

• The KIPS Transactions:PartB
• /
• v.12B no.2 s.98
• /
• pp.123-130
• /
• 2005

In this paper, fingerprint identification algorithm using pixel direction factor in blocks is proposed to minimize false acceptance ratio and to apply security system. The proposed algorithm is that a fingerprint image is divided by 16 blocks, then feature parameters which have direct factors of $0^{\circ},\;45^{\circ},\;90^{\circ}\;and\;135^{\circ}$ is extracted for each block. Membership function of a reference fingerprint and an input fingerprint for the extracted parameters is calculated, then identification of two fingerprint is distinguished using fuzzy inference. False acceptance ratio is evaluated about different fingerprints of In kinds regardless of sex and shape which are obtained from adults, and false rejection ratio is evaluated about fingerprints which are obtained by adding fingerprints of 10 kinds on different fingerprints of 100 kinds. The experiment results is that false acceptance ratio is average $0.34\%$ about experiment of 4,950 times, and false rejection ratio is average $3.7\%$ about experiment of 1,000 times. The proposed algerian is excellent for recognition rate and security.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.51 no.3
• /
• pp.164-173
• /
• 2014

In underwater robotics, vision would be a key element for recognition in underwater environments. However, due to turbidity an underwater optical camera is rarely available. An underwater imaging sonar, as an alternative, delivers low quality sonar images which are not stable and accurate enough to find out natural objects by image processing. For this, artificial landmarks based on the characteristics of ultrasonic waves and their recognition method by a shape matrix transformation were proposed and were proven in Part 1. But, this is not working properly in undulating and dynamically noisy sea-bottom. To solve this, we propose a framework providing a selection phase of likelihood candidates, a selection phase for final candidates, recognition phase and tracking phase in sequence images, where a particle filter based selection mechanism to eliminate fake candidates and a mean shift based tracking algorithm are also proposed. All 4 steps are running in parallel and real-time processing. The proposed framework is flexible to add and to modify internal algorithms. A pool test and sea trial are carried out to prove the performance, and detail analysis of experimental results are done. Information is obtained from tracking phase such as relative distance, bearing will be expected to be used for control and navigation of underwater robots.

• Journal of Digital Convergence
• /
• v.13 no.2
• /
• pp.159-175
• /
• 2015

The two-dimensional QR code has advantages such as directional nature, enough data storage capacity, ability of error correction, and ability of data restoration. There are two major issues like speed and correctiveness of recognition in the two-dimensional QR code. Therefore, this paper proposes a morphology-based algorithm of detecting the interest region of a barcode. Our research contents can be summarized as follows. First, the interest region of a barcode image was detected by close operations in morphology. Second, after that, the boundary of the barcode are detected by intersecting four cross line outside in a code. Three, the projected image is then rectified into a two-dimensional barcode in a square shape by the reverse-perspective transform. In result, it shows that our detection and recognition rates for the barcode image is also 97.20% and 94.80%, respectively and that outperforms than previous methods in various illumination and distorted image environments.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2016.06a
• /
• pp.200-203
• /
• 2016

최근 자율주행자동차에 대한 관심이 증가함에 따라 교통 상황을 인식하는 방법에 대한 연구도 활발하게 진행되고 있다. 특히 교통신호등의 인식은 치명적인 결과를 야기하는 교통사고와 밀접하게 연관된다는 점에서 중요성이 더욱 부각되고 있다. 본 논문에서는 컴퓨터 비전 시스템을 기반으로 한 교통신호등 인식 방법을 제안한다. 차선, 표지판 등과는 다르게 교통신호등은 빛을 발하는 특징이 있으며 그 모양과 형태 또한 규격화 되어 있다. 이러한 특징 중 색상과 형태 특징을 이용하여 두 단계의 추출과정을 거쳐 교통신호등을 인식한다. 먼저 HSV 색 공간에서 적색, 녹색, 주황색의 빛을 발하는 영역을 찾아낸 뒤, 신호의 원형 특징을 이용해 가로, 세로 사이즈와 크기로 신호의 후보를 추출한다. 다음, 신호등의 검은 박스 영역을 찾기 위해 추출한 신호 후보군의 주변부가 검정색인지를 확인한다. 최종적으로 신호등의 박스 부분을 검출하여 신호를 발하는 위치를 기반으로 신호를 인식한다. 실험결과 많은 계산량을 요구하는 기계학습을 사용하지 않고도 실시간 처리와 높은 인식률로 교통 신호를 인식할 수 있음을 확인하였다.

