Research On Illumination Consistency Technology Of Augmented Reality Based On3D Shadow And Background Fusion

In the multimedia field, augmented reality (AR) technology has become one of themost popular key technologies, in which illumination consistency has increasinglybecome the research focus. In recent years, although AR technology is widely used inentertainment, exhibition and military, it is still evolving with some shortages. Forexample, the virtual and real scene is not unified on visual effects. So the research onillumination consistency of AR is very necessary.In this thesis, after the domestic and abroad research status about illuminationconsistency of AR are summarized, as well as the main problems of the field areanalyzed, the focuses then are on the studies of inverse detection of illumination andshadow generation technology.Firstly, the inverse detection algorithms based on the illumination model are widelyused in augmented reality, however they demand the surface of the calibration object tobe purely diffuse and detection errors of the light source's position are also relativelylarge. To solve this problem, a two-step detection algorithm of the light source isproposed, which imports3-D features of the shadows and specular reflection. Thealgorithm is divided into two parts, including position detection of the light source andintensity detection of the light. Under the condition of having known the position of thelight source, the Cook-Torrance model with the specular reflection is introduced, andsuch methods as random sampling partition and error handling are also adopted. Besides,the3-D features of the shadows are used instead of the plane corner features, and theposition of the light source is detected according to the ray tracing. Experiment resultsshow that the position detected by this algorithm is more accurate. Therefore, it solvesthe inaccuracy in detecting the position of the point light source and the limitations inthe calibration objects.Secondly, since soft shadow achieved by existing shadow generation algorithms ofAR is unrealistic, a shadow generation algorithm using adaptive sampling andbackground fusion is proposed. The shadow spatial location distribution of virtualobjects is firstly computed by using planar shadow algorithm which takes occlusion intoaccount. Then, to improve the procedure of soft shadow generation in swell and erodealgorithm, an adaptive sampling method which gets illuminant union according to shapetypes is presented. Finally, since shadow color gotten by gray image method is limited, a method based on multi-channel and background fusion is presented. Experimentresults show that in this algorithm the color of soft shadow is more reasonable and themethod of soft shadow rendering is more effective. Consequently, the new algorithmimproves the realism of soft shadow.Finally, a prototype system of AR based on marker and illumination technologies isdeveloped. The system uses the key technologies including3D registration andillumination consistency.