Search

We consider scenarios where we have zero instances of real pedestrian data (e.g., a newly installed surveillance system in a novel location in which no labeled real data or unsupervised real data exists yet) and a pedestrian detector must be developed prior to any observations of pedestrians. Given a single image and auxiliary scene information in the form of camera parameters and geometric layout of the scene, our approach infers and generates a large variety of geometrically and photometrically accurate potential images of synthetic pedestrians along with purely accurate ground-truth labels through the use of computer graphics rendering engine. We first present an efficient discriminative learning method that takes these synthetic renders and generates a unique spatially-varying and geometry-preserving pedestrian appearance classifier customized for every possible location in the scene. In order to extend our approach to multi-task learning for further analysis (i.e., estimating pose and segmentation of pedestrians besides detection), we build a more generalized model employing a fully convolutional neural network architecture for multi-task learning leveraging the “free" ground-truth annotations that can be obtained from our pedestrian synthesizer. We demonstrate that when real human annotated data is scarce or non-existent, our data generation strategy can provide an excellent solution for an array of tasks for human activity analysis including detection, pose estimation and segmentation. Experimental results show that our approach (1) outperforms classical models and hybrid synthetic-real models, (2) outperforms various combinations of off-the-shelf state-of-the-art pedestrian detectors and pose estimators that are trained on real data, and (3) surprisingly, our method using purely synthetic data is able to outperform models trained on real scene-specific data when data is limited. [ABSTRACT FROM AUTHOR]

In this paper, a procedure for estimation of effective modulus of a multilayered hot mix asphalt (HMA) pavement using intelligent compaction (1C) is investigated. The study is conducted during the construction of an interstate highway (1-35) in Norman, OK. A complete coverage of the level of compaction of each of the asphalt pavement layers was recorded using the intelligent asphalt compaction analyzer (IACA). The collected IACA data allow determination of the level of compaction (density) at any selected location, for each layer, and provided a set of global positioning system (GPS) coordinates. Calibration procedures have previously been tested and validated by the authors to estimate the density of different types of pavements from IACA data. In this paper, a different calibration procedure is used to measure the dynamic modulus instead of the density of a pavement using IACA. Considering the IACA estimated density, the dynamic modulus of each of the selected locations for an individual pavement layer was measured from laboratory developed master curves. Thereafter, an effective modulus of the three-layer pavement system was calculated for all of the selected locations using Odemark's method. The proposed technique was verified by conducting falling-weight deflectometer (FWD) tests at these selected locations. Analyses of the results show that the proposed intelligent compaction technique may be promising in estimating the effective modulus of the pavement layers in a non-destructive manner. In addition, the Witczak model was used to estimate moduli of each of the pavement layers. The comparison of the Witczak model with FWD revealed that the model over-predicted the modulus significantly. [ABSTRACT FROM AUTHOR]

Compaction of hot mix asphalt (HMA) layer is a process of densification through the application of vibratory, dead weight and/or pneumatic tires compactors. These processes are employed independently or in combination, depending on factors such as type of asphalt mix, layer thickness, and weather conditions, to achieve desired density of HMA pavement layers. Appropriate compaction (desired density) of the asphalt mixture is one of the important factors that influence long term performance of an asphalt pavement. During the compaction of HMA layer using vibratory rollers, the applied load and vibration energy cause reorientation and interlocking of asphalt coated aggregates resulting in increase in density. Good compaction is necessary for achieving target specifications such as required density of asphalt layers, stiffness, or volumetric properties of the asphalt mix. These specifications are designed to maximize the resistance of the pavement to deformation, cracking, raveling, moisture damage, and rutting. In the United States of America, target specifications for quality control and assurance of asphalt pavements are usually given in terms of the density of roadway cores extracted from the completed pavement. This paper investigates the level of compaction (in terms of density), that is achievable both in the laboratory and in the field. Compaction quality as a function of compaction time/roller passes and pavement thickness is studied. In addition, the uniformity of compaction over the entire length of the pavement, and the variation in density laterally across the pavement are examined during pavement construction in the field. The results of this study provide insight into the compaction quality that can be achieved during construction. Such information would is helpful for a successful design and the implementation of emerging technologies such as Intelligent Compaction. [ABSTRACT FROM AUTHOR]

Use of a newly developed intelligent compaction technology, called the Intelligent Asphalt Compaction Analyzer (IACA), to estimate the stiffness of an asphalt pavement during construction is addressed in this paper. The IACA is a neural network-based device that functions on the hypothesis that the roller and the underlying pavement layers form a coupled system and that the stiffness of the pavement layer can be determined through the vibration of the roller. For a given asphalt mix, the "Master Curves" for the dynamic modulus are first determined according to the AASHTO TP62-03 test method. During the compaction of the pavement, the IACA records the entire frequency spectrum of the vibrations of the roller and classifies these vibrations into different levels. The master curves are then used to calibrate the IACA to convert these levels of vibrations into a modulus value, representing the stiffness of the asphalt layer being compacted. This two-step process produces stiffness measurements in real-time that are representative of the pavement being constructed. Also, since the neural network is calibrated during the construction of a stretch of the actual pavement, the stiffness measurements are obtained without the need to evaluate the underlying pavement layers and the subgrade. The IACA provides the stiffness values over the entire pavement in a non-destructive manner. Further, the proposed method allows for the verification of the estimated modulus through the falling weight deflectometer (FWD) or similar field tests of the constructed pavement. [ABSTRACT FROM AUTHOR]