This work begins with some basic computer vision techniques that can be very accurate, but only under certain conditions. Next, a more complex and versatile method of machine learning is proposed (Cascade Classifiers), which can be used in a multitude of situations. However, the Cascade Classifiers method still faces difficulties if a simple symbol with too few features needs to be recognized. Therefore, another technique is proposed that uses artificial neural networks to find symbols in large technical drawings. This method includes training that requires a powerful graphics processing unit (GPU), but shows very promising results. Ultimately, a graphical user interface (GUI) is created for the program which enables simple operations without in-depth programming knowledge.
«
This work begins with some basic computer vision techniques that can be very accurate, but only under certain conditions. Next, a more complex and versatile method of machine learning is proposed (Cascade Classifiers), which can be used in a multitude of situations. However, the Cascade Classifiers method still faces difficulties if a simple symbol with too few features needs to be recognized. Therefore, another technique is proposed that uses artificial neural networks to find symbols in large...
»