David Fernández Llorca

Combination of Feature Extraction Methods for SVM Pedestrian Detection

This paper describes a comprehensive combination of feature extraction methods for vision-based pedestrian detection in Intelligent Transportation Systems. The basic components of pedestrians are first located in the image and then combined with a support-vector-machine-based classifier. This poses the problem of pedestrian detection in real cluttered road images. Candidate pedestrians are located using a subtractive clustering attention mechanism based on stereo vision. A components-based learning approach is proposed in order to better deal with pedestrian variability, illumination conditions, partial occlusions, and rotations. Extensive comparisons have been carried out using different feature extraction methods as a key to image understanding in real traffic conditions. A database containing thousands of pedestrian samples extracted from real traffic images has been created for learning purposes at either daytime or nighttime. The results achieved to date show interesting conclusions that suggest a combination of feature extraction methods as an essential clue for enhanced detection performance

Adaptive road crack detection system by pavement classification.

This paper presents a road distress detection system involving the phases needed to properly deal with fully automatic road distress assessment. A vehicle equipped with line scan cameras, laser illumination and acquisition HW-SW is used to storage the digital images that will be further processed to identify road cracks. Pre-processing is firstly carried out to both smooth the texture and enhance the linear features. Non-crack features detection is then applied to mask areas of the images with joints, sealed cracks and white painting, that usually generate false positive cracking. A seed-based approach is proposed to deal with road crack detection, combining Multiple Directional Non-Minimum Suppression (MDNMS) with a symmetry check. Seeds are linked by computing the paths with the lowest cost that meet the symmetry restrictions. The whole detection process involves the use of several parameters. A correct setting becomes essential to get optimal results without manual intervention. A fully automatic approach by means of a linear SVM-based classifier ensemble able to distinguish between up to 10 different types of pavement that appear in the Spanish roads is proposed. The optimal feature vector includes different texture-based features. The parameters are then tuned depending on the output provided by the classifier. Regarding non-crack features detection, results show that the introduction of such module reduces the impact of false positives due to non-crack features up to a factor of 2. In addition, the observed performance of the crack detection system is significantly boosted by adapting the parameters to the type of pavement.

The Benefits of Dense Stereo for Pedestrian Detection

This paper presents a novel pedestrian detection system for intelligent vehicles. We propose the use of dense stereo for both the generation of regions of interest and pedestrian classification. Dense stereo allows the dynamic estimation of camera parameters and the road profile, which, in turn, provides strong scene constraints on possible pedestrian locations. For classification, we extract spatial features (gradient orientation histograms) directly from dense depth and intensity images. Both modalities are represented in terms of individual feature spaces, in which discriminative classifiers (linear support vector machines) are learned. We refrain from the construction of a joint feature space but instead employ a fusion of depth and intensity on the classifier level. Our experiments involve challenging image data captured in complex urban environments (i.e., undulating roads and speed bumps). Our results show a performance improvement by up to a factor of 7.5 at the classification level and up to a factor of 5 at the tracking level (reduction in false alarms at constant detection rates) over a system with static scene constraints and intensity-only classification.

Accurate Global Localization Using Visual Odometry and Digital Maps on Urban Environments

Over the past few years, advanced driver-assistance systems (ADASs) have become a key element in the research and development of intelligent transportation systems (ITSs) and particularly of intelligent vehicles. Many of these systems require accurate global localization information, which has been traditionally performed by the Global Positioning System (GPS), despite its well-known failings, particularly in urban environments. Different solutions have been attempted to bridge the gaps of GPS positioning errors, but they usually require additional expensive sensors. Vision-based algorithms have proved to be capable of tracking the position of a vehicle over long distances using only a sequence of images as input and with no prior knowledge of the environment. This paper describes a full solution to the estimation of the global position of a vehicle in a digital road map by means of visual information alone. Our solution is based on a stereo platform used to estimate the motion trajectory of the ego vehicle and a map-matching algorithm, which will correct the cumulative errors of the vision-based motion information and estimate the global position of the vehicle in a digital road map. We demonstrate our system in large-scale urban experiments reaching high accuracy in the estimation of the global position and allowing for longer GPS blackouts due to both the high accuracy of our visual odometry estimation and the correction of the cumulative error of the map-matching algorithm. Typically, challenging situations in urban environments such as nonstatic objects or illumination exceeding the dynamic range of the cameras are shown and discussed.

