Shunsuke Kamijo

Bio: Shunsuke Kamijo is an academic researcher from University of Tokyo. The author has contributed to research in topics: GNSS applications & Global Positioning System. The author has an hindex of 24, co-authored 187 publications receiving 2530 citations. Previous affiliations of Shunsuke Kamijo include Fujitsu & Ibaraki University.

Traffic monitoring and accident detection at intersections

Abstract: We have developed an algorithm, referred to as spatio-temporal Markov random field, for traffic images at intersections. This algorithm models a tracking problem by determining the state of each pixel in an image and its transit, and how such states transit along both the x-y image axes as well as the time axes. Our algorithm is sufficiently robust to segment and track occluded vehicles at a high success rate of 93%-96%. This success has led to the development of an extendable robust event recognition system based on the hidden Markov model (HMM). The system learns various event behavior patterns of each vehicle in the HMM chains and then, using the output from the tracking system, identifies current event chains. The current system can recognize bumping, passing, and jamming. However, by including other event patterns in the training set, the system can be extended to recognize those other events, e.g., illegal U-turns or reckless driving. We have implemented this system, evaluated it using the tracking results, and demonstrated its effectiveness.

3D building model-based pedestrian positioning method using GPS/GLONASS/QZSS and its reliability calculation

Abstract: The current low-cost global navigation satellite systems (GNSS) receiver cannot calculate satisfactory positioning results for pedestrian applications in urban areas with dense buildings due to multipath and non-line-of-sight effects. We develop a rectified positioning method using a basic three-dimensional city building model and ray-tracing simulation to mitigate the signal reflection effects. This proposed method is achieved by implementing a particle filter to distribute possible position candidates. The likelihood of each candidate is evaluated based on the similarity between the pseudorange measurement and simulated pseudorange of the candidate. Finally, the expectation of all the candidates is the rectified positioning of the proposed map method. The proposed method will serve as one sensor of an integrated system in the future. For this purpose, we successfully define a positioning accuracy based on the distribution of the candidates and their pseudorange similarity. The real data are recorded at an urban canyon environment in the Chiyoda district of Tokyo using a commercial grade u-blox GNSS receiver. Both static and dynamic tests were performed. With the aid of GLONASS and QZSS, it is shown that the proposed method can achieve a 4.4-m 1ź positioning error in the tested urban canyon area.

GPS Error Correction With Pseudorange Evaluation Using Three-Dimensional Maps

Abstract: The accuracy of the positions of a pedestrian is very important and useful information for the statistics, advertisement, and safety of different applications. Although the GPS chip in a smartphone is currently the most convenient device to obtain the positions, it still suffers from the effect of multipath and nonline-of-sight propagation in urban canyons. These reflections could greatly degrade the performance of a GPS receiver. This paper describes an approach to estimate a pedestrian position by the aid of a 3-D map and a ray-tracing method. The proposed approach first distributes the numbers of position candidates around a reference position. The weighting of the position candidates is evaluated based on the similarity between the simulated pseudorange and the observed pseudorange. Simulated pseudoranges are calculated using a ray-tracing simulation and a 3-D map. Finally, the proposed method was verified through field experiments in an urban canyon in Tokyo. According to the results, the proposed approach successfully estimates the reflection and direct paths so that the estimate appears very close to the ground truth, whereas the result of a commercial GPS receiver is far from the ground truth. The results show that the proposed method has a smaller error distance than the conventional method.

GNSS/Onboard Inertial Sensor Integration With the Aid of 3-D Building Map for Lane-Level Vehicle Self-Localization in Urban Canyon

Abstract: Lane-level vehicle self-localization is a challenging and significant issue arising in autonomous driving and driver-assistance systems. The Global Navigation Satellite System (GNSS) and onboard inertial sensor integration are among the solutions for vehicle self-localization. However, as the main source in the integration, GNSS positioning performance is severely degraded in urban canyons because of the effects of multipath and non-line-of-sight (NLOS) propagations. These GNSS positioning errors also decrease the performance of the integration. To reduce the negative effects caused by GNSS positioning error, this paper proposes to employ an innovative GNSS positioning technique with the aid of a 3-D building map in the integration. The GNSS positioning result is used as an observation, and this is integrated with the information from the onboard inertial sensor and vehicle speedometer in a Kalman filter framework. To achieve stable performance, this paper proposes to evaluate and consider the accuracy of the employed GNSS positioning method in dynamic integration. A series of experiments in different scenarios is conducted in an urban canyon, which can demonstrate the effectiveness of the proposed method using various evaluation and comparison processes.

NLOS Correction/Exclusion for GNSS Measurement Using RAIM and City Building Models.

Abstract: Currently, global navigation satellite system (GNSS) receivers can provide accurate and reliable positioning service in open-field areas. However, their performance in the downtown areas of cities is still affected by the multipath and none-line-of-sight (NLOS) receptions. This paper proposes a new positioning method using 3D building models and the receiver autonomous integrity monitoring (RAIM) satellite selection method to achieve satisfactory positioning performance in urban area. The 3D building model uses a ray-tracing technique to simulate the line-of-sight (LOS) and NLOS signal travel distance, which is well-known as pseudorange, between the satellite and receiver. The proposed RAIM fault detection and exclusion (FDE) is able to compare the similarity between the raw pseudorange measurement and the simulated pseudorange. The measurement of the satellite will be excluded if the simulated and raw pseudoranges are inconsistent. Because of the assumption of the single reflection in the ray-tracing technique, an inconsistent case indicates it is a double or multiple reflected NLOS signal. According to the experimental results, the RAIM satellite selection technique can reduce by about 8.4% and 36.2% the positioning solutions with large errors (solutions estimated on the wrong side of the road) for the 3D building model method in the middle and deep urban canyon environment, respectively.

