Self-adaptive control-based circuit product correlation algorithm realization system

Abstract

The invention discloses a self-adaptive control-based circuit product correlation algorithm realization system. The system comprises a self-adaptive control circuit, a data parallel splitting module, an algorithm function unit, a summing unit and an output control circuit, wherein the self-adaptive control circuit is connected with the data parallel splitting module, the adding unit and the output control circuit; and the algorithm function unit and the adding unit are connected with the output control circuit. Along with the change of size condition of an image, the control circuit is capable of self-adaptively splitting the image into a plurality of sub-parts which are suitable for calculation, the algorithm function unit is utilized to respectively calculating the multiply add value of each sub-part, and the multiply add values are added to obtain the value of the multiply accumulate items to be calculated; and through calculating the number of the algorithm function units required by the matching of one image, the corresponding parallel circuits are correspondingly decreased until only one circuit is left, so that the FPGA interior resources are fully utilized, the maximum algorithm parallelizing is realized and the operational performance of a co-processor is improved.

Description

technical field The invention belongs to the technical field of product correlation algorithms, and in particular relates to a product correlation algorithm realization system based on an adaptive control circuit. The adaptive control circuit is used to optimize and control the product correlation algorithm circuit to realize the self-adaptation of the product correlation algorithm to different image sizes. , to achieve the optimal algorithm efficiency and hardware resource usage. Background technique In many fields such as image processing, pattern recognition and information search, it is necessary to perform matching operations on images. Commonly used matching operations include absolute difference, mean square error, and product correlation, among which the product correlation algorithm has the advantages of high precision, strong adaptability, and good robustness, and has been widely used in many fields. In engineering practice, normalized product correlation is often...

AI Technical Summary

Problems solved by technology

The shortcomings at this time are obvious. On the one hand, when the matching operation with a size smaller than 64 is actually performed, a large amount of storage resources and multiply-accumulate resources are wasted. Attachment 3 shows that when the small image size is 32×32, the FPGA product Underutilized storage and computing resources inside the algorithm coprocessor, the actual resource utilization rate is less than 50%; on the other hand, when the size of the small image exceeds 64, the original design can no longer meet the computing requirements and needs to be redesigned

Images