• IEMEK Journal of Embedded Systems and Applications
• /
• v.17 no.1
• /
• pp.41-49
• /
• 2022

This paper proposes a novel lane detection algorithm based on inverse perspective transformation and machine learning in lightweight embedded system. The inverse perspective transformation method is presented for obtaining a bird's-eye view of the scene from a perspective image to remove perspective effects. This method requires only the internal and external parameters of the camera without a homography matrix with 8 degrees of freedom (DoF) that maps the points in one image to the corresponding points in the other image. To improve the accuracy and speed of lane detection in complex road environments, machine learning algorithm that has passed the first classifier is used. Before using machine learning, we apply a meaningful first classifier to the lane detection to improve the detection speed. The first classifier is applied in the bird's-eye view image to determine lane regions. A lane region passed the first classifier is detected more accurately through machine learning. The system has been tested through the driving video of the vehicle in embedded system. The experimental results show that the proposed method works well in various road environments and meet the real-time requirements. As a result, its lane detection speed is about 3.85 times faster than edge-based lane detection, and its detection accuracy is better than edge-based lane detection.

• Journal of Embryo Transfer
• /
• v.32 no.3
• /
• pp.153-157
• /
• 2017

This study was conducted to investigate the optimal artificial insemination (AI) time with diagnostic kit at ovulation time. We already applied the patent about the protein in the cow heat mucose in external reproductive tract. And we would examine the accuracy for detection of cow heat by the kit produced with the protein. Evaluation of optimal heat detection was tried two time at 12 hrs and 24 hrs after the heat. And then, AI service also performed two times with no relation to the results of heat diagnosis by heat detection kit and pregnancy rates were checked with rectal palpation on $60^{th}$ day after AI. Heat diagnostic results by kit in natural heat after 12 hrs in Hanwoo cows were showed 31.3~75.0% on positive in first heat detection and 33.3~100.0% on positve in second heat detection. In the $1^{st}$ positive results were significant different (p<0.05), but $2^{nd}$ positive were not. The results of heat detection showed different result on regional influence and individual cow effects. The pregnancy rates of first trial of heat detection were showed 34.4~78.7% on positive and 21.3~68.8% on negative after the diagnosis by heat detection kit. And the pregnancy rates of next trial of heat detection were showed 33.3~85.7% on positive and 14.3~66.6% on negative after the heat diagnosis. Both positive results of first trial and next trial also were showed significant different (p<0.05), but negative results were not. In positive result, first trial of total pregnancy rates was higher than the next trial of pregnancy, but there showed opposite results on negative results. In conclusion, the optimal heat detection kit is suitable to ordinary Hanwoo cows and it suggested that we have to improve the kit's accuracy by detecting the materials like proteins related optimal AI time.

• Journal of Korea Multimedia Society
• /
• v.19 no.2
• /
• pp.180-188
• /
• 2016

In this paper, we propose a fire detection algorithm based on motion information and color information analysis. Conventional fire detection algorithms have as main problem the difficulty to detect fire due to external light, intensity, background image complexity, and little fire diffusion. So we propose a fire detection algorithm that accurate and fast. First, it analyzes the motion information in video data and then set the first candidate. Second, it determines this domain after analyzing the color and the domain. This algorithm assures a fast fire detection and a high accuracy compared with conventional fire detection algorithms. Our algorithm will be useful to real-time fire detection in real world.

• Proceedings of the KSR Conference
• /
• 2011.10a
• /
• pp.69-74
• /
• 2011

In the curving detection method by using an accelerometer, the ride comfort in the first car is worse than one in the others due to spend the time to calculate the tilting command and drive the tilting mechanism after entering in the curve. In order to enhance the ride comfort in the first car, the preditive curve detection method which predicts the distance from a train to the starting point of curve by using the GPS, Tachometer, Ground balise and position DB for track. In this study, we predicted and evaluated the ride comfort for predictive curve detection method in transient curves according to the shape and dimension of transient curve and the various driving speed. Also, we predicted the improvement of the ride comfort for predictive curve detection method by comparing with the result of the ride comfort for predictive curve detection method and for curve detection method using an accelerometer in the short transient curve.

