ME751 Project Proposal

Computer Aided Engineering Design ME 751/451 Jul-Nov 99 -:Mukerjee

Extraction of Toleranced CAD Models from Digital Images

Gagan Deep Arora (96095)
Ravi Prakash Srivastava (96226)
Indian Institute of Technology - Kanpur : September 1999

Motivation
Example
Past Work
Methodology
Sample Input-Output
Limitations
Bibliography
Back to Report

Motivation

Actual Design Process involves assignment of shapes and their dimensions which in turn serve as the guidelines for the manufactureres, developers, users, and the designers themselves. The various stages of development of a design, particularly the machinig fabrication and testing involve considering it as an approximation in the sense of defining tolerances for both shape and size parameters. These tolerances are inevitably encountrered due to the limitations of each of the stages of the design and development process. It is therefore only essential that the designer be equipped with tools which allow him to design and test with real-life considerations of tolerances.

The present-day CAD tools have little to offer in this area and much work needs to be done before tolerances can be handled by CAD platforms which can identify shape classes rather than rigidly considering shapes alone. The present work is a step in this direction wherein we attempt to extract toleranced CAD models from images of simple, yet representative, 3-D models and providing a shape class for the identified shape.

Example: Extraction of Tolerance model from snapshot of letter object 'L'

The Image taken is preprocessed for improving the brightness and contrast properties of the image so as to facilitate edge detection for the object which is essential for defining the shape boundary.
.

.
Figure: The Snapshot of object in full color and converted to greyscale.
.

Figure: Edge detected images - from greyscale image (left) and with pixel inversion with shadow edges removed (right). To be used as a start-point for the shape class definition algorithm.

As observed in the above edge detected image, the sharpness and the strength of the edges depend on the orientation of the edges with respect to the camera. Hence, also, when the same object is viewed from different angles, the shape definition may vary depending upon the light intensity and the relative positioning of the edges. Thus, several snapshots need to be taken of the object and each would then be identified with a set of vertices and edges. Some admissiable value of tolerance would decide the extent of deivation of these vertices/edges from the corresponding vertices/edges in the various images.

Past Work

Identifying Vertices and Edges

Extraction of vertices and edges in the input image is a primary concern of the project work. The entered image can, at best (in ascii format), be identified as an array of numbers ( in the range 0 to 255 for greyscale format wherein each pixel is represented by a single number ). Identification of straight lines and curves would be based on techniques similar to or directly based on Hough Transforms. Binarised edge-detected images would be traversed along the points of high intensity values, the path being guided by the vicinity crieteria for high-intensity points. Traversal for a particular edge would be ensured by the continuity of the high-intensity points. The exercise would essentially identify edges and hence the approximate vertices for the image. The accuracy of the edges detected would depend on the choice of the "freedom-to-deviate" factor. A high value of this factor would result in gross approximation and loss of information while too low a value for this factor would encourage high level of noise in the extracted model.

Defining Shape Classes from identified set of edges and vertices

The edges, as obtained from several such images of the same object would be used to construct shape class for the object by recording the effect of induced deflections in vertices' position, edges' position(and angle) or both. Thus, there would be primarily two methods of inducing disturbances in the image and accepting or rejecting the so produced images based on the range of tolerances allowed.

Disturbance in vertices - These would consist of allowing each vertex to deviate slightly from its original position and then noting the changes thus produced.
Edge disturbances - In this, the edges would be allowed certain amount of extension/compression and also angular deviation.

Sample Input and Expected Output

The input consists of edge detected binarised images like the one show in example. For each object, the input would be several such images actually taken from varying angles. The problem would be taken up for an increasingly complex set of input data. It is worth noting that the problem of extracting CAD models and assigning tolerances would be an increasingly difficult one as we consider object from the range of I to Q and similar such complicated ones ( with several edges and particularly "holes").

The above snapshots were taken under a single light source and thus contain severe shadow effects. This is undesirable for obvious reasons and fresh images need to be taken for each of these.
Also, the change in angular orientation for the two images of 'Q' is drastic and shall not be helpful in defining shape classes. The different images need to be taken such that the major shape of the object is preserved. One such method would be taking the snaps from ovrhead.
The output would be a set of shape classes which shall be all be different in the strict sense of CAD modelling as used today, but each of these would be identified as the same object from where they originated. Moreover, based on these shapes, as mentioned in the methodology, more shapes would be generated by inducing vertex and edge disturbances. All these again, would be identifiable as the same object.

Limitations

The current study shall be kept limited to shapes which can be approximated with straight lines and second order curves, particularly circular arcs, for the sake of simplicity. The results can be later extended to include higher order curves and other shapes as well.
The work attempted here would be limited to taking only edge detected images as inputs, in pgm ( greyscale ascii / raw )format. The image processing part is left to be done by standard image processing utilities available. We make use of GIMP for the preprocessing needs.
Presence of Shadows in the input image would crucially affect the results since the edge detection algorithms would identify these as shape boundaries. It is therefore required to take care to minimise shadow while taking the images ( should be taken in diffused sunlight ) and required preprocessing should be allowed to eliminate any present shadows.

References

Goswami,A. Identification of Partially visible Shape Classes, July 1999 Indian Institute of Technology, Kanpur, Dept of mechanical engg.
Two Dimensional Image Processing
ICRA'98 Article Abstracts May 16-20, 1998 in Leuven, Belgium
Srinivas Akella's Research

The references at the moment are not annotated in the standard format. They serve only as indicators to the reference we took. Later in the study, the bibliography shall be completed

Source Snaps as taken by the digital camera.

This proposal was prepared by Gagan Deep Arora and Ravi P.Srivastava as a part of the project component in the Course on Computer Aided Engineering Design in the July-December Semester, 1999.
(Instructor : Amitabha Mukerjee )

[CAED COURSE LECTURES HOME PAGE]

Contents