A method to enhance the accuracy of time of flight measurement systems

摘要

A ToF-camera operates with modulated visible and near-infrared radiation, which is detected andrndemodulated simultaneously by a 2-D array of pixels. The phase shift between the sent signal and thernreceived signal is an indicator for the object's distance. The amplitude of the radiation gives additionalrninformation for its reflectance. Unfortunately, measurements of distances performed with almost allrncameras available on the market show very high deviations. The main reason for them seems to be therninfluence of stray light, which is detected by the camera with the modulated radiation. Another source ofrnuncertainties is the low sampling rate along the received signal. In the first part of this paper, experimentsrnusing different cameras and alternating environments have been performed in order to determine therndeviations, which are up to 20 cm over a distance of 150 cm. A survey of existing cameras as well asrntheir performance has taken place. In the second part of the present work two approaches for therneliminations of the deviations have been introduced. In the first approach spline functions have beenrncalculated using reference measurements for distances between 5 cm and 200 cm. Those functions arernused for the calculation of the uncertainties of each pixel. Alternatively the splines can be replaced by arngain matrix which is multiplied with the image matrix in order to equalize the error of the distancernmeasurement of the pixel. Subsequently only one spline is needed in order to adjust the distances. In thisrncase the calculation is faster but it cannot be applied in every case. The reflectance of the objects in thernscene has not been taken into account yet. The opposite happens in the second approach in order tornenhance the accuracy of the measurement. This is done by estimating the reflectance of the nearestrnobject in dependency of the measured amplitude and distance. The reflectance of the background,rnrespectively everything behind the nearest object, is estimated by using the measured amplitude of thernbackground and the distance of the nearest object and the background. By using both reflectance factorsrnand the distances as indicators for further smoothing functions the error of measurement due to thernreflectance can be determined. After applying both methods the error of measurement decreased in ourrntests from about 20 cm to 3 cm.