電気学会論文誌C(電子・情報・システム部門誌)
Online ISSN : 1348-8155
Print ISSN : 0385-4221
ISSN-L : 0385-4221
130 巻, 9 号
選択された号の論文の28件中1~28を表示しています
特集:ビジョン技術によるイノベーション
巻頭言
特集論文
<システム・計測・制御>
  • 川西 亮輔, 山下 淳, 金子 透
    2010 年 130 巻 9 号 p. 1494-1503
    発行日: 2010/09/01
    公開日: 2010/09/01
    ジャーナル フリー
    Map information is important for path planning and self-localization when mobile robots execute autonomous tasks. In an unknown environment, mobile robots should measure the environment and construct its map by themselves. Then, we propose a modeling method of 3D environment. To realize wide-ranging environment measurement, we use an omnidirectional camera. We can measure environments efficiently by using the camera, because it has a 360-degree horizontal field of view. Our proposed method is based on structure from motion. A measurement method using feature points is effective in an environment including textured objects. Conversely, if environments mostly have non-textured objects, it is difficult to estimate camera movement and construct its environment model precisely. However, non-textured objects often have straight-line edges. Edge information is available for modeling of environments including non-textured objects. Against complex environments including both textured objects and non-textured objects, we should utilize both feature point and edge information for 3D environment modeling. Our proposed method constructs a 3D environment model by using both feature point and edge information. Experimental results show the effectiveness of our proposed method.
  • 松井 建樹, 山下 淳, 金子 透
    2010 年 130 巻 9 号 p. 1504-1512
    発行日: 2010/09/01
    公開日: 2010/09/01
    ジャーナル フリー
    A lot of plumbings such as gas pipes and water pipes exist in public utilities, factories, power plants and so on. The use of an omni-directional camera that can take images of 360° in surroundings at a time is effective for the pipe inspection. However, shape measurement is difficult only with the omni-directional camera. Therefore, in this paper, we propose a reconstruction method of piping shape by using an omni-directional camera and an omni-directional laser with a light section method and a structure from motion. Our method calculates 3-D coordinates by the light section method. The individual measurement data is integrated with the information of camera motion estimated by the structure from motion technique. We construct a triangular mesh by using 3-D Delaunay triangulation. By texture-mapping, a 3-D Pipe model is generated. The validity of the proposed method is shown through experiments.
<音声画像処理・認識>
  • 小野 友也, 岩井 儀雄, 石黒 浩
    2010 年 130 巻 9 号 p. 1513-1523
    発行日: 2010/09/01
    公開日: 2010/09/01
    ジャーナル フリー
    Recently, research fields of augmented reality and robot navigation are actively investigated. Estimating a relative posture between an object and a camera is an important task in these fields. In this paper, we propose a novel method for posture estimation by using high frequency markers and kernel regressions. The markers are embedded in an object's texture in the high frequency domain. We observe the change of spatial frequency of object's texture to estimate a current posture of the object. We conduct experiments to show the effectiveness of our method.
  • Atsushi Shimada, Rin-ichiro Taniguchi
    2010 年 130 巻 9 号 p. 1524-1529
    発行日: 2010/09/01
    公開日: 2010/09/01
    ジャーナル フリー
    Background modeling has been widely researched to detect moving objects from image sequences. It is necessary to adapt the background model various changes of illumination condition. Recent years, a hybrid type of background model which consists of more than one background model has been used for object detection since it is very adaptable to illumination changes. In this paper, we also propose a new hybrid type of background model named “Hybrid Spatial-Temporal Background Model”. Our model consists of two different kinds of background models. One is pixel-level background model which adapts to long-term illumination changes. The other is spatial-temporal background model which adapts to short-term illumination changes. Our experimental results demonstrate superiority of our method to some related works.
