Search

A major challenges of deep learning (DL) is the necessity to collect huge amounts of training data. Often, the lack of a sufficiently large dataset discourages the use of DL in certain applications. Typically, acquiring the required amounts of data costs considerable time, material and effort. To mitigate this problem, the use of synthetic images combined with real data is a popular approach, widely adopted in the scientific community to effectively train various detectors. In this study, we examined the potential of synthetic data-based training in the field of intelligent transportation systems. Our focus is on camera-based traffic sign recognition applications for advanced driver assistance systems and autonomous driving. The proposed augmentation pipeline of synthetic datasets includes novel augmentation processes such as structured shadows and gaussian specular highlights. A well-known DL model was trained with different datasets to compare the performance of synthetic and real image-based trained models. Additionally, a new, detailed method to objectively compare these models is proposed. Synthetic images are generated using a semi-supervised errors-guide method which is also described. Our experiments showed that a synthetic image-based approach outperforms in most cases real image-based training when applied to cross-domain test datasets (+10% precision for GTSRB dataset) and consequently, the generalization of the model is improved decreasing the cost of acquiring images.

In this study, the authors analyse the exponential growth of advanced driver assistance systems based on video processing in the past decade. Specifically, they focus on how research and innovative ideas can finally reach the market as cost-effective solutions. They explore well-known computer vision methods for services like lane departure warning systems, collision avoidance systems and point out potential future trends according to a review of the state-of-the-art. Along this study, the authors' own contributions are described as examples of such systems from the perspective of real-time by design, pursuing a trade-off between the accuracy and reliability of the designed algorithms and the restrictive computational, economical and design requisites of embedded platforms.

In this paper, we present a robust vision-based system for vehicle tracking and classification devised for traffic flow surveillance. The system performs in real time, achieving good results, even in challenging situations, such as with moving casted shadows on sunny days, headlight reflections on the road, rainy days, and traffic jams, using only a single standard camera. We propose a robust adaptive multicue segmentation strategy that detects foreground pixels corresponding to moving and stopped vehicles, even with noisy images due to compression. First, the approach adaptively thresholds a combination of luminance and chromaticity disparity maps between the learned background and the current frame. It then adds extra features derived from gradient differences to improve the segmentation of dark vehicles with casted shadows and removes headlight reflections on the road. The segmentation is further used by a two-step tracking approach, which combines the simplicity of a linear 2-D Kalman filter and the complexity of a 3-D volume estimation using Markov chain Monte Carlo (MCMC) methods. Experimental results show that our method can count and classify vehicles in real time with a high level of performance under different environmental situations comparable with those of inductive loop detectors.

The efficient detection and tracking of persons in videos has widrespread applications, specially in CCTV systems for surveillance or forensics applications. In this paper we present a new method for people detection and tracking based on the knowledge of the perspective information of the scene. It allows alleviating two main drawbacks of existing methods: (i) high or even excessive computational cost associated to multiscale detection-by-classification methods; and (ii) the inherent difficulty of the CCTV, in which predominate partial and full occlusions as well as very high intra-class variability. During the detection stage, we propose to use the homograhy of the dominant plane to compute the expected sizes of persons at different positions of the image and thus dramatically reduce the number of evaluation of the multiscale sliding window detection scheme. To achieve robustness against false positives and negatives, we have used a combination of full and upper-body detectors, as well as a Data Association Filter (DAF) inspired in the well-known Rao-Blackwellization-based particle filters (RBPF). Our experiments demonstrate the benefit of using the proposed perspective multiscale approach, compared to conventional sliding window approaches, and also that this perspective information can lead to useful mixes of full-body and upper-body detectors.

An automatic method for rail inspection is introduced in this paper. The method detects rail flaws using computer vision algorithms. Unlike other methods designed for the same goal, we propose a method that automatically fits a 3D rail model to the observations. The proposed strategy is based on the novel combination of a simple but effective laser-camera calibration procedure with the application of an MCMC (Markov Chain Monte Carlo) framework. The proposed particle filter uses the efficient overrelation slice sampling method, which allows us to exploit the temporal coherence of observations and to obtain more accurate estimates than with other sampling techniques. The results show that the system is able to robustly obtain measurements of the wear of the rail. The two other contributions of the paper are the successfull introuction of the slice sampling technique into MCMC particle filters and the proposed online and flexible method for camera-laser calibration.

This paper presents a single camera vehicle detection technique for forward collision warning systems suitable to be integrated in embedded platforms. It combines the robustness of detectors based on classification methods with an innovative perspective multi-scale procedure to scan the images that dramatically reduces the computational cost associated with robust detectors. In our experiments we compare different implementation classifiers in search for a trade-off between the real-time constraint of embedded platforms and the high detection rates required by safety applications.

A multi-classifier formed by specialised classifiers for noise produced by an image is shown in this work. A study has been carried out in the case of cut images, where tree cases of specialization are considered. Classifiers based on neighbourhood criteria are used, the zoning global feature and the Euclidean distance too. Furthermore, the paper explains a modification of the Euclidean distance for classifying cut digits. The experiments have been carried out with images of typewritten digits, taken from real forms. Trying to obtain a strong database to support the experiments, we have cut images deliberately. The recognition rate improves from 84.6% to 97.70%, but whether the system provides information about the disturbance of the image, it can achieve a 98.45%.

This article introduces a 3D vehicle tracking system in a traffic surveillance environment devised for shadow tolling applications. It has been specially designed to operate in real time with high correct detection and classification rates. The system is capable of providing accurate and robust results in challenging road scenarios, with rain, traffic jams, casted shadows in sunny days at sunrise and sunset times, etc. A Bayesian inference method has been designed to generate estimates of multiple variable objects entering and exiting the scene. This framework allows easily mixing different nature information, gathering in a single step observation models, calibration, motion priors and interaction models. The inference of results is carried out with a novel optimization procedure that generates estimates of the maxima of the posterior distribution combining concepts from Gibbs and slice sampling. Experimental tests have shown excellent results for traffic-flow video surveillance applications that can be used to classify vehicles according to their length, width, and height. Therefore, this vision-based system can be seen as a good substitute to existing inductive loop detectors.