Research On 3D Reconstruction And Measurement Methods Of Indoor Scenes Based On RGB-D Images

The indoor environment is the most common environment in human life.Most of the time in human daily life is indoors.Therefore,when entering a strange environment,people will be eager to understand the environment and location.With the application of augmented reality(AR),virtual reality(VR)and other fields in 3D reconstruction of indoor environment,people's requirements for indoor maps are also increasing.However,due to the small scale and complexity of indoor environment,how to build a highprecision indoor 3D map and quickly measure the size of objects in indoor environment are worth further study.In our paper,simultaneous localization and mapping(SLAM)technology,deep learning and 3D reconstruction method are combined.Firstly,our paper studies the domestic and foreign research status of 3D reconstruction technology and target detection technology.Secondly,the theoretical basis of 3D reconstruction and image measurement is described.Finally,our paper studies the indoor 3D reconstruction algorithm based on RGB-D images and the semantic target measurement algorithm based on target detection,and obtains the following results:Aiming at the problems of low accuracy and immature 3D map construction algorithm in complex indoor environment,an indoor 3D reconstruction algorithm based on RGB-D images is proposed in our paper.Our algorithm can effectively improve the density and accuracy of 3D map construction,solve the camera pose of each image and build the image at the same time,and select key images to build offline RGB-D database.The experimental results show that the image-based 3D reconstruction algorithm in indoor environment can quickly establish 3D dense color point cloud map of indoor corridor environment.And the accuracy of the map is high,which can restore the indoor scene information well,measure the distance between any two points in the map,and increase the readability of the map.Aiming at many kinds of objects in complex indoor environment,complex and timeconsuming selection of objects for image measurement,and low accuracy of measurement in indoor environment,a semantic target measurement algorithm based on target detection is proposed for the first time in this paper.Firstly,the user's RGB image is input,and the image with the highest similarity is selected by ORB feature matching in the offline database for target detection and semantic target measurement.The algorithm we proposed avoids the manual time-consuming of selecting the object manually,and can directly output the size of the object to recognize the type of the object,which provides convenience for the extraction of the object size information in the indoor environment.The experimental results show that the proposed algorithm can effectively identify and measure the size of semantic objects,reduce the time consuming of object measurement,and improve the efficiency of measurement.In summary,the indoor environment 3D reconstruction and semantic measurement algorithm we proposed are simulated and verified in our paper.Experimental results show that the proposed algorithm can accomplish 3D reconstruction and semantic target measurement in indoor scenes.