Jesús Nuevo

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

Gaze Fixation System for the Evaluation of Driver Distractions Induced by IVIS

We present a method to monitor driver distraction based on a stereo camera to estimate the face pose and gaze of a driver in real time. A coarse eye direction is composed of face pose estimation to obtain the gaze and drivers fixation area in the scene, which is a parameter that gives much information about the distraction pattern of the driver. The system does not require any subject-specific calibration; it is robust to fast and wide head rotations and works under low-lighting conditions. The system provides some consistent statistics, which help psychologists to assess the driver distraction patterns under influence of different in-vehicle information systems (IVISs). These statistics are objective, as the drivers are not required to report their own distraction states. The proposed gaze fixation system has been tested on a set of challenging driving experiments directed by a team of psychologists in a naturalistic driving simulator. This simulator mimics conditions present in real driving, including weather changes, maneuvering, and distractions due to IVISs. Professional drivers participated in the tests.

Visual Monitoring of Driver Inattention

The increasing number of traffic accidents due to driver inattention has become a serious problem for society. Every year, about 45,000 people die and 1.5 million people are injured in traffic accidents in Europe. These figures imply that one person out of every 200 European citizens is injured in a traffic accident every year and that around one out 80 European citizens dies 40 years short of the life expectancy. It is known that the great majority of road accidents (about 90–95%) are caused by human error. More recent data has identified inattention (including distraction and falling asleep at the wheel) as the primary cause of accidents, accounting for at least 25% of the crashes [15]. Road safety is thus a major European health problem. In the “White Paper on European Transport Policy for 2010,” the European Commission declares the ambitious objective of reducing by 50% the number of fatal accidents on European roads by 2010 (European Commission, 2001). According to the U.S. National Highway Traffic Safety Administration (NHTSA), falling asleep while driving is responsible for at least 100,000 automobile crashes annually. An annual average of roughly 70,000 nonfatal injuries and 1,550 fatalities results from these crashes [32, 33]. These figures only cover crashes happening between midnight and 6 a.m., involving a single vehicle and a sober driver traveling alone, including the car departing from the roadway without any attempt to avoid the crash. These figures underestimate the true level of the involvement of drowsiness because they do not include crashes at daytime hours involving multiple vehicles, alcohol, passengers or evasive maneuvers. These statistics do not deal with crashes caused by driver distraction either, which is believed to be a larger problem. Between 13 and 50% of crashes are attributed to distraction, resulting in as many as 5,000 fatalities per year. Increasing use of in-vehicle information systems (IVISs) such as cell phones, GPS navigation systems, satellite radios and DVDs has exacerbated the problem by introducing additional sources of distraction. That is, the more IVISs the more sources of distraction from the most basic task at hand, i.e., driving the vehicle. Enabling drivers to benefit from IVISs without diminishing safety is an important challenge. This chapter presents an original system for monitoring driver inattention and alerting the driver when he is not paying adequate attention to the road in order to prevent accidents. According to [40] the driver inattention status can be divided into two main categories: distraction detection and identifying sleepiness. Likewise, distraction can be divided in two main types: visual and cognitive. Visual distraction is straightforward, occurring when drivers look away from the roadway (e.g., to adjust a radio). Cognitive distraction occurs when drivers think about something not directly related to the current vehicle control task (e.g., conversing on a hands-free cell phone or route planning). Cognitive distraction impairs the ability of drivers to detect targets across the entire visual scene and causes gaze to be concentrated in the center of the driving scene. This work is focused in the sleepiness category. However, sleepiness and cognitive distraction partially overlap since the context awareness of the driver is related to both, which represent mental occurrences in humans [26]. The rest of the chapter is structured as follows. In Sect. 2 we present a review of the main previous work in this direction. Section 3 describes the general system architecture, explaining its main parts. Experimental

Estimating surrounding vehicles' pose using computer vision

This paper presents a computer vision-based approach to tracking surrounding vehicles and estimating their trajectories, in order to detect potentially dangerous situations. Images are acquired using a camera mounted in the egovehicle. Estimations of the distance, velocity and orientation of other vehicles on the road are obtained by detecting their lights and shadow. Because 3D information is not readily available in a mono-camera system, several sets of constraints and assumptions on the geometry of both road and vehicles are proposed and tested in this paper. Kalman filters are used to track the detected vehicles. We also study the advantages of tracking the vehicles in road space (world coordinates), or tracking the position of the lights and shadows on the image. The performance of the approaches is evaluated on video recorded in urban environment.

