Research Of Target Recognition And Localization In CNC Machining Simulation Base On Augmented Reality

The CNC machining simulation technology is always at the foreland of the related research fields as one of core parts of the CNC technology. The CNC machining simulation based on virtual reality emphasizes that the users immerse in a virtual environment producted by computer, so this isolates completely the users and the real machining environment.This problem can be solved effectively when Augmented Reality (AR) technology is application in the area of CNC maching simulation. AR technology merges the virtual enviroment producted by computer into the real world around users,so it can enhance the user the ability to sense the real world.This paper discusses deeply the target recogintion and localization technology in AR faced CNC machining simulation.At first,based on introduction AR technology,CNC machining simulation based on AR is brought out,and the frame of its system is put forward further.among them, in focus,the target recogintion and localization technology is described in detail.According to the particularity of the CNC system,we bring out the target recogintion and localization based on the nature feature of the target.In target recogintion and localization based on nature feature, in view of special enviroment ,this paper describes in detail this system image processing link:image gathering,image pretreatment,feature detection.We summarize and analyze merits and faults of the each existent algorithm,at the same time improve some of algorithms,so we can get the most excellent algorithms adapt to this system.Then this paper focus on the principle and progress of the target recogintion and localization method based on nature feature.For target recognition ,we first bulid the hash table through training image,and then search this hash table using the target advantage feature,at last, vote the target model to accumulate. Target localization achieve mainly according as the projection matrix between the target model and real-time image.we can get the projection matrix through the matching point between the target model and real-time image.In this paper,we link the target model to real-time image through training image as an intermediary.At last, an application instance whose target uses the machine clamping fixture comes true to achieve target recognition and loaclization.This instance verifies the feasibility that the target recognition and localization applies to CNC machining simulation.