A new method of defining objects of interest and 3D visualization in FOTOM NG system

Transcription

1 A new method of defining objects of interest and 3D visualization in FOTOM NG system Lachezar Lichev, Jakub Hendrych, Radim Kunchicky, Karolina Feberova Department of Computer Science, VSB Technical University of Ostrava, Faculty of Electrical Engineering and Computer Science, Ostrava, Czech Republic, Daniela Gotseva Computer Systems Department, FCSC, TU-Sofia, Sofia, Bulgaria,

2 PAPER STRUCTURE INTRODUCTION DEFINING OBJECTS OF INTEREST Basic Tools for Defining Objects of Interest Defining Objects of Interest Using the Radon Transform OBJECT VISUALIZATION IN 3D Transformation Projections Modeling by Using Java3D API CALCULATIONS FOR THE ANALYSIS OF 3D OBJECT RESULTS, DISCUSSIONS, CONCLUSIONS

3 INTRODUCTION The current trend of modern applications is to quickly and accurately analyse the data. Among the types of such analysis we can consider 3D visualization. Thanks to the visualization of the measured object, we get another perspective on examining data than as from tables and graphs. Another big advantage is the preservation of the measurement, and therefore we don t need to measure new one. The system provides the ability to analyse data from the long term - it is possible to compare the measurement and thus observe various changes, such as deformation of mining holes or extra-high voltage pylons. This leads to assuring of other factors such as safety, simplicity, efficiency and others.

4 DEFINING OBJECTS OF INTEREST The defining objects of interest, is a very important element of the FOTOM NG system. For this purpose are implemented a large variety of tools. With these tools, we are able to express real-world objects in images and defined for them the position and size. It should be mentioned that the ratio between the image units (pixels) and real-world units (mm, cm, m) are expressed by Base points - define reference points in image. According to these points is then possible to determine the position of the images in the series, where we observe the same object. It also serves to the mutual positioning of each image in the series.

5 Basic Tools for Defining Objects of Interest When working with measuring image we will most often use those tools, which has an internal surface (circle tool and polygon tool). These tools are suitable for the marking the light areas on the image. Because they have own area, they can be used to calculate the volume and for creating a 3D model. Intersection tool and point tool are from the perspective of the 3D visualization essential for creating polylines from points and two intersecting planes. Polygon tool Circle tool Intersection and Point tool

6 Defining Objects of Interest Using the Radon Transform This new feature in FOTOM NG will allow us to create a polygonal representation of objects using a completely different approach. The main idea of Radon transform is based on the fact that the observed object can be fully reconstructed from a set of projections of the same scene, which contains this object. By the term of projection is meant a set of infinitely narrow focused beam projections acquired at one angle shooting.

7 OBJECT VISUALIZATION IN 3D In 3D modeling, we have the ability to create different complex shapes from the basic ones to the final 3D model. However, each model is composed from elementary parts - a point and a vector. Why do we mention only these two objects? If we focus on some specific API (in our implementation Java3D) for creating 3D models, we find that they always use these two elementary objects. From them are subsequently created others shapes (polygons, etc.). Most of APIs requires for work with points and vectors their coordinates.

8 Transformation If we have created an object in 3D space, we need to manipulate with it - for the interaction or for various calculations that are necessary for proper orientation of the images in the series against the reference profile. Default or reference profile is usually the first image of the series. During the implementation we used only three basic transformations - rotation, translation and scaling. In all cases the transformations will work with homogeneous coordinates.

9 Projections To displaying a three-dimensional scene (3D) on a computer monitor that has the physical ability to display only twodimensional objects (2D), it is necessary to convert this 3D scene to 2D. This is the purpose of the projections. In practice there are a variety of projections, but in our module for 3D modeling we use only two types.

10 Modeling by Using Java3D API Java3D is an addition to Java for displaying 3D graphics. Classes provide extensive interface, sufficient to create highquality 3D models, animations and games. Java3D builds on existing technologies such as DirectX, OpenGL and can incorporate objects created by 3D modeling tools like TrueSpace and VRML models. Java3D provides the classes, from which we can create the overall scene. It is made up of different class objects that represent geometry of shape, appearance, shapes, lights, etc. The scene can be defined as a data structure consisting of nodes (which are instances of Java3D classes) and the relations between them (parent - child or a reference).

11 Creating an object When creating an object, we need to know some information - where we want to create an object, how it will look and what will be composition. We must mention that Java3D utilizes clockwise coordinate system. Whenever we want to create a 3D model, which is composed of a variable number of shapes, we have to decide how to create these shapes. This is the purpose of the GeometryArray classes. They provide PointArray for creating points, vectors using Lin-eArray, triangles using TriangleArray or quadrangles using QuadArray. For each of these classes we define the 3D coordinates of each point. Thus formed geometry is ap-plied (addgeometry()) to shape of Shape3D class together with a defined appearance (setappearance()). If we want to expose final shape to the transformations, we placed this shape as a child of TG (addchild()). Transform3D class provides transformation, which can be applied (settransform()) for the whole group. TG can be cover by using another BG. Thus, we created the visual part of the scene.

12 Creating a camera If we want to create a camera or cameras observing 3D model, we need to define camera properties. The base is ViewPlatform that controls the position, orientation and scale of the camera. If we place it into the TG, we will be able to manipulate with it and set it according to our ideas. Another important element is the view (View) that contains all the important parameters for rendering 3D scenes from a single point of observation. Parameters are specifics of camera (PhysicalBody), physical environment of scene (PhysicalEnvironment) and mainly canvas for 3D rendering (Canvas3D).

13 CALCULATIONS FOR THE ANALYSIS OF 3D OBJECT The fact that the model is created in 3D space, allows us to perform various calcula-tions, which are very important for the overall analysis - volume calculations, area, cen-ter of gravity and the surface area based on defined objects of interest from Chapter 2. From these values, we can generate graphs, so we do not need any additional external tools. For objects type intersection and point are volume and area calculations com-pletely irrelevant.

14 RESULTS, DISCUSSIONS, CONCLUSIONS The designed module allows us to view 3D model based on polygonal representa-tion of objects of interest from the series of measuring images. This provides a new per-spective, how to express the measured data. We have the ability to work freely with model through the interactions such as translation, rotation, scaling, dyeing a concrete objects, etc. We use two projections for the display of 3D model. The big advantage is the function for creating graphs based on the selected values without using external tools. We are also able to animate the progressive rendering of the entire model, com-pare measurements (series) among themselves, applying the texture to the model, print, and many other utilities.

15 RESULTS, DISCUSSIONS, CONCLUSIONS

16 RESULTS, DISCUSSIONS, CONCLUSIONS In this paper we describe, among other things, methods for defining objects of interest on a measurement images, including a new approach that is Radon transform and mathematical theory for creating a model in 3D space. Furthermore, we described method using Radon transform for 3D modeling of structure of extra-high voltage py-lons. The main intention of this paper is to introduce a new module of FOTOMNG system for visualization of the observed object in 3D. In conclusion we can say that in some cases there may be possibilities of 3D visualization based on polygonal representation insufficient. Therefore, the next step of development we see in the use of voxel graphics for creating 3D models. ACKNOWLEDGEMENTS The following grant is acknowledged for the financial support provided for this re-search: partially supported by Grant of SGS No. SP2015/142.