• Proceedings of the Korean Society of Surveying, Geodesy, Photogrammetry, and Cartography Conference
• /
• 2006.04a
• /
• pp.469-473
• /
• 2006

In general, a classification process between ground data and non-ground data, which include building objects, is required prior to producing a DEM for a certain surface reconstruction from LiDAR data in which the DEM can be produced from the ground data, and certain objects like buildings can be reconstructed using non-ground data. Thus, an exact classification between ground and non-ground data from LiDAR data is the most important factor in the ground reconstruction process using LiDAR data. In particular, building objects can be largely used as digital maps, orthophotos, and urban planning regarding the object in the ground and become an essential to providing three dimensional information for certain urban areas. In this study, an entropy theory, which has been used as a standard of disorder or uncertainty for data used in the information theory, is used to apply a more objective and generalized method in the recognition and segmentation of buildings from raw LiDAR data. In particular, a method that directly uses the raw LiDAR data, which is a type of point shape vector data, without any changes, to a type of normal lattices was proposed, and the existing algorithm that segments LiDAR data into ground and non-ground data as a binarization manner was improved. In addition, this study proposes a generalized building extraction method that excludes precedent information for buildings and topographies and subsidiary materials, which have different data sources.

• Journal of the Korea Society of Computer and Information
• /
• v.3 no.3
• /
• pp.73-81
• /
• 1998

In this paper suggest for methods disease region feature extraction of medical image using wavelet. In the preprocessing, the shape informations of medical image are selected by performing the discrete wavelet transform(DWT) with four level coefficient matrix. In this approach, based on the characteristics of the coefficient matrix, 96 feature parameters are calculated as follows: Firstly. obtaining 32 feature parameters which have the characteristics of low frequency from the parameters according to the horizontal high frequency are calculated from the coefficient matrix of horizontal high frequency. In the third place, 16 vertical feature parameters are also calculated using the same kind of procedure with respect to the vertical high frequency. Finally, 32 feature parameters of diagonal high frequency are obtained from the coefficient matrix of diagonal high frequency. Consequently, 96 feature aprameters extracted. Using suggest algorithm in this paper will, implamentation can automatic recognition system, increasing efficiency of picture achieve communication system.

• Journal of Institute of Control, Robotics and Systems
• /
• v.20 no.9
• /
• pp.957-963
• /
• 2014

This paper proposes a predictive control for an efficient human following robot using Kinect sensor. Especially, this research is focused on detecting of foot-end-point and foot-vector instead of human body which can be occluded easily by the obstacles. Recognition of the foot-end-point by the Kinect sensor is reliable since the two feet images can be utilized, which increases the detection possibility of the human motion. Depth image features and a decision tree have been utilized to estimate the foot end-point precisely. A tracking point average algorithm is also adopted in this research to estimate the location of foot accurately. Using the continuous locations of foot, the human motion trajectory is estimated to guide the mobile robot along a smooth path to the human. It is verified through the experiments that detecting foot-end-point is more reliable and efficient than detecting the human body. Finally, the tracking performance of the mobile robot is demonstrated with a human motion along an 'L' shape course.

• Journal of Bio-Environment Control
• /
• v.1 no.1
• /
• pp.37-51
• /
• 1992

Quantizing and extracting visual features of mushroom(Lentinus edodes L.) are crucial to the sorting and grading automation, the growth state measurement, and the dried performance indexing. A computer image processing system was utilized for the extraction and measurement of visual features of front and back sides of the mushroom. The image processing system is composed of the IBM PC compatible 386DK, ITEX PCVISION Plus frame grabber, B/W CCD camera, VGA color graphic monitor, and image output RGB monitor. In this paper, an automatic thresholding algorithm was developed to yield the segmented binary image representing skin states of the front and back sides. An eight directional Freeman's chain coding was modified to solve the edge disconnectivity by gradually expanding the mask size of 3$\times$3 to 9$\times$9. A real scaled geometric quantity of the object was directly extracted from the 8-directional chain element. The external shape of the mushroom was analyzed and converted to the quantitative feature patterns. Efficient algorithms for the extraction of the selected feature patterns and the recognition of the front and back side were developed. The developed algorithms were coded in a menu driven way using MS_C language Ver.6.0, PC VISION PLUS library fuctions, and VGA graphic functions.