Search

Visual object tracking has achieved remarkable progress in recent years and has been broadly applied in intelligent transportation systems such as autonomous vehicles and drones to monitor and analyze the behavior of specific targets. One typical tracking approach is the discriminative tracker, which branches into two main categories: the correlation filter (CF) and the convolutional neural network (CNN). However, most of the current researches consider both categories as two separate techniques and only rely on one of them. Thus, a dense cooperation between the CF and the CNN still remains less discovered and the question of how to effectively join both techniques to further boost the tracking performance is still open. To address this issue, in this paper, we propose a collaborative architecture which incorporates models constructed with both techniques and dynamically aggregates their response maps for target inference. By an alternating optimization, both models are learned on each other’s errors to persistently improve the classification power of the whole tracker. For further efficiency, we present a faster solver for our utilized CF and an analytical solution for dynamic model weighting. Through experiments on standard benchmarks, we reveal the influence of key factors on the joint learning architecture and show that it outperforms the state-of-the-art approaches.

Visual tracking of multiple objects is an essential component for a perception system in autonomous driving vehicles. One of the favorable approaches is the tracking-by-detection paradigm, which links current detection hypotheses to previously estimated object trajectories (also known as tracks) by searching appearance or motion similarities between them. As this search operation is usually based on a very limited spatial or temporal locality, the association can fail in cases of motion noise or long-term occlusion. In this paper, we propose a novel tracking method that solves this problem by putting together information from both enlarged structural and temporal domain. For efficiency without loss of optimality, this approach is decomposed in to three stages, with each dealing with only one constrained association task, and thus, it follows the alternating optimization fashion. In our approach, detections are first assembled into small tracklets based on meta-measurements of object affinity. The association task for tracklets-to-tracks is solved by structural information based on a motion pattern between them. Here, we propose new rules to decouple the processing time from the tracklet length. Furthermore, constraints from temporal domain are introduced to recover objects, which are long-time disappearing due to failed detection or long-term occlusion. By putting together the heterogeneous domain information, our approach exhibits an improved state-of-the-art performance on standard benchmarks. With relatively little processing time, an online and real-time tracking is also permitted in our approach.

Despite the fact that in recent years, vision-based tracking approaches have made significant progress, the task of tracking vehicles at night still remains challenging. Visual information is strongly deteriorated or at least degraded due to poor illumination conditions. This reduces the perceptive ability of vision systems significantly and can even lead to target loss, resulting in false estimation and/or false prediction of object behavior. In this paper, we propose a novel online-learning method to track vehicles at night. Our method is based on the kernelized correlation filter and assembles different feature channels to kernelized experts. By estimating their reliabilities, we force the appearance model to focus on the most discriminative visual features to accomplish the classification. In addition, a temporal optimization step in conjunction with a memory model is used to remove outliers and keep the most reliable samples to train the tracker models. Experiments over various daytime and weather conditions show that our approach outperforms existing trackers at night and in case of bad weather while offering state-of-the-art performance in more favorable situations. As our tracker has only little computational cost, it is appropriate for use cases with real-time requirements like in automotive or industrial applications.

Vision-based object tracking approach has been drawing increased attention from both the academia and the industry in recent years. One of its most successful applications is the monitoring system, which can be installed on infrastructures or on mobile platforms (e.g., vehicles) to track pedestrians or bicyclists in a pre-defined region and further to prevent probable accidents. Despite tremendous progress achieved, the task of visual tracking is still challenging, especially in dealing with severe occlusions, where the tracker may fail due to the abrupt change of object appearance. Such case is common not only in traffic scenarios but also in other tracking tasks. Aiming to tackle this problem, in this paper, we propose a new tracking approach by adopting part based trackers, which are built in the form of correlation filter. In this approach, the occluded object parts are identified by leveraging the knowledge derived from image features and filter responses. With the help of a masking process, visible object areas are acquired in a pixel-wise precision and utilized to build part filters. As both the number and size of part filters are adapted to the current object appearance, the influence of occlusions can be significantly suppressed. Experimental results on traffic sequences demonstrate that our tracker performs robust against occlusion, especially in cases, where long term and severe occlusions appear. A further experiment on the standard benchmark proves that our approach outperforms state-of-the-art methods in tracking various object classes under varied circumstances. Furthermore, the proposed tracker is sophisticatedly designed and is feasible for real time applications.

