Shigeharu Miyata

Improvement of adaptive cruise control performance

This paper describes the Adaptive Cruise Control system (ACC), a system which reduces the driving burden on the driver. The ACC system primarily supports four driving modes on the road and controls the acceleration and deceleration of the vehicle in order to maintain a set speed or to avoid a crash. This paper proposes more accurate methods of detecting the preceding vehicle by radar while cornering, with consideration for the vehicle sideslip angle, and also of controlling the distance between vehicles. By making full use of the proposed identification logic for preceding vehicles and path estimation logic, an improvement in driving stability was achieved.

Traffic Sign Recognition Utilizing an Eigen Space Method Based on the KL Transform

This study explains that a method utilizing the eigen spaces obtained by the KL transform for automatic recognition by camera of the speed on a speed limit sign has the following advantages: it is robust in response to changes in intensity patterns caused by the direction the sign is facing and by the amount of light striking the sign, and it is able to reduce the recognition processing time by reducing the number of feature vector dimensions during analysis. The method for recognition of traffic signs previously proposed by the authors of this study was a method for recognition based on extracting geometric shapes from the sign and recognizing them based on their aspect ratios. As such, this method was not able to identify the numbers on a speed limit sign, all of which have identical aspect ratios. It will be shown that the method in this study is able to recognize nearly all speed limits indicated on traffic signs within several 100s of ms after image acquisition. This method was applied to still images and its effectiveness was verified from the perspective of the following requirements for providing accurate information concerning the vehicle surroundings to the driver: high processing speed, high recognition accuracy, detection of all detectable objects without omission, and robustness in response to changes in the surrounding environment.

Automatic Path Search for Roving Robot Using Reinforcement Learning

Rapid advances in robot technology have been made in recent years. In connection with these advances, robots are expected to be utilized in a variety of places and environments. This study describes 1) a method which allows a robot to measure the location of its destination in the real world based on an image obtained from a single camera, and 2) a method of navigating a robot to a destination which is selected by a user on a display showing the forward robot view. Consideration is also given to cases in which there are obstacles between the robot and the destination. Through the use of reinforcement learning, which is considered a promising candidate among autonomous control techniques, the roving robot tries to find the shortest way to the destination based on information concerning the locations of obstacles and the destination. This study also describes an image-based method of measuring a selected location, the results from a simulation of path finding using reinforcement learning, and the results from an experiment of navigation in a real environment. Finally, a summary of the main conclusions is provided.

Automated license plate detection using a support vector machine

This paper proposes a new method of detecting license plates in images of vehicles where the license plate is shown, and reports the detection results when this method was applied to detection of license plates on vehicles in Japan. This license plate detection process detects only the edge vertical components, and the candidate license plates are narrowed down using the contours obtained by dilation and erosion processing and region fill processing. A SVM (Support Vector Machine) based on negative and positive examples is used to determine whether or not a candidate area is a license plate, and finally the position of the license plate is identified. This study examined how the license plate detection results in license plate and non-license plate images were affected by differences in aspect ratios, differences in brightness between the vehicle body and license plate, and the number of positive and negative examples used for learning. The effectiveness of this method was confirmed to yield a license plate detection rate of approximately 90%.

Method for recognition of numbers on speed limit signs utilizing an eigen space method based on the KL transform

This study explains that a method utilizing the eigen spaces obtained by the KL transform for automatic recognition by camera of the speed on a speed limit sign has the following advantages: it is robust in response to changes in intensity patterns caused by the direction the sign is facing and by the amount of light striking the sign, and it is able to reduce the recognition processing time by reducing the number of feature vector dimensions during analysis. The method for recognition of traffic signs previously proposed by the authors of this study was a method for recognition based on extracting geometric shapes from the sign and recognizing them based on their aspect ratios. As such, this method was not able to identify the numbers on a speed limit sign, all of which have identical aspect ratios. It will be shown that the method in this study is able to recognize nearly all speed limits indicated on traffic signs within several 100s of ms after image acquisition. This method was applied to still images and its effectiveness was verified from the perspective of the following requirements for providing accurate information concerning the vehicle surroundings to the driver: high processing speed, high recognition accuracy, detection of all detectable objects without omission, and robustness in response to changes in the surrounding environment and to geometric changes in the sign image as the vehicle approaches it.

