Table 1 Comparative characteristics of means of the parallel hierarchical transformation
From: Modeling of a method of parallel hierarchical transformation for fast recognition of dynamic images
Number | Parameter | Indices | |||
|---|---|---|---|---|---|
Known soft hardware (imitational modeling based on MLP and RBF networks[23]) | Known soft hardware (imitational modeling based on neural-like network technology[22]) | Known soft hardware (central processor, PH transformation method based on the Q-decomposition[19]) | Suggested soft hardware (graphical processor, method of the PH network training based on the normalizing equation) | ||
1 | Average size of good route fragments (%) | 70 | 74 | 50 | 18 |
2 | Average value of correct recognition (%) | 92 | 92,5 | 84,8 | 94 |
3 | Accuracy of the energy center determination, decomposition elements | 1.5 | 1.5 | 1.2 | 0.01 |
4 | Average recognition time of network processing(s) | - | - | 8.4 | 1.52 |
5 | Average time of preliminary processing of route fragments(s) | - | - | 3.32 | 0.6 |