  • 柏木 利幸, 檜垣 真二, 宮脇 俊幸, 大恵 俊一郎
    2010 年 130 巻 9 号 p. 1530-1536
    発行日: 2010/09/01
    公開日: 2010/09/01
    ジャーナル フリー
    We propose a visualization method of uneven areas on single layer thin films formed on optical filters or semiconductor materials. High uniformity of the thickness of thin films is required to produce those devices, in spite of the difficulty of visual inspection. Our method visualizes uneven areas on the film by detecting intensity variances of light interference. To capture images, we utilize a fluorescent lamp with 3 lighting spectrum peaks and a high sensitive color line sensor camera. In the taken color image, uneven areas of thickness of thin films are observed as uneven color areas. Our method displays the uneven color areas based on the average vector of normal area pixels, which are selected by the original techniques to detect non-uniform color areas using color histogram. In our experiments to visualize uneven areas on a thin oxide layer on silicon wafers, we can discern the uneven area of which height has the difference of 2 nanometers from the normal area.
  • 堀江 勝大, 鹿嶋 雅之, 佐藤 公則, 渡邊 睦
    2010 年 130 巻 9 号 p. 1537-1545
    発行日: 2010/09/01
    公開日: 2010/09/01
    ジャーナル フリー
    The necessity of ultrasonic diagnosis tools increases every year. We propose an automatic endocardium tracing method by applying prepared “Standard Left Ventricles Shape Model (SLVSM)”. The cross section of heart wall in ultrasonic image is decided depending on the position and the angle of this probe. The initial contour is adaptively determined as crossing curve line between the SLVSM and the cross section. And the endocardium contour is extracted by active contour model(ACM) in two stages. In the first stage, an endocardium contour is detected using the result of an edge extraction based on the separability of image features. In the second stage, the endocardium contour is extracted using shape correction processing. “Mitral valve processing” not only detects the position of the mitral valve at the end diastolic period, but also corrects the detected contour after the first stage of ACM. Experimental results using one healthy case and three diseased cases have shown the effectiveness of the proposed method.
  • 重森 清史, 菊野 敏博, 井上 高宏
    2010 年 130 巻 9 号 p. 1546-1553
    発行日: 2010/09/01
    公開日: 2010/09/01
    ジャーナル フリー
    An automatic visual inspection method for an IC lead frame, which is effective to detect low contrast defects called stains and irregular luster, is desired. Low contrast defects are usually difficult to be distinguished clearly from the normal area by the difference of intensity level. Therefore, a simple subtraction technique using a good product as a reference image is not effective for detecting low contrast defects. This paper proposes a novel automatic visual inspection method effective to detect low contrast defects by using a spline function. In our method, a virtual good-product image is created for each product under the visual inspection by using a spline function. And the intensity subtraction technique is applied between a target image and a virtual good-product image. The intensity deviation of the virtual good-product image from the normal one can be made small enough to distinguish low contrast defects. Thus, the proposed method realizes an effective detection of low contrast defects and the reduction of false detection for the good product.
  • 池村 翔, 藤吉 弘亘
    2010 年 130 巻 9 号 p. 1554-1560
    発行日: 2010/09/01
    公開日: 2010/09/01
    ジャーナル フリー
    This paper presents a method for action classification by using Joint Boosting with depth information obtained by TOF camera. Our goal is to classify action of a customer who takes the goods from each of the upper, middle and lower shelf in the supermarkets and convenience stores. Our method detects of human region by using Pixel State Analysis (PSA) from the depth image stream obtained by TOF camera, and extracts the PSA features captured from human-motion and the depth features (peak value of depth) captured from the information of human-height. We employ Joint Boosting, which is a multi-class classification of boosting method, to perform the action classification. Since the proposed method employs spatiotemporal and depth feature, it is possible to perform the detection of action for taking the goods and the classification of the height of the shelf simultaneously. Experimental results show that our method using PSA feature and peak value of depth achieved a classification rate of 93.2%. It also had a 3.1% higher performance than that of the CHLAC feature, and 2.8% higher performance than that of the ST-patch feature.
