CS 5320/6320 Project

Interpretation of Engineering Drawings

 

 

Team Members:  Thomas C. Henderson

 

Week 1: January 11 2007

 

 

Goal

 

The semantic interpretation of industrial drawings is a very difficult

problem.  Tombre (Tombre 1998a) summarizes the state of the field for CAD drawings and as a document image analysis problem (Tombre 1998b, Tombre 1998c}.  An earlier overview by Haralick and Kanungo (Kanungo 1995} gives insight into the scale of the technical drawing problem (e.g., about 250 million drawings are generated annually), as well as a discussion of performance measurement.  A major problem pointed out by these reviewers is that there is a dearth of work relating high-level models of documents, drawings or graphics to the various levels of  analysis.

The goal of technical drawing analysis is to interpret the contents of

an image, such as that shown here (part of an engineering drawing):

 

 

 

 

 

Several systems have been developed along these lines, although none works well in an automated fashion.  Various problems arise; e.g., when text touches graphics or when there is noise in the scanned image or when the thresholds of the image-analysis codes support only a few classes of drawings.  One notable system is CELESSTIN (Ahsoon 1995, Tombre 1993a}; however, as pointed out by Haralick, CELESSTIN suffers from the fact that it is based on a hierarchical rule system and has many rules; furthermore, whatever interpretive models the rules instantiate exist only in the thresholds and logic of the hand-coded rules.

 

In the standard form of analysis, the scanned drawing is digitized, noise is removed, text and graphics are recognized, graphics is vectorized, dimensions are extracted and the whole drawing is analyzed by applying knowledge rules. Each of these steps normally uses just a single set of thresholds.  Our goal is to allow a wide range of thresholds to generate a potentially large search space (hopefully including the correct components), and then to apply domain constraints to prune the space.

 

Issues

 

The interpretation of engineering drawings requires:

 

  1. An image formation model.  This involves not only the scanning process, but also the physical defects that can be found on a drawing; for example, creases, stains, tears, etc.

 

  1. An extraction of basic image elements; this includes:
    1. Straight edge segments
    2. Text
    3. Arrows
    4. Boxes
    5. Circles
    6. Etc.

 

  1. Identification of structural drawing elements; this includes:
    1. Dimension sets
    2. Graphics
    3. Title block
    4. Revisions list
    5. Materials list
    6. Etc.

 

Thus, we will develop models based on these issues.  Models include:

 

  • Physical Image Degradation Processes
  • Image Formation
  • Line segments
  • Text
  • Boxes
  • Arrows
  • Circles
  • Structural model.

 

Schedule

 

Week  1:  Define Project

Week  2:  Develop physical degradation model

          Develop image formation model

Week  3:  Explore use of color in models

Week  4:  Develop line segment model

          Develop image smoothing approach

          Explore use of linear (and nonlinear) filters

Week  5:  Explore use of gradient, 1st derivative, etc.

          Explore line segment detector alternatives

Week  6:  Define performance measures

          Develop benchmark dataset and ground truth

Week  7:  Implement first cut image analysis system

Week  8:  Measure performance of image analysis

Week  9:  Explore various structural methods

Week 10:  Compare various structural methods

Week 11:  Compare structural methods

Week 12:  Explore complete system sensitivity analysis

Week 13:  Thanksgiving

Week 14:  Measure system performance

Week 15:  Optimize system and draw conclusions

Week 16:  Final results and report

 

Bibliography

 

Tombre 1998a     "Analysis of Engineering Drawings: State of Art and Challenges," Karl Tombre, Graphics Recognition - Algorithms and Systems, Lecture Notes in Computer Science, Vol. 1389, Springer-Verlag, Berlin, April, 1998, pp 257-264.

 

Tombre 1993a     "Don't Tell Mom I'm Doing Document Analysis; She Believes I'm in the Computer Vision Field," Suzanne Collin, Karl Tombre and Pascal Vaxiviere, Proceedings of 2nd International Conference on Document Analysis and Recognition, Tsukuba Science City (Japan), October, 1993, pp. 619-622.

 

Ahsoon 1995      "A Step Towards Reconstruction of 3-D CAD Models from Engineering Drawings," Christian Ah-Soon and Karl Tombre, Proceedings 3rd International Conference on Document Analysis and Recognition", Montréal (Canada), 1995, pp. 331-334.

 

Tombre 1998b     "Ten Years of Research in the Analysis of Graphics  Documents: Achievements and Open Problems," Karl Tombre, Proceedings of the 10th Portuguese Conference on Pattern Recognition, Lisbon (Portugal), 1998. 

}

 

Tombre 1998c     Graphics Documents: Achievements and Open Problems," Karl Tombre, Proceedings of the 10th Portuguese Conference on Pattern Recognition, Lisbon, Portugal, 1998.

 

Kanungo 1995      "Understanding Engineering Drawings: A Survey," Tapas Kanungo, Robert M. Haralick and Dov Dori, Proceedings of First IARP Workshop on Graphics Recognition, University Park, PA, pp. 217-228, 1995.

 

 

Week 2: January 15 – 2 September 2004

 

 

Physical Degradation Model:

 

Physical paper engineering drawings can undergo several types of degradation:

 

  • Tearing: this includes rips, holes, or areas removed from the paper
  • Staining: this includes spills, splotches or other induced blemishes
  • Folding: this includes folds that induce perceivable lines in the drawing

 

Tearing:

 

A rip will be characterized as a straight line segment starting within some distance of an edge and running at some angle straight to the edge of the image with the following parameters:

:

  • MAX_RIP_DIST: maximum distance from an edge for a rip to start
  • uniform distribution of starting locations from edge to MAX_RIP _DIST
  • x,y: location of start point of rip chosen uniformly in range between 0 and  MAX_RIP_DIST of  the image edge
  • theta: angle of orientation of linear rip line segment (horizontal is 0 degrees).
  • thickness: thickness of rip; selected from normal distribution N(mu,sigma)

 

A hole will be characterized as a circle with the following parameters:

  • x,y location is chosen from a uniform distribution over the complete image,
  • r: radius is selected from a normal distribution over N(mu,sigma).

 

A removed area will be characterized by the following parameters:

 

·        shape: selected from a predefined set of shapes defined as pixel sets

·        scale: selected from a normal distribution over N(mu,sigma)

 

Staining:

 

Staining will be parameterized by:

  • shape: a predefined set of shapes defined as pixel sets
  • scale: selected from a normal distribution over N(mu,sigma)

 

[Note: this model is identical to the removed area model.]

 

Folding:

 

Folding will be modeled as a line through the image and will be characterized by the following parameters:

  • rho: normal distance from origin to line sampled uniformly from 0 to maximum distance which remains in image
  • theta: orientation of line sampled uniformly from 0 to pi.

 

 

Image Formation Model

 

The engineering drawing image is produced by a linear camera array which moves along the x-axis (i.e., along the columns) to scan the image.  The following parameters are used to model the process:

  • pixel noise: the true gray level of the paper drawing is offset by noise sampled from a normal distribution N(mu,sigma)
  • motion blur: each pixel produced by the imaging array represents a patch of the physical drawing
  • pixel diffusion: pixels in the linear array diffuse charge (which results in intensity diffusion) and this is modeled as a simple diffusion process; that is, a diffusion rate occurs along the gradient.