Eun Kyeong Kim

Map building of indoor environment using laser range finder and geometrics

This paper proposes map building method. Sensors such as a laser range finder, a gyroscope, encoders multiply compose the system. Generally, mobile robot can measure its relative position using a gyroscope and encoders in an environment. However, in this case, a large number of errors occur due to accumulative errors of sensor over time. Therefore, a map based on feature points is used. And an absolute position is measured by the feature points and geometrics. As combining the relative position and the absolute position, mobile robot can recognize its position. According to compositional data of a map, in case of a laser range finder, it takes a long time when adding a map or calculating a position of mobile robot. Therefore, it is necessary to arrange the feature points for computational time and map building. In this paper, a map was formed by laser range finder and features of geometrics. As a result of proposed method, the map was built efficiently in an aspect of time.

Fast Generative Approach Based on Sparse Representation for Visual Tracking

One of issue in generative approach for visual tracking is relates to computation time. It is because generative approach uses particle filter for modeling the motion and as a method to predict the state in the current frame. The system will be more accurate but slower computation if many particles are used. Recently, the combination between particle filter and sparse model is proposed to handle appearance variations and occlusion in visual tracking. Unfortunately, the issue about computation time still remains. This paper presents fast method for sparse generative approach in visual tracking. In this method, l1 minimization is used to calculate sparse coefficient vector for each candidate sample. Then, the maximum weighted is selected to represent the result. Based on simulations, our proposed method demonstrate good result in area under curve parameter and achieve four times faster than other methods with only use fifty particles.

Indoor SLAM application using geometric and ICP matching methods based on line features

Abstract This study presents an autonomous guided vehicle (AGV) with simultaneous localization and map building (SLAM) based on matching method, and extended Kalman filter SLAM. In general, the AGV is a large mobile robot that is used in transportation to carry cargoes, and it is guided using wired or wireless guidance systems. The guidance system based AGV accounts for a majority of robots in the mobile robot industry. However, in semiconductor factories, landmarks are unavailable; hence, the existing system has not been used in the mentioned environments. Therefore, the SLAM technology is applied in the environments, and can guide the AGV without landmarks. However, the accuracy of the SLAM can be low owing to measurement error of sensors and a cumulative calculation caused by localization sensors. Therefore, the accuracy is frequently assumed to be incorrect; moreover, the accuracy of the built map is low. In order to solve the problems, this study proposes the AGV with the SLAM based on matching methods; two matching method; geometric matching method and iterative closest point algorithm. The performance of the proposed method is compared with typical methods such as singular value decomposition / RIGID transformation based technologies using feature-point-based SLAM and is compared with the aforementioned two methods using the extended Kalman filter SLAM. The proposed method is more efficient than the typical methods used in the comparison.

Automatic brightness adjustment system by fuzzy inference system for object recognition

A camera has been widely used in practical fields with a diversity of purposes recently. There is a variety purpose of photography: images for memory, medical images for diagnosis, images for object recognition, surveillance images, and so on. In case of images for object recognition, the clarity of images is necessary to analyze the images which are obtained using vision sensors. However, a brightness of the image highly depends on the intensity of illumination in the certain environment. Therefore, we propose a method to solve the problems mentioned above by adjusting brightness automatically by utilizing CIE L∗a∗b∗ color space and fuzzy inference system. At first, the proposed method adjusts the brightness of a given image by considering both RGB component and L component of CIE L∗a∗b∗ color space. Secondly, the proposed method applies the fuzzy inference system to determine adjustment coefficients of each pixel for adjusting brightness of the image. Through the processes as mentioned above, we can obtain the result which is adjusted its brightness. To verify the proposed method, we compare the result image with two different images, a reference image, and an adjusted image by using offset. It is confirmed that the proposed method can adjust a given image efficiently and automatically.

Visual Tracking Based on Complementary Learners with Distractor Handling

The representation of the object is an important factor in building a robust visual object tracking algorithm. To resolve this problem, complementary learners that use color histogram- and correlation filter-based representation to represent the target object can be used since they each have advantages that can be exploited to compensate the other’s drawback in visual tracking. Further, a tracking algorithm can fail because of the distractor, even when complementary learners have been implemented for the target object representation. In this study, we show that, in order to handle the distractor, first the distractor must be detected by learning the responses from the color-histogram- and correlation-filter-based representation. Then, to determine the target location, we can decide whether the responses from each representation should be merged or only the response from the correlation filter should be used. This decision depends on the result obtained from the distractor detection process. Experiments were performed on the widely used VOT2014 and VOT2015 benchmark datasets. It was verified that our proposed method performs favorably as compared with several state-of-the-art visual tracking algorithms.

