The most important input for development and evaluation of crash protection is real life accidents data. The data is however time consuming to collect and the output is in relation to what can be measured in laboratory testing, primitive. Important parameters are often collected by untrained people as secondary task in the rescue work at the accident scene. The precision and accuracy of the data can therefore often be questioned and the possibilities to draw conclutions are often limited. By using modern technology, such as photogrammetric measurements of exterior and interior of the vehicle deformations, cheap crash pulse recorders (CPR), and by training workshop and rescue personal, the possibility to collect high quality accident data can be dramatically increased. In this presentation, such a comprehensive system is described, in terms of potentials and possible output, as well as a theoretical background for increasing precision of collected data.