METHOD FOR THE MODEL-BASED COMPILATION OF TEST DATA SETS FOR EXAMINING IMAGE PROCESSING ALGORITHMS

摘要

The invention relates to a method for compiling test data sets (50) for examining an image processing method (80) or a sensor (91), wherein the image processing method (80) and/or the image processing algorithm (80) is a part of said sensor, wherein the image-processing method (80) to be examined and/or the image-processing algorithm (80) to be examined is designed to determine result values (7) from digital images, wherein a domain model (1) is determined by means of which, using a number of pre-settable parameters (6), typically three-dimensional scenes (2) can be defined, three-dimensional scenes (2) are compiled by establishing parameters (6) of the domain model (1), wherein for each of the scenes (2) the position and viewing direction of an observer (120) in relation to the particular three-dimensional scene (2) represents a further parameter (6) which, together with the remaining parameters (6), is defined for the respective scene (2), a digital image (5) of each scene (2) is compiled starting from the position and orientation of the observer (120) by means of a rendering method (40), comparison values (4) are determined or derived from the scene (2), and the comparison values (4) are stored together with the digital images (5) as a common data set (50).