Image Processing

Abstract: MUCKE aims to mine a large volume of images, to structure them conceptually and to use this conceptual structuring in order to improve large-scale image retrieval. The last decade witnessed important progress concerning low-level image representations. However, there are a number problems which need to be solved in order to unleash the full potential of image mining in applications. The central problem with low-level representations is the mismatch between them and the human interpretation of image content. This problem can be instantiated, for instance, by the incapability of existing descriptors to capture spatial relationships between the concepts represented or by their incapability to convey an explanation of why two images are similar in a content-based image retrieval framework. We start by assessing existing local descriptors for image classification and by proposing to use co-occurrence matrices to better capture spatial relationships in images. The main focus in MUCKE is on cleaning large scale Web image corpora and on proposing image representations which are closer to the human interpretation of images. Consequently, we introduce methods which tackle these two problems and compare results to state of the art methods. Note: some aspects of this deliverable are withheld at this time as they are pending review. Please contact the authors for a preview.

Real-World Anomaly Detection in Surveillance Videos

Abstract: Surveillance videos are able to capture a variety of realistic anomalies. In this paper, we propose to learn anomalies by exploiting both normal and anomalous videos. To avoid annotating the anomalous segments or clips in training videos, which is very time consuming, we propose to learn anomaly through the deep multiple instance ranking framework by leveraging weakly labeled training videos, i.e. the training labels (anomalous or normal) are at video-level instead of clip-level. In our approach, we consider normal and anomalous videos as bags and video segments as instances in multiple instance learning (MIL), and automatically learn a deep anomaly ranking model that predicts high anomaly scores for anomalous video segments. Furthermore, we introduce sparsity and temporal smoothness constraints in the ranking loss function to better localize anomaly during training. We also introduce a new large-scale first of its kind dataset of 128 hours of videos. It consists of 1900 long and untrimmed real-world surveillance videos, with 13 realistic anomalies such as fighting, road accident, burglary, robbery, etc. as well as normal activities. This dataset can be used for two tasks. First, general anomaly detection considering all anomalies in one group and all normal activities in another group. Second, for recognizing each of 13 anomalous activities. Our experimental results show that our MIL method for anomaly detection achieves significant improvement on anomaly detection performance as compared to the state-of-the-art approaches. We provide the results of several recent deep learning baselines on anomalous activity recognition. The low recognition performance of these baselines reveals that our dataset is very challenging and opens more opportunities for future work. The dataset is available at: http://crcv.ucf.edu/projects/real-world/

ShiDianNao: shifting vision processing closer to the sensor

Abstract: In recent years, neural network accelerators have been shown to achieve both high energy efficiency and high performance for a broad application scope within the important category of recognition and mining applications. Still, both the energy efficiency and performance of such accelerators remain limited by memory accesses. In this paper, we focus on image applications, arguably the most important category among recognition and mining applications. The neural networks which are state-of-the-art for these applications are Convolutional Neural Networks (CNN), and they have an important property: weights are shared among many neurons, considerably reducing the neural network memory footprint. This property allows to entirely map a CNN within an SRAM, eliminating all DRAM accesses for weights. By further hoisting this accelerator next to the image sensor, it is possible to eliminate all remaining DRAM accesses, i.e., for inputs and outputs. In this paper, we propose such a CNN accelerator, placed next to a CMOS or CCD sensor. The absence of DRAM accesses combined with a careful exploitation of the specific data access patterns within CNNs allows us to design an accelerator which is 60× more energy efficient than the previous state-of-the-art neural network accelerator. We present a full design down to the layout at 65 nm, with a modest footprint of 4.86mm2 and consuming only 320mW, but still about 30× faster than high-end GPUs.

Deep Learning for IoT Big Data and Streaming Analytics: A Survey

Abstract: In the era of the Internet of Things (IoT), an enormous amount of sensing devices collect and/or generate various sensory data over time for a wide range of fields and applications. Based on the nature of the application, these devices will result in big or fast/real-time data streams. Applying analytics over such data streams to discover new information, predict future insights, and make control decisions is a crucial process that makes IoT a worthy paradigm for businesses and a quality-of-life improving technology. In this paper, we provide a thorough overview on using a class of advanced machine learning techniques, namely deep learning (DL), to facilitate the analytics and learning in the IoT domain. We start by articulating IoT data characteristics and identifying two major treatments for IoT data from a machine learning perspective, namely IoT big data analytics and IoT streaming data analytics. We also discuss why DL is a promising approach to achieve the desired analytics in these types of data and applications. The potential of using emerging DL techniques for IoT data analytics are then discussed, and its promises and challenges are introduced. We present a comprehensive background on different DL architectures and algorithms. We also analyze and summarize major reported research attempts that leveraged DL in the IoT domain. The smart IoT devices that have incorporated DL in their intelligence background are also discussed. DL implementation approaches on the fog and cloud centers in support of IoT applications are also surveyed. Finally, we shed light on some challenges and potential directions for future research. At the end of each section, we highlight the lessons learned based on our experiments and review of the recent literature.