Automotive Camera Modeling and Integration with Standardized Interfaces
- Various technological challenges still prevail when it comes to the development and validation process of advanced driver-assistance systems and autonomous functions. This dissertation aims to introduce new approaches and methods that still adhere to the process flow of the V-cycle, and to address some of the existing development challenges. To satisfy these approaches,
a physical sensor model for an advanced driver assistance camera
has been created, and in the context of the proposed modeling approach, pertinent levels of abstraction and relevant optical and image sensor effects are identified. Additionally, applied concepts and employed methods for the camera model parametrization are presented as well. Moreover, this dissertation illustrates the effectiveness of hybrid-development and test strategies. For the development part, the usage of a camera model in training image-based neural network algorithms is illustrated. Besides,
concerning the test part, several classical computer vision algorithms and image-based neural networkVarious technological challenges still prevail when it comes to the development and validation process of advanced driver-assistance systems and autonomous functions. This dissertation aims to introduce new approaches and methods that still adhere to the process flow of the V-cycle, and to address some of the existing development challenges. To satisfy these approaches,
a physical sensor model for an advanced driver assistance camera
has been created, and in the context of the proposed modeling approach, pertinent levels of abstraction and relevant optical and image sensor effects are identified. Additionally, applied concepts and employed methods for the camera model parametrization are presented as well. Moreover, this dissertation illustrates the effectiveness of hybrid-development and test strategies. For the development part, the usage of a camera model in training image-based neural network algorithms is illustrated. Besides,
concerning the test part, several classical computer vision algorithms and image-based neural network algorithms are tested and evaluated in a dedicated but generic/modular framework. Finally, co-simulation frameworks are presented for the integration and coupling process of various software components (sensor models, highly automated functions and simulation
software) with standardized interfaces like the functional mockup interface and the open simulation interface.…
![show more show more](/opus4-hs-kempten/layouts/opus4/img/arrow_down.png)
![show less show less](/opus4-hs-kempten/layouts/opus4/img/arrow_up.png)