  • 長原 一, 神吉 良典, 岩井 儀雄, 谷内田 正彦
    2010 年 130 巻 9 号 p. 1561-1571
    発行日: 2010/09/01
    公開日: 2010/09/01
    ジャーナル フリー
    A resolution of camera has been drastically improved under a current request for high-quality digital images. For example, digital still camera has several mega pixels. Although a video camera has the higher frame-rate, the resolution of a video camera is lower than that of still camera. Thus, the high-resolution is incompatible with the high frame rate of ordinary cameras in market. It is difficult to solve this problem by a single sensor, since it comes from physical limitation of the pixel transfer rate.
    In this paper, we propose a multi-sensor camera for capturing a resolution and frame-rate enhanced video. Common multi-CCDs camera, such as 3CCD color camera, has same CCD for capturing different spectral information. Our approach is to use different spatio-temporal resolution sensors in a single camera cabinet for capturing higher resolution and frame-rate information separately. We build a prototype camera which can capture high-resolution (2588×1958 pixels, 3.75 fps) and high frame-rate (500×500, 90 fps) videos. We also proposed the calibration method for the camera. As one of the application of the camera, we demonstrate an enhanced video (2128×1952 pixels, 90 fps) generated from the captured videos for showing the utility of the camera.
  • Dipankar Das, Yoshinori Kobayashi, Yoshinori Kuno
    2010 年 130 巻 9 号 p. 1572-1580
    発行日: 2010/09/01
    公開日: 2010/09/01
    ジャーナル フリー
    In this paper, we present an object detection technique that uses scale invariant local edgel structures and their properties to locate multiple object categories within a range image in the presence of partial occlusion, cluttered background, and significant scale changes. The fragmented local edgels (key-edgel, ek) are efficiently extracted from a 3D edge map by separating them at their corner points. The 3D edge maps are reliably constructed by combining both boundary and fold edges of 3D range images. Each key-edgel is described using our scale invariant descriptors that encode local geometric configuration by joining the edgel to adjacent edgels at its start and end points. Using key-edgels and their descriptors, our model generates promising hypothetical locations in the image. These hypotheses are then verified using more discriminative features. The discriminative feature consists of a bag-of-words histogram constructed by key-edgels and their descriptors, and a pyramid histogram of orientation gradients. To find the similarities between different feature types in a discriminative stage, we use an exponential χ2 merging kernel function. Our merging kernel outperforms the conventional rbf kernel of the SVM classifier. The approach is evaluated based on ten diverse object categories in a real-world environment.
  • 倉見 義幸, 伊藤 裕司, 名取 道也, 大関 和夫, 青木 義満
    2010 年 130 巻 9 号 p. 1581-1587
    発行日: 2010/09/01
    公開日: 2010/09/01
    ジャーナル フリー
    In recent years, development of neonatal care is strongly hoped, with increase of the low-birth-weight baby birth rate. Especially respiration of low-birth-weight baby is incertitude because central nerve and respiratory function is immature. Therefore, a low-birth-weight baby often causes a disease of respiration. In a NICU (Neonatal Intensive Care Unit), neonatal respiration is monitored using cardio-respiratory monitor and pulse oximeter at all times. These contact-type sensors can measure respiratory rate and SpO2 (Saturation of Peripheral Oxygen). However, because a contact-type sensor might damage the newborn's skin, it is a real burden to monitor neonatal respiration. Therefore, we developed the respiratory monitoring system for newborn using a FG (Fiber Grating) vision sensor. FG vision sensor is an active stereo vision sensor, it is possible for non-contact 3D measurement. A respiratory waveform is calculated by detecting the vertical motion of the thoracic and abdominal region with respiration. We attempted clinical experiment in the NICU, and confirmed the accuracy of the obtained respiratory waveform was high. Non-contact respiratory monitoring of newborn using a FG vision sensor enabled the minimally invasive procedure.
  • 圓田 直樹, 福井 真二, 倉橋 渉, 武知 啓介, 岩堀 祐之
    2010 年 130 巻 9 号 p. 1588-1596
    発行日: 2010/09/01
    公開日: 2010/09/01
    ジャーナル フリー
    This paper proposes an approach of robust tracking for crossing of similar pattern objects. It is based on the particle filter. As far as only appearance information of a target object is used for the particle filter, it fails in tracking the target object when it intersects other objects with similar patterns. The proposed method uses velocity information of the target object and distance information between the target object and other moving objects in addition to appearance information. The situation judges where the target object exists and how to calculate the likelihood of each particle is changed according to the situation. The proposed method can track targets efficiently and accurately by using the result of background subtraction. Moreover, the method is improved so that it can track objects even when they are hidden wholly by a background object. Results are demonstrated by experiments using real video sequences.