This paper describes an algorithm that enables a railway vehicle to determine its position in a track network. The system is based solely on onboard sensors such as a velocity sensor and a Global Navigation Satellite System (GNSS) sensor and does not require trackside infrastructure such as axle counters or balises. The paper derives a probabilistic modeling of the localization task and develops a sensor fusion approach to fuse the inputs of the GNSS sensor and the velocity sensor with the digital track map. We describe how we can treat ambiguities and stochastic uncertainty adequately. Moreover, we introduce the concept of virtual balises that can be used to replace balises on the track and evaluate the approach experimentally. This paper focuses on an accurate modeling of sensor and estimation uncertainties, which is relevant for safety critical applications.

The ever-increasing requirements on gas turbine efficiency, which are at least partially met by increasing firing temperatures, and the simultaneous demand for reduced emissions, necessitate much more accurate calculations of the combustion process and combustor wall temperatures. Thermocouples give locally very accurate measurements of these temperatures, but there is a practical limit to the amount of measurement points. Thermal paints are another established measurement technique, but are toxic and at the same time require dedicated, short-duration tests. Thermal History Paints (THPs) provide an innovative alternative to the aforementioned techniques, but so far only a limited number of tests has been conducted under real engine conditions. THPs are similar in their chemical and physical make-up to conventional thermographic phosphors which have been successfully used in gas turbine applications for on-line temperature detection before. A typical THP comprises of oxide ceramic pigments and a water based binder. The ceramic is synthesized to be amorphous and when heated it crystallizes, permanently changing the microstructure. The ceramic is doped with lanthanide ions to make it phosphorescent. The lanthanide ions act as atomic level sensors and as the structure of the material changes, so do the phosphorescent properties of the material. By measuring the phosphorescence the maximum temperature of exposure can be determined through calibration, enabling post operation measurements at ambient conditions. This paper describes a test in which THP was applied to an impingement-cooled front panel from a combustor of an industrial gas turbine. Since this component sees a wide range of temperatures, it is ideally suited for the testing of the measurement techniques under real engine conditions. The panel was instrumented with a thermocouple and thermal paint was applied to the cold side of the impingement plate. THP was applied to the hot-gas side of this plate for validation against the other measurement techniques and to evaluate its resilience against the reacting hot gas environment. The durability and temperature results of the three different measurement techniques are discussed. The results demonstrate the benefits of THPs as a new temperature profiling technique. It is shown that the THP exhibited greater durability compared to the conventional thermal paint. Furthermore, the new technology provided detailed measurements down to millimeters indicating local temperature variations and global variations over the complete component.

In this paper, we present the concepts and methods developed for the autonomous vehicle known as AnnieWAY, which is our winning entry to the 2011 Grand Cooperative Driving Challenge. We describe algorithms for sensor fusion, vehicle-to-vehicle communication, and cooperative control. Furthermore, we analyze the performance of the proposed methods and compare them with those of competing teams. We close with our results from the competition and lessons learned.

This paper presents a novel technique of stereo vision based on the combination of an omnidirectional camera and a perspective camera. The technique combines the 360° field of view of the omnidirectional camera with the long field of view of a perspective camera. We describe the setup of such a camera system and how it can be used to achieve 3D-position estimates. Furthermore, we develop a maximum likelihood approach and a Bayesian approach that are able to fuse monocular and binocular observations of the same object to estimate its position and movement and show how this technique can be applied successfully in the RoboCup MiddleSizeLeague.