Multi-scale color features based on correlation filter for visual tracking

The change of appearance of the target object is one of important issue in visual tracking. It is because some factors such as camera motion, illumination change, motion change, occlusion, and size change are influenced to the object target during tracking. Recently, discriminative correlation filters (DCF) gave good results to handle these problems. Unfortunately, the DCF only works in the single-resolution features maps. In this paper, multi-scale color features are investigated to to solve this limitation. The multi-scale color features are implemented with correlation filter where it works in the continuous domain. In consequence of this reason, the implicit model of the target object is needed. So that, an interpolation is used to solve this problem. The output of this method is selected from the circular convolution response which has maximum value. Extensive experimental results on VOT2015 benchmark dataset which consists of 60 challenging videos show that the multi-scale color features based on correlation filter performs favorably against several state-of-the-art methods.

Performance Comparison of Probabilistic Methods Based Correction Algorithms for Localization of Autonomous Guided Vehicle

This paper presents performance comparison of probabilistic methods based correction algorithms for localization of AGV (Autonomous Guided Vehicle). Wireless guidance systems among the various guidance systems guides the AGV using position information from localization sensors. Laser navigation is mostly used to the AGV of a wireless type, however the performance of the laser navigation is influenced by a slow response time, big error of rotation driving and a disturbance with light and reflection. Therefore, the localization error of the laser navigation by the above-mentioned weakness has a great effect on the performance of the AGV. There are many different methods to correct the localization error, such as a method using a fuzzy inference system, a method with probabilistic method and so on. Bayes filter based estimation algorithms (Kalman Filter, Extended Kalman Filter, Unscented Kalman Filter and Particle Filter) are mostly used to correct the localization error of the AGV. This paper analyses performance of estimation algorithms with probabilistic method at localization of the AGV. Algorithms for comparison are Extended Kalman Filter, Unscented Kalman Filter and Particle Filter. Kalman Filter is excluded to the comparison, because Kalman Filter is applied only to a linear system. For the performance comparison, a fork-type AGV is used to the experiments. Variables of algorithms is set experiments based heuristic values, and then variables of same functions on algorithms is set same values.

Anomalous Propagation Echo Classification of Imbalanced Radar Data with Support Vector Machine

A number of technologically advanced devices, such as radars and satellites, are used in an actual weather forecasting process. Among these devices, the radar is essential equipment in this process because it has a wide observation area and fine resolution in both the time and the space domains. However, the radar can also observe unwanted nonweather phenomena. Anomalous propagation echo is one of the representative nonprecipitation echoes generated by an abnormal refraction phenomenon of a radar beam. Abnormal refraction occurs when the temperature and the humidity change dramatically. In such a case, the radar recognizes either the ground or the sea surface as an atmospheric object. This false observation decreases the accuracy of both quantitative precipitation estimation and weather forecasting. Therefore, a system that can automatically recognize an anomalous propagation echo from the radar data needs to be developed. In this paper, we propose a classification method for separating anomalous propagation echoes from the rest of the weather data by using a combination of a support vector machine classifier and the synthetic minority oversampling technique, to solve the problem of imbalanced data. By using actual cases of anomalous propagation we have confirmed that the proposed method provides good classification results.

Discrete Wavelet Transform for Watermarking Three-Dimensional Triangular Meshes from a Kinect Sensor

We present a simple method to watermark three-dimensional (3D) triangular meshes that have been generated from the depth data of the Kinect sensor. In contrast to previous methods, which maintain the shape of 3D triangular meshes and decide the embedding place, requiring calculations of vertices and their neighbors, our method is based on selecting one of the coordinate axes. To maintain shape, we use discrete wavelet transform and constant regularization. We know that the watermarking system needs the information to be embedded; we used a text to provide that information. We used geometry attacks such as rotation, scales, and translation, to test the performance of this watermarking system. Performance parameters in this paper include the vertices error rate (VER) and bit error rate (BER). The results from the VER and BER indicate that using a correction term before the extraction process makes our system robust to geometry attacks.

Image brightness adjustment system based on ANFIS by RGB and CIE L∗a∗b∗

This paper proposes the method to adjust brightness information by applying CIE L∗a∗b∗ color space and adaptive neuro-fuzzy inference system. The image which is already captured by vision sensor should be adjusted brightness to recognize objects in an image. In case of proper intensity of lights, the clarity of an image is good to recognize objects. However, in case of improper intensity of lights, the image has darkish regions. It will leads to reduce success of object recognition. To make up for this week point, we adjust the image, which is a darkish image, by controlling brightness information of an image. Brightness information can be represented by CIE L∗a∗b∗ color space. So based on CIE L∗a∗b∗ color space, adaptive neuro-fuzzy inference system is implemented as control function. Control function carries out adjusting of brightness information by dealing with the value of L component of CIE L∗a∗b∗ color space. L component describes brightness information of an image. The values which is calculated by adaptive neuro-fuzzy inference system is called the adjustment coefficient. Finally, the adjustment coefficient is added to L component for adjusting brightness information. To verify the propose method, we calculated color difference with respect to RGB and CIE L∗a∗b∗ color space. As experimental results, the propose method can reduce color difference and makes the target image will be similar with reference image under proper intensity of lights.