Airbag volume measurements

Download this article as a PDF here

In the automotive industry there is a need to measure the volume of an airbag during the inflation process. There are several reasons for this: to verify the inflation
process, to compare with simulated models, and to perform quality testing during the production. The main features to be measured are the volume, area and position at different times. 

The methods for measuring used today are very time-consuming and complex. Special markers are attached to the airbag before the test and the final analysis includes many operator interactions before a final result can be presented. For a long time the industry has wanted an alternative method with shorter preparation time,  ecreased complexity and high accuracy. A new solution from Image Systems in Sweden now provides this by measuring the volume of an unprepared airbag in a simple and fast process by using multiple cameras. The airbag is imaged, during the course of inflation, from at least three cameras simultaneously. The cameras are positioned as in a standard 3D test setup, which means that all cameras must be positioned (X, Y, Z, roll, pitch and yaw) in a common lab coordinate system. The cameras must be calibrated due to lens distortion and the actual focal length must be measured to achieve an accurate result. All this is done in the same way as for a standard 3D measurement setup with all the cameras synchronized when recording the event. 

After the event, the image sequences from all the cameras are loaded into the TEMA software package for analysis. The new technique, derived from computer vision research, is based on identifying the boundary of the airbag in each camera view. The algorithm used to detect the boundary is based on edge detection, which decreases the sensitivity to varying light, and the effect from static disturbances is eliminated using background image subtraction. By combining the boundaries of the same airbag from different views, it can be reconstructed in 3D space and the volume can be estimated.

From Visual Hull theory, it is known that the airbag is ‘over-sized’. However, this ‘error’ has been determined in simulations, and is based on a number of parameters, the most influential being the number of cameras used. After compensating for these factors, the accuracy of the volume for a fully inflated airbag is +/-1%, although in the early stage of inflation the accuracy is less.