The second preparation workshop for the Grand Cooperative Driving Challenge (GCDC) was held in Helmond, The Netherlands, in December 2009. Like the first workshop in May, it was organized excellently by TNO, the Dutch research organization which, together with HTAS, has initiated the GCDC. The main objective of the workshop was to discuss the rules and regulations with potential participants starting from a draft document provided by TNO. The workshop was attended by 80 researchers from different universities, knowledge institutes, and manufacturers from eighteen countries spread over four continents.

Imaging of OH* or CH* chemiluminescence with intensified cameras is often employed for the determination of heat release in premixed flames. Proportionality is commonly assumed, but in the turbulent case this assumption is not justified. Substantial deviations from proportionality are observed, which are due to turbulence-chemistry interactions. In this study a model based correction method is presented to obtain a better approximation of the spatially resolved heat release rate of lean turbulent flames from OH* measurements. The correction method uses a statistical strain rate model to account for the turbulence influence. The strain rate model is evaluated with time-resolved velocity measurements of the turbulent flow. Additionally, one-dimensional simulations of strained counterflow flames are performed to consider the non-linear effect of turbulence on chemi-luminescence intensities. A detailed reaction mechanism, which includes all relevant chemiluminescence reactions and deactivation processes, is used. The result of the simulations is a lookup table of the ratio between heat release rate and OH* intensity with strain rate as parameter. This lookup table is linked with the statistical strain rate model to obtain a correction factor which accounts for the non-linear relationships between OH* intensity, heat release rate, and strain rate. The factor is then used to correct measured OH* intensities to obtain the local heat release rate. The corrected intensities are compared to heat release distributions which are measured with an alternative method. For all investigated flames in the lean, partially premixed regime the corrected OH* intensities are in very good agreement with the heat release rate distributions of the flames.Copyright © 2011 by ASME

The determination of the heat release in technical flames is commonly done via bandpass filtered chemiluminescence measurements in the wavelength range of OH* or CH* radicals, which are supposed to be a measure for the heat release rate. However, these indirect heat release measurements are problematic, because the measured intensities are the superposition of the desired radical emissions and contributions from the broadband emissions of CO2 * . Furthermore, the chemiluminescence intensities are strongly affected by the local air excess ratio of the flame and the turbulence intensity in the reaction zone. To investigate the influence of these effects on the applicability of chemiluminescence as a measure for heat release rate in turbulent flames with mixture gradients, a reference method is used, which is based on the first law of thermodynamics. It is shown that although the integral heat release can be correlated with the integral chemiluminescence intensities, the heat release distribution is not properly represented by any signal from OH* or CH*. No reliable information about the spatially resolved heat release can be obtained from chemiluminescence measurements in flames with mixture gradients.Copyright © 2009 by ASME

In order to assess the stability of gas turbine combustors measured flame transfer functions are frequently used in thermoacoustic network models. Although many combustion systems operate at high pressure, the measurement of flame transfer functions was essentially limited to atmospheric conditions in the past. With the test rig employed in the study presented in the paper transfer function measurements were made for a wide range of combustor pressures. The results show similarities of the amplitude response in the entire pressure range investigated. However, the increase of the pressure leads to a considerable amplitude gain at higher frequencies. In the low frequency regime the phase is also independent of pressure, whereas above this region the pressure increase results in a considerably smaller phase lag. These observations are particularly important when evaluating Rayleigh’s criterion: Interestingly, the choice of the operating pressure can render a system stable or unstable, so that the common procedure of applying flame transfer functions measured at ambient pressure for the high pressure engine case may not always be appropriate. The detailed analysis of high speed camera images, which were recorded to get locally resolved information on the flame response reveal different regions of activity within the flame that change in strength, size and location with changing operating conditions. The observed transfer function phase behavior is explained by the interaction of those regions and it is shown that the region of highest dynamic activity dominates the phase.Copyright © 2006 by ASME