Vehicle logo recognition in traffic images using HOG features and SVM

In this paper a new vehicle logo recognition approach is presented using Histograms of Oriented Gradients (HOG) and Support Vector Machines (SVM). The system is specifically devised to work with images supplied by traffic cameras where the logos appear with low resolution. A sliding-window technique combined with a majority voting scheme are used to provide the estimated car manufacturer. The proposed approach is assessed on a set of 3.579 vehicle images, captured by two different traffic cameras that belong to 27 distinctive vehicle manufacturers. The reported results show an overall recognition rate of 92.59%, which supports the use of the system for real applications.

An Experimental Study on Pitch Compensation in Pedestrian-Protection Systems for Collision Avoidance and Mitigation

This paper describes an improved stereovision system for the anticipated detection of car-to-pedestrian accidents. An improvement of the previous versions of the pedestrian-detection system is achieved by compensation of the cameras pitch angle, since it results in higher accuracy in the location of the ground plane and more accurate depth measurements. The system has been mounted on two different prototype cars, and several real collision-avoidance and collision-mitigation experiments have been carried out in private circuits using actors and dummies, which represents one of the main contributions of this paper. Collision avoidance is carried out by means of deceleration strategies whenever the accident is avoidable. Likewise, collision mitigation is accomplished by triggering an active hood system.

Clavileño: Evolution of an autonomous car

Cars capable of driving in urban environments autonomously are today one of the most challenging topics in the intelligent transportation systems (ITS) field. This paper deals with the evolution of Clavileño -a gas propelled vehicle-in its automation process towards a fully autonomous car driving in a real word. So, the required modifications for a mass-produced car in order to equip it with automatic driving capabilities; the on-board sensor systems to analyze the environment; the autonomous guidance system as well as the cooperative maneuvers implemented and the local evaluation system are presented. The system has been tested in a controlled area with other vehicles in several experiments with good results.

Error Analysis in a Stereo Vision-Based Pedestrian Detection Sensor for Collision Avoidance Applications

This paper presents an analytical study of the depth estimation error of a stereo vision-based pedestrian detection sensor for automotive applications such as pedestrian collision avoidance and/or mitigation. The sensor comprises two synchronized and calibrated low-cost cameras. Pedestrians are detected by combining a 3D clustering method with Support Vector Machine-based (SVM) classification. The influence of the sensor parameters in the stereo quantization errors is analyzed in detail providing a point of reference for choosing the sensor setup according to the application requirements. The sensor is then validated in real experiments. Collision avoidance maneuvers by steering are carried out by manual driving. A real time kinematic differential global positioning system (RTK-DGPS) is used to provide ground truth data corresponding to both the pedestrian and the host vehicle locations. The performed field test provided encouraging results and proved the validity of the proposed sensor for being used in the automotive sector towards applications such as autonomous pedestrian collision avoidance.

Traffic data collection for floating car data enhancement in V2I networks

This paper presents a complete vision-based vehicle detection system for floating car data (FCD) enhancement in the context of vehicular ad hoc networks (VANETs). Three cameras (side-, forward- and rear- looking cameras) are installed onboard a vehicle in a fleet of public buses. Thus, a more representative local description of the traffic conditions (extended FCD) can be obtained. Specifically, the vision modules detect the number of vehicles contained in the local area of the host vehicle (traffic load) and their relative velocities. Absolute velocities (average road speed) and global positioning are obtained after combining the outputs provided by the vision modules with the data supplied by the CAN Bus and the GPS sensor. This information is transmitted by means of a GPRS/UMTS data connection to a central unit which merges the extended FCD in order to maintain an updated map of the traffic conditions (traffic load and average road speed). The presented experiments are promising in terms of detection performance and computational costs. However, significant effort is further necessary before deploying a system for large-scale real applications.

Robust visual odometry for vehicle localization in urban environments

This paper describes a new approach for estimating the vehicle motion trajectory in complex urban environments by means of visual odometry. A new strategy for robust feature extraction and data post-processing is developed and tested on-road. Images from scale-invariant feature transform (SIFT) features are used in order to cope with the complexity of urban environments. The obtained results are discussed and compared to previous works. In the prototype system, the ego-motion of the vehicle is computed using a stereo-vision system mounted next to the rear view mirror of the car. Feature points are matched between pairs of frames and linked into 3D trajectories. The distance between estimations is dynamically adapted based on re-projection and estimation errors. Vehicle motion is estimated using the non-linear, photogrametric approach based on RAndom SAmple Consensus (RANSAC). The final goal is to provide on-board driver assistance in navigation tasks, or to provide a means of autonomously navigating a vehicle. The method has been tested in real traffic conditions without using prior knowledge about the scene or the vehicle motion. An example of how to estimate a vehicles trajectory is provided along with suggestions for possible further improvement of the proposed odometry algorithm.