Gabriel Sanchez-Perez

An Early Fire Detection Algorithm Using IP Cameras

The presence of smoke is the first symptom of fire; therefore to achieve early fire detection, accurate and quick estimation of the presence of smoke is very important. In this paper we propose an algorithm to detect the presence of smoke using video sequences captured by Internet Protocol (IP) cameras, in which important features of smoke, such as color, motion and growth properties are employed. For an efficient smoke detection in the IP camera platform, a detection algorithm must operate directly in the Discrete Cosine Transform (DCT) domain to reduce computational cost, avoiding a complete decoding process required for algorithms that operate in spatial domain. In the proposed algorithm the DCT Inter-transformation technique is used to increase the detection accuracy without inverse DCT operation. In the proposed scheme, firstly the candidate smoke regions are estimated using motion and color smoke properties; next using morphological operations the noise is reduced. Finally the growth properties of the candidate smoke regions are furthermore analyzed through time using the connected component labeling technique. Evaluation results show that a feasible smoke detection method with false negative and false positive error rates approximately equal to 4% and 2%, respectively, is obtained.

A sub-block-based eigenphases algorithm with optimum sub-block size

Several algorithms have been proposed for constrained face recognition applications. Among them the eigenphases algorithm and some variations of it using sub-block processing, appears to be desirable alternatives because they achieves high face recognition rate, under controlled conditions. However, their performance degrades when the face images under analysis present variations in the illumination conditions as well as partial occlusions. To overcome these problems, this paper derives the optimal sub-block size that allows improving the performance of previously proposed eigenphases algorithms. Theoretical and computer evaluation results show that, using the optimal block size, the identification performance of the eigenphases algorithm significantly improves, in comparison with the conventional one, when the face image presents different illumination conditions and partial occlusions respectively. The optimal sub-block size also allows achieving a very low false acceptance and false rejection rates, simultaneously, when performing identity verification tasks, which is not possible to obtain using the conventional approach; as well as to improve the performance of other sub-block-based eigenphases methods when rank tests are performed.

Improvement of a Video Smoke Detection Based on Accumulative Motion Orientation Model

Early fire-alarming is very important to avoid serious human being and materials losses. The traditional sensor-based methods can detect fire when the situation already has been dangerous. The video-based smoke detection can overcome these drawbacks. This paper proposes improvements of Yuans video-based smoke detection, which employs accumulative motion orientation to detect smoke. In the proposed improvements, optimal thresholds for motion and chrominance detection are established and isolated noisy blocks are eliminated. The motion detection threshold is experimentally determined, and the chrominance detection thresholds are deduced from observation and testing of many videos with or without smoke. The elimination of isolated noisy blocks is achieved using the connected component labeling algorithm, which allows only processing the smoke regions, reducing the computational cost. Experimental results show that the proposed scheme increase the accuracy of the smoke detection and reduce the computation time.

Feature extraction and face verification using Gabor and Gaussian mixture models

This paper proposes a faces verification in which the feature extraction is carried out using the discrete Gabor function (DGF), while the Gaussian Mixture Model (GMM) is used in the face verification stage. Evaluation results using standard data bases with different parameters, such as the number of mixtures and the number of face used for training show that proposed system provides better results that other proposed systems with a correct verification rate larger than 95%. Although, as happens in must face recognition systems, the verification rate decreases when the target faces present some rotation degrees.

