Key Technology Research Of Augmented Reality To Assistance Maintenance

Augmented reality technology could be applied into the complex equipments maintenance,which provides users with the more intuitive and flexible operation style and the fully immersive maintenance environment mixing virtual with real world.It also has great significance to shorten the equipment maintenance cycle,improve the maintenance efficiency and reduce the maintenance cost.Furthermore,the research of key technology to augmented reality system to assistance maintenance has the important guiding significance and reference value of improving the natural features of 3D objects recognition,tracking and registration,human-computer interaction or related algorithm,and the optimizing of the system design and implementation.Based on the analysis of the current development of augmented reality to assistance maintenance technology,several related key technologies including natural features recognition,tracking and registration,human-computer interaction,and evaluation method were studied in the realization processing of augmented reality prototype system to assistance maintenance and some innovative research results were obtained.For the problem of multiple vision-angle fast recognition of maintenance objects in augmented reality assistance maintenance scenes,the edge orient histogram of a 2D image was used to describe their global typical features,and the corner points were used to describe the appearance features of uniqueness and diversity.As a result,the natural features recognition method of maintenance objects was built based on the corner-edge composite feature.Furthermore,the classifier was used to train composite features to get multi-classification predition model.Intra-class matching method based on K-D tree was used to locating objects accurately.Therefore,it was realized to quickly identifying and locating the maintenance objects,which laid the foundation for the study of maintenance objects tracking and registering in the augmented reality assistance maintenance system without markers.In order to solve the tracking and registration problem of maintenance component in the augmented reality to assistance maintenance system without markers,a tracking algorithm of Region of Interest(ROI)which produced by means of the identifying and locating result was proposed based on multiple compressive features within the particle filter framework.The multi-feature presentation model was combined by extracting the texture features and grayscale average features of maintenance objects.The particle filter was used to track the region of interest,and the reference feature vectors were updated as soon as the the great changing of ROI occurred in the process of tracking,which solved effectively the low tracking accuracy and the poor adaptability to target shape change of the basic particle filter tracking method,and improved the real-time performance and the robustness of tracking and registration.Meanwhile,the characteristics were constructed by the eigenvalues differences of image random point pairs between object planes in ROIs in video frames,and then the homography matrix was solved to obtain the 3D/2D projection matrix for 3D virtual models' displaying.As result,the effect of virtual information-real maintenance scene synthesis was shown.Aiming at the problem of different gesture recognition in human-computer interaction process in augmented reality assistance maintenance system,the bare hand gesture recognition method was proposed based on the assistance maintenance status analysis.Firstly,the skin accumulation method was used for gesture image segmentation in YCbCr color space.Secondly,gestures were divided into static gestures and dynamic gestures by extracting the state information of gesture shape and position.Meanwhile,static gestures features based on image properties and dynamic gestures based on direction coding were extracted respectively.Thirdly,the gesture feature vectors were put into the Support Vector Machine(SVM)to train and get the prediction model of the gesture type.Therefore,human-computer interaction function including paging,selecting,stretching and scaling could be implement effectively by capturing,analyzing,processing and recognizing the various kinds of gestures instruction information under the rich gesture definition.Because the existing method lacked of the completeness of evaluation indexes and the systemic verification method to augmented reality assistance maintenance prototype system,a set of ARSAM prototype system was built in lab condition.The miniaturized optical system design of OST-HMD and the light weight free-form-surface system design were the important research content.Furthermore,a group of subjective assessment experiment was designed to observe the maintainability of ARSAM compared to the traditional maintenance method based on the equipment maintenance service manual with actual maintenance task.The experimental results were beneficial to the optimizing design of ARSAM and the improved measure of interactive action.The paper would have an important guiding significance for the key technologies research to the natural characteristics recognition,tracking and registration,gesture recognition,and the system implementation during augmented reality assistance maintenance prototype system.Meanwhile,it also provides the theoretical significance and the practical value for augmented reality technology used in complex equipment maintenance,which can improve the maintenance support efficiency.