RSMAT: Robust simultaneous modeling and tracking

This paper describes a robust on-line appearance modeling and tracking method, based on simultaneous modeling and tracking (SMAT). The appearance model is defined by a series of clusters, built in a video sequence using previously encountered samples. This model is used to search for the corresponding element in the following frames. Three alternative incremental clustering methods are proposed to increase the robustness and description capabilities of the model. The proposal is evaluated on an application of face tracking for driver monitoring. The test set comprises sequences of drivers recorded outdoors and in a truck simulator, which contain examples of occlusions and self-occlusions, as well as illumination changes. The performance is evaluated and compared with that of the original SMAT proposal and the recently presented stacked trimmed active shape model (STASM). Our proposal shows better results than the original SMAT and similar fitting error levels to STASM, with much faster execution times and better robustness to self-occlusions.

Road Vehicle Recognition in Monocular Images

This paper describes a monocular vision-based Vehicle Recognition System in which the basic components of road vehicles are first located in the image and then combined with a SVM-based classifier. The challenge is to use a single camera as input. This poses the problem of vehicle detection and recognition in real, cluttered road images. A distributed learning approach is proposed in order to better deal with vehicle variability, illumination conditions, partial occlusions and rotations. The vehicle searching area in the image is constrained to the limits of the lanes, which are determined by the road lane markings. By doing so, the rate of false positive detections is largely decreased. A large database containing thousands of vehicle examples extracted from real road images has been created for learning purposes. We present and discuss the results achieved up to date.

Analysing Driver's Attention Level using Computer Vision

This paper presents a system for evaluating the attention level of a driver using computer vision. The system detects head movements, facial expressions and the presence of visual cues that are known to reflect the users level of alertness. The fusion of these data allows our system to detect both aspects of inattention (drowsiness and distraction), improving the reliability of the monitoring over previous approaches mainly based on detecting only one (drowsiness). Head movements are estimated by robustly tracking a 3D face model with RANSAC and POSIT methods. The 3D model is automatically initialized. Facial expressions are recognized with a model-based method, where different expressions are represented by a set of samples in a low-dimensional manifold in the space of deformations. The system is able to work with different drivers without specific training. The approach has been tested on video sequences recorded in a driving simulator and in real driving situations. The methods are computationally efficient and the system is able to run in real-time.

Face pose estimation and tracking using automatic 3D model construction

This paper presents a method for robustly tracking and estimating the face pose of a person in both indoor and outdoor environments. The method is invariant to identity and that does not require previous training. A face model is automatically initialized and constructed on-line, when the face is frontal to the stereo camera system. To build the model, a fixed point distribution is superposed over the frontal face, and several appropriate points close to those locations are chosen for tracking. Using the stereo correspondence of the two cameras, the 3D coordinates of these points are extracted, and the 3D model is created. RANSAC and POSIT are used for tracking and 3D pose calculation at each frame. The approach runs in real time, and has been tested on sequences recorded in the laboratory and in a moving car.

Real-time robust face tracking for driver monitoring

In this paper we present an active appearance model and a fitting algorithm to track a drivers face, as a component for a driver alertness monitoring system. The proposed method represents some improvements respecting a previous prototype developed by the authors, based on computer vision using active illumination. We test the tracker under different conditions where our previous tracking system fails or exhibits poor performance, such as changing light conditions, occlusions, daylight or drivers wearing glasses. The algorithm is very efficient and able to run in real-time. Some experimental results and conclusions are presented

Real-Time Stereo Visual SLAM in Large-Scale Environments based on SIFT Fingerprints

This paper presents a new method for real-time SLAM calculation applied to autonomous robot navigation in large-scale environments without restrictions. It is exclusively based on the visual information provided by a cheap wide-angle stereo camera. Our approach divide the global map into local sub-maps identified by the so-called SIFT fingerprint. At the sub-map level (low level SLAM), 3D sequential mapping of natural land-marks and the robot location/orientation are obtained using a top-down Bayesian method to model the dynamic behavior. A high abstraction level to reduce the global accumulated drift, keeping real-time constraints, has been added (high level SLAM). This uses a correction method based on the SIFT fingerprints taking for each sub-map. A comparison of the low SLAM level using our method and SIFT features has been carried out. Some experimental results using a real large environment are presented.