Face Recognition Algorithm Using the Discrete Gabor Transform

This paper proposes a face recognition algorithm in which the discrete Gabor transform is used to extract the image face features vector that is then feed into a multilayer perceptron to be carried out the recognition task. The features vector, estimated using the Gabor transform, presents a small intra-person variation while the inter-persons variation is considerably large. This fact provides robustness against changes in illumination, wardrobe, facial expressions, scale, and position inside the captured image, as well as inclination, noise contamination and filtering. Proposed scheme also provides some tolerance to changes on the age of the person under analysis. Evaluation results using the proposed scheme with identification and verification configurations are given to show the desirable features of proposed algorithm

Cross View Gait Recognition Using Joint-Direct Linear Discriminant Analysis

This paper proposes a view-invariant gait recognition framework that employs a unique view invariant model that profits from the dimensionality reduction provided by Direct Linear Discriminant Analysis (DLDA). The framework, which employs gait energy images (GEIs), creates a single joint model that accurately classifies GEIs captured at different angles. Moreover, the proposed framework also helps to reduce the under-sampling problem (USP) that usually appears when the number of training samples is much smaller than the dimension of the feature space. Evaluation experiments compare the proposed framework’s computational complexity and recognition accuracy against those of other view-invariant methods. Results show improvements in both computational complexity and recognition accuracy.

Detección de Movimiento de Vehículos en Secuencias de Video Basados en la Diferencia Absoluta entre Fotogramas y la Combinación de Bordes

Resumen Se presenta un metodo para la deteccion de movimiento en secuencias de video, mediante la combinacion de la Diferenciacion Absoluta entre dos Fotogramas (DAF) y el analisis de bordes en la region considerada movimiento en el tiempo t. Esto permite resolver el fenomeno de apertura del primer plano al unir las regiones que pertenecen a un objeto en movimiento, una de las principales desventajas de la DAF. Las imagenes que son consideradas objetos de interes se someten a operaciones morfologicas para eliminar objetos con componentes conectados menores a un umbral dado. Se realiza una segunda diferenciacion absoluta de fotogramas considerando solo las regiones determinadas con movimiento en la primera diferenciacion, de manera que al combinar los bordes con la deteccion de movimiento de la segunda diferencia absoluta, se obtiene la imagen del primer plano. El algoritmo desarrollado tiene la ventaja de presentar la forma y principales parametros del objeto en movimiento.

Social Sentiment Sensor in Twitter for Predicting Cyber-Attacks Using ℓ1 Regularization

In recent years, online social media information has been the subject of study in several data science fields due to its impact on users as a communication and expression channel. Data gathered from online platforms such as Twitter has the potential to facilitate research over social phenomena based on sentiment analysis, which usually employs Natural Language Processing and Machine Learning techniques to interpret sentimental tendencies related to users’ opinions and make predictions about real events. Cyber-attacks are not isolated from opinion subjectivity on online social networks. Various security attacks are performed by hacker activists motivated by reactions from polemic social events. In this paper, a methodology for tracking social data that can trigger cyber-attacks is developed. Our main contribution lies in the monthly prediction of tweets with content related to security attacks and the incidents detected based on ℓ1 regularization.

Compact image steganalysis for LSB-matching steganography

In this paper, we propose a compact image steganalysis method for the LSB-matching steganography, in which a feature vector composed by only 12 elements is extracted from the image. We analyze the statistical artifact occurred in images when a secret data is embedded in it by the LSB-matching steganography. We selected 12 most relevant features based on the probability density function (PDF) of difference of adjacent pixels and the co-occurrence matrix of the image, which can distinguish stegoimages from the natural images. The Support Vector Machine (SVM) is employed as classifier using the training vectors with 12 elements. The experimental results show that the proposed scheme provides a better discriminate performance than previously proposed methods that require a larger amount of feature elements, such as 27, 35 and 225 feature elements, respectively, for their discriminations.

A Growing Cell Neural Network Structure for Off-Line Signature Recognition

The signature recognition is a topic of intensive research due to its great importance, among others, in the financial system. However it does not exist yet an enough reliable method for signature recognition and verification, especially in the forgeries detection. This paper presents an off-line signature recognition using features extracted from the off-line signature and an array of five growing cell neural network. The proposed system was evaluated using 950 signatures of 19 different persons. Experimental results show that proposed system achieves a fairly good recognition rate with a relatively low computational complexity.