"Scientist's Component Toolkit" (SCT) is yet another "box-and-wire" component software construction set, but this time specifically for the Macintosh. It is being developed primarily for my own software development needs, rather than in anticipation of becoming a commercial product for sale.

Here are some images of current and past work on this project. Click on the images for a bigger picture:

Realtime acquisition and spectral analysis of EEG:

Electric dipole field:

NASA PDS Voyager spacecraft image decompression and processing:

Complex spectral analysis in cylindrical coordinates:

Most of the work being done on and with the SCT software is currently shown at my Macintosh EEG/EKG project website. Although SCT can be used for much more than EEG and EKG acquisition and analysis, this is the primary use that I have for it.

In addition, I have also created two other small examples of using SCT:

• A small digital oscilloscope application for testing electronic equipment.
  
  
• Reading, decompresing, and processing NASA PDS spacecraft images.

Overall Project

SCT is based on, but a complete rewrite of (and significant improvement on), the Scientist's Workbench project I did for Apple's Advanced Technology Group back in 1992-1993.

A short digression... Scientist's Workbench


---



Apple's Scientist's Workbench project for the Macintosh was a modular construction set for building distributed scientific applications. At first glance, Scientist's Workbench (SWB) resembles other component-based analysis systems such as SGI Explorer or AVS, which are available on Unix workstations. However, SWB extends this metaphor to include a rich component user interface, the use of concurrently running applications as components, arbitrary levels of component hierarchy, distributed and remote processing, and network interfaces to Unix, VMS, and Dos. Many data import/export types are supported, such as HDF and FITS. Typical SWB applications may include "hot links" to other scientific applications such as Mathematica and Spyglass. SWB was part of Apple's OpenDoc Scientific Solutions Builder project, and was expected to be released as part of the OpenDoc application suite sometime in 1995 or 1996.

As a test case for SWB, an autonomous agent capable of recognizing and counting arbitrary features (e.g. craters) in spacecraft photographs was constructed. Neural net technology using multi-layer feed-forward/back-propagation was used to construct a programmable scanner which could be trained to recognize arbitrary bitmaps in images. Training sets of artificial, random craters with adjustable parameters were input to the net until it was capable of recognizing features in simulated images with acceptable accuracy. A complete autonomous agent capable of reading multiple images, extracting features, and compiling coordinates and statistics in background mode was then constructed and used to locate actual features in Viking and Mariner images of Mars. This agent was capable of being trained to recognize arbitrary features in any image, and can be used as is or modified as desired.



I was a contractor for this project, and wrote the entire system (dynamic shell application and suite of scientific analysis components) first in Common Lisp (and CLOS), and then again in C++.

It should be pretty obvious what one could do with SCT. SCT consists of a shell application and an external set of palettes of dynamically linked components. The shell application manages the construction and editing of "data flow" documents containing various components linked together with "wires", while the components themselves do the actual work of processing data. Each component can have its own GUI consisting of windows, menus, buttons, and other controls. Writing new components is pretty easy, and can be carried out independently of the main app or other components. By the end of 2001, I was doing most of my new application development in SCT, before moving to OS X and ObjC/Cocoa.

Here is an image of the current set of components, as of 5/27/01:

Current components consist of:

• Infix equation parser and functions for real data.
• Numerical input.
• Numerical display as text.
• Clock counter.
• 1d plot.
• 1d vector generation (row or column).
• Text file reading and conversion to data.
• Conversion of data to text and file writing.
  
  
• 2d matrix generation.
• 1d non-blocking slider.
• Bar meter.
• 2d scatter/line plot.
• 3d scatter/line plot.
• 2d image.
• 2d to 1d data binning.
• Selectable output switch.
  
  
• Realtime data acquisition from Mac sound input port.
• Discrete Fourier transform and inverse.
• Convolution of real and complex data.
• Deconvolution of real and complex data.
• Menu-selectable functions of complex data.
• Magnitude, phase, and triggering of acquired data.
• Downsampling of data.
• Fist-in-first-out data accumulation queue and reformatter.
  
  
• 2d wireframe and filled surface.
• Data combiner and time-slicer.
• Data statistics and modifications.
• Maxima and minima finder.
• Discrete integration.
• Gradient and other "del" operations.
• 2d vector plot.
• Multiple 1d plots.

SCT was a System 9 program, and so has been asleep as a project for several years now. However, there is occasionally some interest in reviving the project and porting it to OS X. If that were to happen, it would probably be rewritten again in Python and some cross-platform graphics toolkit (e.g. Wx), so that it could run under several flavors of Linux as well.