Python-Based Image Recognition Techniques for Testing vehicle IPC in HIL environment

The world of production has been strongly influenced by the spread of Artificial Intelligence (AI) technologies in the last decade. The concept of automation of repetitive and mechanical operations has been a focal point for manufacturing companies in order to reduce human error, human effort and increase safety and speed of production. AI in particular have revolutionized many fields and Machine Learning (ML) is one of its most important applications which introduces the possibility to learn to the system from dataset and execute operations even if it’s not programmed to do them. Object detection is a specific task of ML which requires Convolutional Neural Networks. Object recognition provides improvements in the automotive field not only for the manufacturing processes but even in testing operations. ECUs are mechatronic devices that are tested by means of HIL (Hardware-in-the-Loop) environments because dangerous or complex conditions for the test could present. During these tests some actuators are usually implemented in the hardware configuration which are controlled by ECUs’ signals. The correct motion and timing of activity need to be controlled by an operator who is in charge to declare if test executed at the HIL is successful or not. This boring and repetitive activity could be automated by the implementation of a camera with the goal of detecting the presence of specific motions of mechanical components or, like in the case analyzed in this thesis activity, check if the warning lights of a screen switch on after a fault has been set. To do that Python has been identified as the best high-level programming language to achieve a powerful and reliable instrument due to its libraries and it allows to minimize effort even for the control of the signal transmitted to the ECU of interest. In this thesis activity an Instrument Panel Cluster (IPC) has been used and has been implemented a Python code which allows to edit the CAPL (CAN Access Programming Language) on VectorCANalyzer during simulations in a HIL environment. The final results expected are a tool able to recognize specific warning lights on the IPC, which is controlled by means of a Python code during a test and, at the end of it, determine if the test is passed or not. The performance of the camera results to be fundamental for the production of an efficient product. That’s why the camera’s software has been tested in order to get the lowest latency possible between the CAN signal modification and the image recognition.