Evaluation of Kinect vision sensor for bin-picking applications: Improved component separation accuracy with combined use of depth map and color image

This report describes a problem involved with use of Kinect depth maps for robot picking of randomly stacked components, and also a solution to this problems. When Kinect is installed above stacked parts and processing is performed using only the obtained Kinect depth map information, there are cases when individual small metal components cannot be separately identified. So that the robot can reliably pick up a single component in these cases, this report demonstrates that in areas where the system incorrectly identifies multiple components as a single component, the addition of color image information and blob analysis of the color image results in accurate separation of the individual components, allowing a single item to be identified for picking.

Navigation and path search for roving robot using reinforcement learning

The robot technology is rapidly developing in recent years. In connection with this technology, a robot activity is expected in various places or various environments. Therefore, this study describes 1) how the location of the destination of the robot in real world is measured based on the image obtained by one camera and 2) how the robot is navigated to the destination where a user points out on the display, on which the forward scene is imaged. The cases where there are some obstacles on the way to the destination are considered. The roving robot tries to find the shortest way to the destination based on the information on the locations of the obstacles and the destination by using the reinforcement learning, which is a hopeful candidate in the autonomous control technique. In addition, the measurement method for the indicated location based on the image is described, the simulation result for the path search by using the reinforcement leaning is shown, and the experiment result of navigation in real field is shown. Finally, the main conclusions are summarized.

Development of a real-time range finder interpolating the parallax signals for surfaces

Abstract In order to apply “eyes” to an autonomous running vehicle, a real-time range finder has been developed that can measure both the distance to every point on a plane without any pattern and its inclination. Although it seems contrary to the fact that the parallax is inversely proportional to the distance, this paper derives the mathematical ground that a plane in a real 3-D space ( X , Y , Z ) is mapped also into the plane in a spatial parallax ( x , y , Δx ). On this basis, an electronic circuit has been manufactured which can compute in video rate the interpolation value between the parallaxes at two characteristic points by using several programmable logic devices. It has also been demonstrated that this device is applicable to the measurements of the slant angles of and the directions normal to a simple plane or a slightly curved surface.

ESTABLISHMENT OF WIDE SENSE DIGITAL FILTER FOR WIND NOISE CANCELLATION IN A LOW FREQUENCY SOUND MEASUREMENT

It is well known that a wind noise strongly influences observed data, especially when measuring a low frequency acoustic signal. The main reason for this seems to be that major spectrum component of the wind noise coexists within the same frequency bands as the low frequency sound signals. Therefore, it is principally difficult to eliminate the undesirable effect of wind noise when measuring low frequency acoustic signals, only by employing the traditional electrical circuit type analog filter.

A stochastic response evaluation using the digital filter supported by method for a double‐wall‐type sound insulation system excited by the actual music sound

Recently, for the purpose of reducing a residential environmental noise, many sound insulation systems are often improved acoustically by changing their geometrical scales and/or acoustical characteristics. In this paper, in a direct connection with the stochastic signal information processing along a real physical time, a practical method of identification and probabilistic prediction for insulation systems are theoretically and experimentally proposed in the object‐oriented expression forms by functionally introducing a few system parameters. These functional system parameters are supported by many physical structural factors in close relation to the well‐known statistical energy analysis (SEA) method. First, a new trial of identification of the above functional system parameters and the output probabilistic prediction for a geometrical change of double‐wall‐type sound insulation system altered by the above improvement work, especially, under the existence of a strong background noise, is theoretically ...