Point Pair Feature Based Object Detection for Random Bin Picking

Point pair features are a popular representationfor free form 3D object detection and pose estimation. Inthis paper, their performance in an industrial random binpicking context is investigated. A new method to generaterepresentative synthetic datasets is proposed. This allows toinvestigate the influence of a high degree of clutter and thepresence of self similar features, which are typical to ourapplication. We provide an overview of solutions proposedin literature and discuss their strengths and weaknesses. Asimple heuristic method to drastically reduce the computationalcomplexity is introduced, which results in improved robustness, speed and accuracy compared to the naive approach.