New approach to motion analysis useful for Parkinson's disease, aircraft engines and sport
Small motions are difficult to observe with the naked eye. A team of TU Delft researchers will present an improved method for achieving this at the CVPR conference in Honolulu, Hawaii on Saturday, 22 July. The technique has potential for use in analysing tremors in Parkinson's disease, for example.
The work of Jan van Gemert and his colleagues in the Delft Vision Lab involves magnifying small motions in video footage. The possibility to magnify (or reduce) subtle changes can be used in all kinds of fields, including video editing. However, it is also useful for medical video analysis, product quality control and in sport, for example.
‘In these cases there is often large motion present, which distorts the current methods for magnifying small changes in video footage’, says Van Gemert. ‘This is why we have developed a new approach to dealing with these large motions. What it comes down to is only magnifying acceleration. We just ignore linear motion.’
‘We have tested the method on moving objects and now have quantitative and qualitative proof that our method works better than what is currently considered to be state-of-the-art, a technique developed by MIT’, says Van Gemert. ‘Our techniques do not involve particularly stringent requirements. The images can simply be produced using an iPhone.’
The new techniques could prove useful in countless different areas. The TU Delft researchers have already tested their method successfully in various fields. Examples cited by Van Gemert include analysis of tremors in Parkinson's disease, monitoring heartbeat and respiration (e.g. in babies), drone flights, vibrations in aircraft engines and sport.
On Saturday, 22 July, Jan van Gemert will present a paper on motion analysis at the CVPR conference (21-26 July in Honolulu, Hawaï). CVPR is the premier annual event in the field of computer vision.