a) Operational Task One: Pre-processing and storage
• Pre-processing of data with the destination to storage by:
- Segmentation of data into regular parts
- Measurement of digital objects (attribute calculation)
- Indexation of objects by their attributes
- Classification of objects by their attributes (induction)
- Data approximation (with restitution)
b) Operational Task Two: Interrogation and Retrieval
• Interrogation of recorded data
- Interrogation by address
- Restitution of data according top location
- Interrogation by content
• Multi- criteria selection and interrogation on the closest neighbour with restitution of a list of objects sorted using a similarity index
• Interrogation based on relational algebra as function of the position of objects relative to others.
• Interrogation based on knowledge realizing a semantic network: Instantiating an interrogation space (deduction), and learning function.
2-4.3
Implementation of a man – machine interface
The “glue” between these operational tasks is well served by the COBRA model. The distributed nature of the PACHA system, using high performance SCI interfaces, and the distributed support provided by the system, using high performance SCI interfaces, and the distributed support provided by the system running over the hardware architecture provide a sound basis for the integration of COBRA object request broker and COBRA interfaces to clients performing acquisition or interrogation and retrieval tasks. This model can be extended to include more specific processing methods (for compute intensive data processing e.g. seismic data, groundwater flows simulations, etc…) where according to the interrogation criteria new analysis might be required to complete image feature or object extraction providing results for the interrogation and enriching the data base at the same time.
2-4.4 Computer System Strategy
The PACHA system has been evaluated as being ideal for such a high performance intelligent data bank, already possessing many necessary features which are required to ensure efficient execution of principal actions of such a system. However, there are a number of gaps in the system which require being filled out before integration and testing of production products could start. To shorten the development time the following principal tasks are proposed as an approach to the outstanding design and implementation :adapt ion action required to complete the product:
• Acquisition of data sets and evaluation tests
• Installation of a preliminary test-bed platform
- Adaptation of software to run on a high performance distributed machine
• Evaluation of the platform
2-4.5 Implementation steps
• Acquisition of data test
• Data pre-processing:
- Port and adapt ion to image signal processing
* Data segmentation
* Measurement of digital objects
* Indexation of digital objects
- Development and test of function
for image signal processing
* Classification of digital
objects
* Data compression
* Data approximation
• Interrogation of data:
- Port and adaptation to image signals processing
* Interrogation by address
* Multi-criteria selection and nearest neighbour query
- Development and test of functions for interrogation
* Interrogation based on relational algebra
* Interrogation based on knowledge including learning• Architecture Systems:
- Acquisition of elements based on a high performance distributed architecture
* Installation of a test –bed platform
* Port and integration of software onto platforms
* Acquisition of systems evaluation data sets
* Tests and evaluation of the platform