Research On Camera Planning In 3D Animation Generation

With the development of computer graphics theory and related hardware,the effect of rendering pictures in computer becomes more and more excellent,even enough to achieve the effect of real pictures,which makes the production of games and animations higher and higher,and also provides strong support for the later production of movies.At the same time,there are also rich research results in the field of artificial intelligence creation.For example,AI completes the writing of poetry,painting,composing and other artistic creations.Therefore,using artificial intelligence technology to automatically generate 3D animation has become a hot topic.Camera planning is an important research direction in this topic.The main problems faced by virtual camera planning in 3D animation generation are as follows: First,the description of the movie lens,that is,extracting quantitative elements from the existing movie pictures,used to analyze the composition,transition and arrangement of the movie lens,which can be used to analyze the existing movie style or as a guide in film creation.A complete chronological description of all the lenses in the movie is a shooting scheme that serves as a blueprint for controlling virtual cameras for shooting.The second is to generate a shooting plan from the scene and plot information,which is similar to the work of the director in the actual film shooting,that is,to choose what kind of picture to display the corresponding scene and plot.The third is to use a virtual camera to complete the shooting work that meets the requirements of the shooting program,which is similar to the work of a cinematographer.There are a lot of researches and experiments on the problems in camera planning.The results show that the existing algorithms can not get the practical degree of camera planning results.The main difficulty is that the film photography work is inherently highly complex and subjective,can not use a simple algorithm to solve.This paper studies this problem and obtains a complete set of methods that can complete the camera planning automatically based on the scene and plot information in 3D animation.The details are as follows:(1)Based on the photographic primitives proposed by Academician Lu Ruqian and prose storyboard language,a shooting plan description language can be obtained by excluding contents irrelevant to the camera and adding detailed description of starting and ending frames.This description language is not only the result of shooting program planning,but also the condition of camera parameter calculation.(2)In view of the insufficiency of flexibility in rule-based method and knowledgebased method,this paper adopts probabilistic soft logic method to complete the shooting plan planning,and studies a complete set of processing process.(3)By combining the existing geometric calculation method,particle swarm optimization algorithm and centripetal catmull-rom spline,a method is obtained that can calculate camera parameters in various scenes,and can deal with camera collision and occlusion problems.The experiment of the method studied in this paper was conducted based on Unity3 D game engine,and the results show that this method can completely solve the camera planning problem in 3D animation generation in non-real-time scene.In addition,the method in this paper is also extensible and can be further studied on the basis of it,such as the completion of lighting and color planning.