Publication: Desarrollo de un algoritmo de autocalibración de los parámetros extrínsecos de un sistema estéreo para aplicaciones ADAS
Loading...
Identifiers
Publication date
2014
Defense date
2014-07-24
Authors
Advisors
Tutors
Journal Title
Journal ISSN
Volume Title
Publisher
Impact
Export
Abstract
En este proyecto se realiza un estudio sobre la relación del entorno real que rodea a un vehículo con el sistema de la cámara estéreo instalado en él, durante la conducción del mismo con el fin de analizar la calzada por la que circula.
Para ello se ha procedido a implementar un método que analiza las imágenes captadas por una cámara situada en la parte frontal del vehículo y a partir de ellas obtener la parte de la calzada transitable.
En primer lugar la cámara debe calibrarse, esto se realiza de forma constante en cada instante de captura realizado por las cámaras, intentando minimizar los errores debidos a posibles vibraciones que experimente el vehículo en movimiento, lo que nos permitirá obtener resultados más precisos.
Después analizaremos y etiquetaremos los diferentes píxeles de la imagen para obtener así la calzada por la que puede circular le vehículo y diferenciarla de los distintos obstáculos.
Por último realizaremos una comparación entre diferentes métodos, comprobando cuál de ellos cumple los requisitos deseados., Para ello se realizarán unas simulaciones y se analizaran los valores que tomarían dichos parámetros extrínsecos, con el fin de dar validez a cada uno de los métodos utilizados.
The aim of this project is to study the relationship between the real environment of a moving vehicle, and the model generated by a set of cameras installed inside this vehicle. Thus, this generated model would allow us to analyze the environment where the car drives along. With that purpose images were taken from a stereo system placed in front of the vehicle, and afterwards these images were analyzed using a method which development is the core of this project. The calibration of the cameras was implemented using a steady method that continuously takes images from the installed cameras, and then minimizes the errors due to vibration, consequence of uneven surface. This method allowed us to increase the accuracy of the results. Then, every pixel was first analyzed and the tagged accordingly, by differentiating road and different obstacles. Knowing the position of the road, we can obtain the relative position of the stereo system. That is estimated the extrinsic parameters of the stereo system. Finally, model samples were obtained through different algorithms and then compared among them. This analysis allowed us to check the quality of the results depending on the method, and so proceed to validation. The results obtained support heavily the effectiveness of the methods developed in this project.
The aim of this project is to study the relationship between the real environment of a moving vehicle, and the model generated by a set of cameras installed inside this vehicle. Thus, this generated model would allow us to analyze the environment where the car drives along. With that purpose images were taken from a stereo system placed in front of the vehicle, and afterwards these images were analyzed using a method which development is the core of this project. The calibration of the cameras was implemented using a steady method that continuously takes images from the installed cameras, and then minimizes the errors due to vibration, consequence of uneven surface. This method allowed us to increase the accuracy of the results. Then, every pixel was first analyzed and the tagged accordingly, by differentiating road and different obstacles. Knowing the position of the road, we can obtain the relative position of the stereo system. That is estimated the extrinsic parameters of the stereo system. Finally, model samples were obtained through different algorithms and then compared among them. This analysis allowed us to check the quality of the results depending on the method, and so proceed to validation. The results obtained support heavily the effectiveness of the methods developed in this project.
Description
Keywords
Proceso de imágenes, Visión artificial, Algoritmos