Region of Interest-Based Depth Correction and Multipath Interference Mitigation in Time-of-Flight Cameras for Short-Range Automation: development of a Local Filtering Pipeline for Adaptive Region of Interest Detection and Multi- path Interference Reduction

Abstract

In automation technology, it is important to have a reliable depth measurement for short range automation tasks like quality inspection, component verification and object alignment. To get cost efficient depth measurements the industry uses time-of-flight cameras. Most of the time there is a highly reflective environment in the automation industry which induces an error by multipath interference. This systematic error is caused by indirect light from reflective surfaces. Most of the time-of-flight cameras in the industry are low-cost sensors like the Basler Blaze-101 camera. These cameras do not include a hardware correction for the error caused by multipath interference. This limits their accuracy which makes them less suitable to use all over the industry. To get around this error caused by multipath interference a software-based approach is developed. This approach detects and adaptive region of interest and reduces the error caused by multipath interference through a local filter. The aim of the work is to analyse the error caused by multipath interference on the Basler Blaze-101 camera and to design a local filter for the multipath interference error on an adaptive region of interest. As a software MATLAB is used due to its accessibility when it comes to depth data and filtering methods. The final filter has a shape adaptive region of interest selection and a bias correction with exponential and temporal smoothing of the depth data. The performance of the filter is evaluated with multipath interference and no multipath interference datasets of industrial automation scenarios, like object height verification and product sealing checks. The results demonstrate that the proposed filter successfully eliminates systematic bias relative to the fitted plane and reduces temporal depth variation by nearly an order of magnitude. Within the defined region of interest, the corrected data achieves millimeter accuracy and high repeatability over multiple frames. Validation experiments with industrial reference objects confirm that the method provides stable and reliable depth measurements even under complex reflection conditions. The developed pipeline therefore offers a cost-efficient and sustainable software solution for improving time-of-flight depth sensing performance in automated measurement and inspection systems.