Comparison and Evaluation of varying Noise- and Cross-Validation-Score thresholds for the Two-Step Conditional Filter Band Selection

Description

The majority of the plastic waste produced is thermally recycled, i.e. incinerated. The reason for this is that the plastics often cannot be properly identified. Multispectral imaging enables clear classification. So-called camera arrays are used for this purpose; filters in different wavelengths in the infrared range are fitted in front of the individual cameras. As plastics reflect and absorb long-wave light differently due to their different carbon chain structures, different materials appear in different colors. However, the specific wavelengths of interest have to be found first, which enable a good classification accuracy.

Details can be found in the following document:

Topic_BA

Prerequisites

Experience in programming with Python as well as image and video compression is required. Additional knowledge in Deep Learning (PyTorch/TF) is useful.

Category: Bachelor Theses – Offered