A survey on depth based 3D image Modeling

Dr. Dayanand G Savakar, Ravi Hosur
2017 International Journal of Engineering and Technology  
Depth based 3D image modeling is a biggest challenge that persists in the digital image processing depending on different depths. The method of recognizing in dynamic areas is an unanswered challenge facing the community of robotics. The answer can be framed by including the methods like segmentation, tracking, and classification components, etc. This paper reveals the survey of the works carried on the construction of 3D model with their own method and technologies applied in the process. The
more » ... aper mainly goals the Reality systems to enhance the techniques used for the 3D image construction and also to segment the blurry areas in the image captured from different depths. Also the image is desirable to recognize 3D-objects of the system in the user's environment, in order to avoid manual based method. Many existing systems are surveyed which perform 3D-object model construction. Keyword-Depth, 3D image, modeling, Reconstruction, Acquisition I. INTRODUCTION Depth based 3D-object construction and its recognition is the most challenging area of research in the DIP (Digital Image Processing). The goal of 3D modeling is to correctly identify objects of different pose when given to a system for identification that are present in a 3D scene of a depth present with range image, and also to be used in the security systems and the military systems, and also intensity data systems, range data systems, and both range and intensity systems (sometimes including color). In cases like the intensity-image systems, points and straight line segments are the considered common features. Recently, the alignment method blindly matches the points or line segments thrice from the image with the similar of the model, using little or no contextual information. These all inventions lead to the assumptions that the points or line segments are the reliable features of the class of the identified object to be recognized or located and (b) the view of the object is determined can be unique using these features. They lead to the view of polyhedral objects or those modelled images with sharp and straight edges. Hence they are categorized under curved-surface objects. Three-dimensional reconstruction of objects has become an important technique in the context of inspection: Industry quality inspection -for observing the quality of a product from different views, Archaeology analysisreconstruction of an object from raw data obtained from muddy area, Medical science -for conduction sessions of teaching, and for reconstruction body check up from collected data like details obtained from body check up, Forensic analysis -reconstruction of a scene from raw data for identifying unseen actions. II. LITERATURE SURVEY (Jose L. Herrera Carlos R, et al., (2015) ) proposes a method for constructing 3D using 2D by using a technique centroid clustering process by combining all the depth images in the cluster. (Zhongjun Wu, et al., (2015) proposes a Multi-Depth Generic Elastic Models (MDGEMs) for the construction of 3D face modeling by using varying multiple depth maps of an image. (Yan Zhou, Huiwen Guo, et al., (2015) ) proposes an algorithm for reconstructing a 3D view using a wavelet transform and Support Vector Machine (SVM) models to get the image focus quality and Mean shift algorithm for the depth scene. The results prove to be improving the efficiency better compared to previous techniques. (Masaya Kaneko, et al., (2015) ) proposes a method for detecting the edge by generating a decision tree, a solution 25 times better to a conventional method existing by training the decision tree using supervised learning method in the process (Xuyuan Xu, et al., (2014) ) recovers the object boundaries of captured depth images with sharp that refine the use of adaptive block by reallocation of the same position and expand to increase the depth accuracy by avoiding false depth boundary refinement.
doi:10.21817/ijet/2017/v9i4/170904117 fatcat:gnrrsxanajfsvlg67ptjgwhwn4