Rectángulo redondeado: Research

My research interest are around VLSI Design, Computer Architecture, Image Processing, Smart Surveillance Systems, and Data Fusion. 

Image Fusion
The Electromagnetic spectrum is very wide and cannot be capture by one single type of sensor. Therefore, each kind of sensor captures a very specific band of the spectrum. One way to improve the analysis of the information is to fuse images coming from different sensors. Then, a better look at the problem can be achieved. I worked in a new image fusion technique that can be used in different applications such as remote sensing, surveillance systems, computer vision and medical applications. 

	ImageFusion.org


Data Fusion in Wireless Sensor Network
Wireless sensor network consists of several hundreds of nodes that are composed of sensing unit, processing unit, communication and power unit. Each sensor must sent its information towards a base station where all data is processed. However, the most expensive operation in terms of power that a node does is data transmission. Therefore, the node must eliminate the redundant information before the data is transmitted to save energy. I worked in a data fusion algorithm to save energy and facilitate processing data in the base station. 

	Data fusion Server


Object Detection for Surveillance.
The first block in surveillance system processing flow is the object detection. There are two techniques used to perform this task, optic flow and background subtraction. However, optic flow requires a huge amount of resources and its difficult to implement in real time. Background subtraction is a better option because its low complexity compared with Optic Flow. These techniques have been implemented in software. However, software platforms become an expensive solution when the number of cameras grows. A Hardware-software platforms offer a better solution. In this scenario, software implementation of background subtraction algorithms is not a suitable solution. Therefore, we developed two hardware implementation of the Wronskian Change Detector. The background subtraction algorithms where studied using the Wallflower benchmark.

	Wallflower benchmark


Object Tracking using Visual Information.
Once the object has been detected, the next step in the processing flow is tracking. I worked in a surveillance system that uses an agent based tracking system. The multi-agent system has two class of agents, region and object agent. The area under surveillance is divided into regions. Each region is assigned to a region agent, who is in charge of monitoring the region and to assign an object agent to each object that is inside his region. The object agent is the one that actually tracks the object of interest.
 


 

Universidad Autónoma de San Luis Potosí

School of Science

Electronics Department