Publication: Detección de objetos en movimiento con sensor LIDAR
Loading...
Identifiers
Publication date
2021
Defense date
2021-07
Authors
Advisors
Tutors
Journal Title
Journal ISSN
Volume Title
Publisher
Impact
Export
Abstract
La detección de objetos en movimiento es una tarea fundamental en el mapeado, navegación y seguimientos de objetos. LIDAR es uno de los sensores más reconocidos en la tarea de detectar objetos en movimiento mediante la captura de nubes de puntos de una escena, destacando su función en aplicaciones como robótica, videovigilancia, automatización de vehículos, etc.
El presente trabajo consiste en crear un sistema computacional capaz de detectar objetos en movimiento con un sensor LIDAR.
Para alcanzar este objetivo, se diseña y desarrolla un dispositivo que permita la recopilación de información del entorno. Se utiliza una Raspberry Pi que conecta un LIDAR de corto alcance y un conjunto de sensores (GPS, cámara, barómetro, acelerómetro, giroscopio y magnetómetro). Con este dispositivo se puede registrar información del entorno proporcionada por los distintos sensores con el fin de ser tratado en diferido. Se realiza una agrupación de las nubes de puntos con el algoritmo de clustering DBSCAN, una segmentación de los datos utilizando la técnica de SMOTE y una asociación de las nubes de puntos mediante algoritmo ICP con el fin de detectar el movimiento. Además, se realiza una transformación de los puntos obtenidos por el LIDAR al sistema de coordenadas geodésicas mundial WGS-84, de esta manera se puede expresar los datos obtenidos en posición con respecto al elipsoide de referencia.
Para verificar el movimiento generado en escena se ha desarrollado un sistema comparativo que detecta el movimiento en las imágenes capturas por la cámara que integra la Raspberry Pi, esta detección se basa en la sustracción de imágenes.
Este trabajo puede ser utilizado como punto de partida para aplicaciones relacionadas con detección y escaneado de escenas mediante un dispositivo.
El resultado obtenido finalmente es un dispositivo capaz de registrar conjuntos de datos y un sistema que detecta el movimiento en nubes de puntos pudiendo representarlas en WGS-84.
The detection of moving objects is a fundamental task in the mapping, navigation and tracking of objects. LIDAR is one of the most recognized sensors in the task of detecting moving objects by capturing point clouds from a scene, highlighting its role in applications such as robotics, video surveillance, vehicle automation, etc. The present work consists of creating a computational system capable of detecting moving objects with a LIDAR sensor. To achieve this goal, a device that allows the collection of information from the environment is designed and developed. A Raspberry Pi is used to connect a short-range LIDAR and a set of sensors (GPS, camera, barometer, accelerometer, gyroscope and magnetometer). With this device, information from the environment provided by the different sensors can be recorded in order to be processed in deferred. A clustering of the point clouds is performed using the DBSCAN clustering algorithm, a segmentation of the data using the SMOTE technique and an association of the point clouds using ICP algorithm in order to detect motion. In addition, a transformation of the points obtained by the LIDAR to the WGS-84 world geodetic coordinate system is performed, in this way it is possible to express the data obtained in position with respect to the reference ellipsoid. To verify the movement generated in the scene, a comparative system has been developed to detect the movement in the images captured by the camera that integrates the Raspberry Pi, this detection is based on the subtraction of images. This work can be used as a starting point for applications related to detection and scanning of scenes through a device. The final result is a device capable of recording data sets and a system that detects movement in point clouds and can represent them in WGS-84.
The detection of moving objects is a fundamental task in the mapping, navigation and tracking of objects. LIDAR is one of the most recognized sensors in the task of detecting moving objects by capturing point clouds from a scene, highlighting its role in applications such as robotics, video surveillance, vehicle automation, etc. The present work consists of creating a computational system capable of detecting moving objects with a LIDAR sensor. To achieve this goal, a device that allows the collection of information from the environment is designed and developed. A Raspberry Pi is used to connect a short-range LIDAR and a set of sensors (GPS, camera, barometer, accelerometer, gyroscope and magnetometer). With this device, information from the environment provided by the different sensors can be recorded in order to be processed in deferred. A clustering of the point clouds is performed using the DBSCAN clustering algorithm, a segmentation of the data using the SMOTE technique and an association of the point clouds using ICP algorithm in order to detect motion. In addition, a transformation of the points obtained by the LIDAR to the WGS-84 world geodetic coordinate system is performed, in this way it is possible to express the data obtained in position with respect to the reference ellipsoid. To verify the movement generated in the scene, a comparative system has been developed to detect the movement in the images captured by the camera that integrates the Raspberry Pi, this detection is based on the subtraction of images. This work can be used as a starting point for applications related to detection and scanning of scenes through a device. The final result is a device capable of recording data sets and a system that detects movement in point clouds and can represent them in WGS-84.
Description
Keywords
Detección de objetos en movimiento, Nubes de puntos