  • 藤原 孝幸, 渡邊 裕記, 輿水 大和, 上田 泰広, 重山 吉偉, 中村 淳良
    2010 年 130 巻 9 号 p. 1597-1603
    発行日: 2010/09/01
    公開日: 2010/09/01
    ジャーナル フリー
    Super-resolved imaging method based on OK Quantization Theory is proposed. Several methods have recently been reported for improvement of the resolution to the direction of the image space. On the other hand, we have been considering the resolution to the direction of the gray level. In this paper we propose super-resolved imaging method by using piecewise histogram equalization and show its experimental results.
  • 荒井 伸太郎, 井上 修, 小沢 慎治
    2010 年 130 巻 9 号 p. 1604-1613
    発行日: 2010/09/01
    公開日: 2010/09/01
    ジャーナル フリー
    This paper proposes a novel vehicle detection method. It uses an affine transform to determine the planar of features visible on the vehicle front surface. The feature points are detected on the image obtained from an embedded camera. In general, it is said that the feature points can perform stable vehicle detection since they are robust to weather and illumination changes. However, the feature points are also detected from artifacts which appear in the background image. For removing unnecessary feature points, we consider the vehicle front surface as planar and assume it follows an affine transform. We find that this affine transform assumption is valid, and the optical flow of the plane is used to determine whether it belongs to the background or to a following vehicle. Namely, our system can differentiate the front face of following vehicle and the background. Our experiments confirm that our system can locate and track the rear and side vehicles accurately and robustly.
  • 樋口 未来, 志磨 健, 村松 彰二, 入江 耕太, 門司 竜彦
    2010 年 130 巻 9 号 p. 1614-1621
    発行日: 2010/09/01
    公開日: 2010/09/01
    ジャーナル フリー
    This paper proposes a novel crosswalk detection technique for self-localization of automobile. The self-localization function based on the proposed technique can estimate highly accurate self-position by comparing results of image recognition with positions of crosswalks in map database. This paper focuses on a robust method to detect a crosswalk and its reference point which is used to calculate the distance between the self-position of host vehicle and the crosswalk. Our method can detect crosswalks and reference points from rear camera image sequences in real time. The previous road marking detection techniques hardly detect crosswalks with robustness because the rear camera images have some noise such as damages of road markings, halation, and shadows. Our method estimates the state including rough relative position of the crosswalk by Dynamic Bayesian Network in order to detect crosswalks and reference points robustly. The proposed method uses also the specification of crosswalk to reduce computational cost. The proposed method was tested on real images to confirm the accuracy and computational cost. The experimental results show that our method can detect crosswalks with a high degree of stability in real time.
  • 尾崎 竜史, 佐藤 雄隆, 岩田 健司, 坂上 勝彦
    2010 年 130 巻 9 号 p. 1622-1629
    発行日: 2010/09/01
    公開日: 2010/09/01
    ジャーナル フリー
    Based on the Statistical Reach Feature method, a image registration method which has the robustness for disturbances (e.g. illumination variation, noise) is proposed. The proposed method is based on the selection of point-pairs with stochastic consistency of sign of intensity difference. The robustness of the selected point-pairs is guaranteed from the statistical point of view. The detailed description of the proposed method is given, together with the statistical analysis. Also, the experimental results are given to show the effectiveness of the proposed method.
