Publicação
Object detection and localization: an application inspired by RobotAtFactory using machine learning
| Resumo: | The evolution of artificial intelligence and digital cameras has made the transformation of the real world into its digital image version more accessible and widely used. In this way, the analysis of information can be carried out with the use of algorithms. The detection and localization of objects is a crucial task in several applications, such as surveillance, autonomous robotics, intelligent transportation systems, and others. Based on this, this work aims to implement a system that can find objects and estimate their location (distance and angle), through the acquisition and analysis of images. Having as motivation the possible problems that can be introduced in the robotics competition, RobotAtFactory Lite, in future versions. As an example, the obstruction of the path developed through the printed lines, requiring the robot to deviate, and/or the positioning of the boxes in different places of the initial warehouses, being positioned so that the robot does not know its previous location, having to find it somehow. For this, different methods were analyzed, based on machine leraning, for object detection using feature extraction and neural networks, as well as object localization, based on the Pinhole model and triangulation. By compiling these techniques through python programming in the module, based on a Raspberry Pi Model B and a Raspi Cam Rev 1.3, the goal of the work is achieved. Thus, it was possible to find the objects and obtain an estimate of their relative position. In the future, in a possible implementation together with a robot, this data can be used to find objects and perform tasks. |
|---|---|
| Autores principais: | Pilarski, Leonardo |
| Assunto: | Machine learning Neural networks Object detection Localization Robotics competition |
| Ano: | 2023 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Instituto Politécnico de Bragança |
| Idioma: | inglês |
| Origem: | Biblioteca Digital do IPB |
| Resumo: | The evolution of artificial intelligence and digital cameras has made the transformation of the real world into its digital image version more accessible and widely used. In this way, the analysis of information can be carried out with the use of algorithms. The detection and localization of objects is a crucial task in several applications, such as surveillance, autonomous robotics, intelligent transportation systems, and others. Based on this, this work aims to implement a system that can find objects and estimate their location (distance and angle), through the acquisition and analysis of images. Having as motivation the possible problems that can be introduced in the robotics competition, RobotAtFactory Lite, in future versions. As an example, the obstruction of the path developed through the printed lines, requiring the robot to deviate, and/or the positioning of the boxes in different places of the initial warehouses, being positioned so that the robot does not know its previous location, having to find it somehow. For this, different methods were analyzed, based on machine leraning, for object detection using feature extraction and neural networks, as well as object localization, based on the Pinhole model and triangulation. By compiling these techniques through python programming in the module, based on a Raspberry Pi Model B and a Raspi Cam Rev 1.3, the goal of the work is achieved. Thus, it was possible to find the objects and obtain an estimate of their relative position. In the future, in a possible implementation together with a robot, this data can be used to find objects and perform tasks. |
|---|