Testing the database of optical spectra

Spectra. Indexing. Recognition. Search. (PDF-presentation)

Testable objects:
Random wandering swinging RGB-spectra consisting of 128 values of RGB type (3 bytes).

What are “Random wandering swinging RGB-spectra”?
– See video Swinging optical spectra

1. Number of spectra 250 million
(32 billion
2. Size of tables of source spectra on hard disk

Source spectra
89.4 GB
3. Size of the table of masks of spectra

61.5 GB
4. Size of the index of masks of spectra

93.3 MB
5. The size of index related to the size of the indexed data (related to the size of the table of masks of spectra) 0.148 %
6. Time needed to build the table of masks of spectra, including the time of reading the functions from the hard disk 01:06:20.203
7. Time of indexing masks of spectra 04:35:46.547
8. The size of RAM used by the shell program 3.8 MB
9. The size of RAM buffers allocated to organize fuzzy search 5.0 MB
10. Average time to receive one search query result in CROSS mode with a 30 % coefficient of similarity of spectra to a pattern ~ 28 millisecond


How is the process of controlled search realized?
– See video
Controlled exact/fuzzy search for spectra.

Computer used in tests is a standard home desktop computer bought for $ 1000: Intel(R) Pentium(R) Dual CPU E2200 @ 2.20 GHz, 2.99 GB RAM.

Prior to generating the data, indexing and testing, there has been no preparation of the computer done nor has there been any disk defragmentation software used, i.e. the station was used in the current status "as is".

During test queries a disk space of the station was used up by 95 % (full size of disk space is 390 GB).

Platform: Microsoft Windows XP Professional, version 2002, Service Pack 2.

During the tests the computer was connected to the Internet, anti-virus system was active.

When saving the source base on a portable hard disk Seagate FreeAgent, the system works via USB-channel at approximately the same speed. Search time is the same as in case when the database is located on the hard disk.

Search Technology developed with support from FASIE
foundation formed by the Government of Russian Federation
Novosib-BIT LLC 2004 - 2017