José Manuel Mossi

Detection of Parked Vehicles Using Spatiotemporal Maps

This paper presents a video-based approach to detect the presence of parked vehicles in street lanes. Potential applications include the detection of illegally and double-parked vehicles in urban scenarios and incident detection on roads. The technique extracts information from low-level feature points (Harris corners) to create spatiotemporal maps that describe what is happening in the scene. The method neither relies on background subtraction nor performs any form of object tracking. The system has been evaluated using private and public data sets and has proven to be robust against common difficulties found in closed-circuit television video, such as varying illumination, camera vibration, the presence of momentary occlusion by other vehicles, and high noise levels.

Robust motion detector for video surveillance applications

This paper presents a robust motion-detector video sensor. It is intended to operate in surveillance applications for long periods of time with time-varying noise level. It makes use of the fact that whenever there is no motion a similarity measure between frames tends to have similar values.

A Comparative Study of Facial Landmark Localization Methods for Face Recognition Using HOG descriptors

This paper compares several approaches to extract facial landmarks and studies their influence on face recognition problems. In order to obtain fair comparisons, we use the same number of facial landmarks and the same type of descriptors (HOG descriptors) for each approach. The comparative results are obtained using FERET and FRGC datasets and show that better recognition rates are obtained when landmarks are located at real facial fiducial points. However, if the automatic detection of these is compromised by the difficulty of the images, better results are obtained using fixed landmarks grids.

Real-time traffic analysis at night-time

This paper presents a video-based approach to traffic analysis and monitoring in night light conditions. In this kind of scenarios the headlights of the vehicles are the main features of the image taken from an urban or inter-urban traffic camera. The body of the vehicles is very low contrasted and many of the algorithms used in day-time decrease their performance. In our algorithm, we detect car headlights, and using this information, we obtain the three main magnitudes used in traffic monitoring: number of vehicles per time unit, i.e. intensity, mean speed, and occupancy. Extensive evaluations show that the system exhibits an excellent performance with real-time video sequences from cameras of the Traffic Authority of the city of Valencia, Spain.

Video-based traffic queue length estimation

This paper presents an approach to estimate traffic queue lengths. It does not rely on any sort of object tracking or background subtraction. The method is based on the detection of low level-features (corners) which can be associated to the presence of vehicles. Then, features are classified as moving or static. An analysis of the locations of static corners allows to measure queue lengths in meters. Errors are smaller than vehicle lengths. The algorithm has been tested with around the clock recordings in different real scenarios.

Re-identifying people in the wild

People re-identification in uncontrolled scenarios is a difficult task since people appearance may significantly vary along time due to changes in illumination, changes in the person pose or the presence of undesired objects in the scene. In order to cope with this temporal variability in the person appearance, we introduce the concept of Bags of Appearances (BoA) to describe each person. A BoA is a container of color features that fully represents a person by collecting all their different appearances along time. Matching of bags is performed in a probabilistic framework by accumulating the probability of matching for all of the elements of each bag. Experiments have been conducted in a real shop where clients were re-identified at the entrance and exit. Results improve state-of-the art methods and confirm that our proposal successfully copes with rough changes in the people appearance.

Color HOG-EBGM for face recognition

Most face recognition algorithms make only use of intensity information of the images discarding color as a distinctive cue. This paper extends the HOG-EBGM face recognition algorithm to use color information. In HOG-EBGM, each face is represented by a graph described by HOG features at specific landmarks. The color extension of the method here proposed intends to make the algorithm robust against color changes, while keeping its former robustness against scale, position, rotation and intensity variations. In the paper, several color representations of the faces are studied. Also, to reduce the higher dimensionality of the new color features, a comparison of dimensionality reduction techniques is included. Results on the Experiment 4 of the Face Recognition Grand Challenge show that color HOG-EBGM outperforms the gray-scale version of the algorithm in all cases. The best results were obtained using the Opponent color space with LDA.

Estimating Point of Regard with a Consumer Camera at a Distance

In this work, we have studied the viability of a novel technique to estimate the POR that only requires video feed from a consumer camera. The system can work under uncontrolled light conditions and does not require any complex hardware setup. To that end we propose a system that uses PCA feature extraction from the eyes region followed by non-linear regression. We evaluated three state of the art non-linear regression algorithms. In the study, we also compared the performance using a high quality webcam versus a Kinect sensor. We found, that despite the relatively low quality of the Kinect images it achieves similar performance compared to the high quality camera. These results show that the proposed approach could be extended to estimate POR in a completely non-intrusive way.

Using latent features for short-term person re-identification with RGB-D cameras

This paper presents a system for people re-identification in uncontrolled scenarios using RGB-depth cameras. Compared to conventional RGB cameras, the use of depth information greatly simplifies the tasks of segmentation and tracking. In a previous work, we proposed a similar architecture where people were characterized using color-based descriptors that we named bodyprints. In this work, we propose the use of latent feature models to extract more relevant information from the bodyprint descriptors by reducing their dimensionality. Latent features can also cope with missing data in case of occlusions. Different probabilistic latent feature models, such as probabilistic principal component analysis and factor analysis, are compared in the paper. The main difference between the models is how the observation noise is handled in each case. Re-identification experiments have been conducted in a real store where people behaved naturally. The results show that the use of the latent features significantly improves the re-identification rates compared to state-of-the-art works.

Two-Dimensional LMS Adaptation Strategies for Nonstationary Signals

Abstract In this paper we propose a new two-dimensional least mean squares algorithm (2D-LMS) which is able to track nonstationarities in both vertical and horizontal directions with a computational load comparable to 1D-LMS methods of the same number of weights. The main difference of our method consists in the proposed strategy to run the image in order to update the filter weights. Smaller initial transients, as well as a reduction in computational load and storage are achieved. Simulations comparing the behavior of our method to recently published methods of 2D-LMS adaptive filtering, have been carried out, showing the main advantages of the proposed method.