論文
<電子物性・デバイス>
  • 橋本 貴之, 川島 徹也, 白石 正樹, 秋山 登, 宇野 友影, 松浦 伸悌
    2010 年 130 巻 9 号 p. 1630-1635
    発行日: 2010/09/01
    公開日: 2010/09/01
    ジャーナル フリー
    This paper presents Cu-plate-bonded and capacitor-mounted SiPs for voltage regulators. The Cu-plate-bonded SiP reduces the power loss by 23% compared to those of a SiP with wire bonding. Copper plates reduce the spreading resistance of the topside electrodes in the MOSFETs, leading to lower power loss. The parasitic inductance of the capacitor-mounted SiP is reduced to 56% of that of the SiPs having the input capacitor mounted on the PCB. This reduction is due to the short current loop from the input capacitor to the MOSFETs. As a result, the power loss can be reduced by 20% for the same spike voltage. The high-side MOSFET die is flipped so that the drain electrode faces up, facilitating the connection of the drain electrode of the high-side MOSFET and the source electrode of the low-side MOSFET to the mounted input capacitor.
<メディア情報,ユーザ・インタフェース>
<音声画像処理・認識>
  • Amnach Khawne, Kazuhiko Hamamoto, Orachat Chitsobhuk
    2010 年 130 巻 9 号 p. 1644-1656
    発行日: 2010/09/01
    公開日: 2010/09/01
    ジャーナル フリー
    This paper proposes a digital image watermarking based on adaptive models of human visual perception. The algorithm exploits the local activities estimated from wavelet coefficients of each subband to adaptively control the luminance masking. The adaptive luminance is thus delicately combined with the contrast masking and edge detection and adopted as a visibility threshold. With the proposed combination of adaptive visual sensitivity parameters, the proposed perceptual model can be more appropriate to the different characteristics of various images. The weighting function is chosen such that the fidelity, imperceptibility and robustness could be preserved without making any perceptual difference to the image quality.
  • H. Kipsang Choge, Tadahiro Oyama, Stephen Karungaru, Satoru Tsuge, Min ...
    2010 年 130 巻 9 号 p. 1657-1666
    発行日: 2010/09/01
    公開日: 2010/09/01
    ジャーナル フリー
    Biometric applications based on the palmprint have recently attracted increased attention from various researchers. In this paper, a method is presented that differs from the commonly used global statistical and structural techniques by extracting and using local features instead. The middle palm area is extracted after preprocessing for rotation, position and illumination normalization. The segmented region of interest is then divided into blocks of either 8×8 or 16×16 pixels in size. The type-II Discrete Cosine Transform (DCT) is applied to transform the blocks into DCT space. A subset of coefficients that encode the low to medium frequency components is selected using the JPEG-style zigzag scanning method. Features from each block are subsequently concatenated into a compact feature vector and used in palmprint verification experiments with palmprints from the PolyU Palmprint Database. Results indicate that this approach achieves better results than many conventional transform-based methods, with an excellent recognition accuracy above 99% and an Equal Error Rate (EER) of less than 1.2% in palmprint verification.
<ソフトコンピューティング・学習>
  • 小澤 誠一, 岡本 圭介
    2010 年 130 巻 9 号 p. 1667-1673
    発行日: 2010/09/01
    公開日: 2010/09/01
    ジャーナル フリー
    To avoid the catastrophic interference in incremental learning, we have proposed Resource Allocating Network with Long Term Memory (RAN-LTM). In RAN-LTM, not only new training data but also some memory items stored in long-term memory are trained either by a gradient descent algorithm or by solving a linear regression problem. In the latter approach, radial basis function (RBF) centers are not trained but selected based on output errors when connection weights are updated. The proposed incremental learning algorithm belongs to the latter approach where the errors not only for a training data but also for several retrieved memory items and pseudo training data are minimized to suppress the catastrophic interference. The novelty of the proposed algorithm is that connection weights to be learned are restricted based on RBF activation in order to improve the efficiency in learning time and memory size. We evaluate the performance of the proposed algorithm in one-dimensional and multi-dimensional function approximation problems in terms of approximation accuracy, learning time, and average memory size. The experimental results demonstrate that the proposed algorithm can learn fast and have good performance with less memory size compared to memory-based learning methods.
<情報システム,エレクトロニック・コマース>
<情報処理・ソフトウェア>
研究開発レター
<音声画像処理・認識>
部門記事
feedback
Top