An Evaluation of Human Detection Methods on Camera Images in Heavy Industry Environments
Nico Zengeler, Matthias Grimm, Colja Borgmann, Marc Jansen, Sabrina Cornelia Eimler, Uwe Handmann · 2019
In this paper we evaluate different machine learning models for human body detection in heavy industry environments. Contributing a framework to asses the reliability of a detection system in industrial environments, we compare techniques of feature extraction for support vector machines to artificial neural networks. To accommodate for common environmental challenges in heavy industry, such as dust, difficult light conditions and partially covered persons, we apply programmatic changes to our test image set and evaluate the accuracy of person detection, foot point estimation and the tendency of erroneous detections.