• Journal of Korea Multimedia Society
• /
• v.7 no.12
• /
• pp.1692-1699
• /
• 2004

Many researchers have developed various techniques for detection of human faces in ordinary still images. Face detection is the first imperative step of human face recognition systems. The two main problems of human face detection are how to cutoff the running time and how to reduce the number of false positives. In this paper, we present frontal and near-frontal face detection algorithm in still gray images using a thick edge image and neural network. We have devised a new filter that gets the thick edge image. Our overall scheme for face detection consists of two main phases. In the first phase we describe how to create the thick edge image using the filter and search for face candidates using a whole face detector. It is very helpful in removing plenty of windows with non-faces. The second phase verifies for detecting human faces using component-based eye detectors and the whole face detector. The experimental results show that our algorithm can reduce the running time and the number of false positives.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.15 no.6
• /
• pp.141-146
• /
• 2015

In order to classify and analyze variously compounded sound and voice signal from FPGA microphone, there are numerous systems to detect abnormality signal, however, they have a lot of problems to implement the abnormality signal detection efficiently and effectively. Therefore, we proposed a method that implements classifying the signal effectively and outputting the detection efficiently based on the algorithm applied FIFO structure (First-in First-out) by using microphone sensor which able to input the sound signal, and statistical variance and coefficient of variation (CV). The result showed 96.3% detection when the experiment was performed more than 100 times with the proposed algorithm applied system.

• Journal of the Semiconductor & Display Technology
• /
• v.17 no.4
• /
• pp.105-108
• /
• 2018

Face Detection is an essential first step of the face recognition, and this is significant effects on face feature extraction and the effects of face recognition. Face detection has extensive research value and significance. In this paper, we present and analysis the principle, merits and demerits of the classic AdaBoost face detection and ASM algorithm based on point distribution model, which ASM solves the problems of face detection based on AdaBoost. First, the implemented scheme uses AdaBoost algorithm to detect original face from input images or video stream. Then, it uses ASM algorithm converges, which fit face region detected by AdaBoost to detect faces more accurately. Finally, it cuts out the specified size of the facial region on the basis of the positioning coordinates of eyes. The experimental result shows that the method can detect face rapidly and precisely, with a strong robustness.

• Proceedings of the KSRS Conference
• /
• 2003.11a
• /
• pp.461-463
• /
• 2003

The information on the cloud presence within a instantaneous field of view is the first step toward the derivation of many other geophysical parameters. Here, we first applied the current MODIS cloud detection algorithm developed by University of Wisconsin and compared the results to a visual interpretation of composite data, especially during the daytime. Most of cases, the detection algorithm performs very well, except a few cases with over-detection. One of the reasons for the false detection is due to the time independent use of land information which affects the threshold values of visible channel test. In the presentation, we show detailed analysis of the current cloud detection algorithm and suggest possible way to overcome the current shortfall.

• Communications for Statistical Applications and Methods
• /
• v.19 no.3
• /
• pp.333-344
• /
• 2012

We consider some statistical properties of the first order difference-based error variance estimator in nonparametric regression models with a single outlier. So far under an outlier(s) such difference-based estimators has been rarely discussed. We propose the first order difference-based estimator using the leave-one-out method to detect a single outlier and simulate the outlier detection in a nonparametric regression model with the single outlier. Moreover, the outlier detection works well. The results are promising even in nonparametric regression models with many outliers using some difference based estimators.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.50 no.7
• /
• pp.258-266
• /
• 2013

We propose adaptive detection algorithm to reduce a false alarm by considering the characteristics of the random noise on the detection system based on a seismic sensor. The proposed algorithm consists of the first step detection using kernel function and the second step detection using detection classes. Kernel function of the first step detection is obtained from the threshold of the Neyman-Pearon decision criterion using the probability density functions varied along the noise from the measured signal. The second step detector consists of 4 step detection class by calculating the occupancy time of the footstep using the first detected samples. In order to verify performance of the proposed algorithm, the detection of the footsteps using measured signal of targets (walking and running) are performed experimentally. The detection results are compared with a fixed threshold detector. The first step detection result has the high detection performance of 95% up to 10m area. Also, the false alarm probability is decreased from 40% to 20% when it is compared with the fixed threshold detector. By applying the detection class(second step detector), it is greatly reduced to less than 4%.