Reconfigurable hardware acceleration method and system for Gaussian pyramid construction

Abstract

The invention discloses a reconfigurable hardware acceleration method and system for Gaussian pyramid construction, and belongs to the field of hardware accelerator design. The system provided by the invention comprises an SRAM (Static Random Access Memory) group, an FIFO (First In First Out) group, a switch network, a shift register array, an adder tree module, a demultiplexer, a reconfigurable PE (Provider Edge) array and a Gaussian difference module. According to the method, the reconfigurable PE array resources can be configured according to the requirements of different scenes and tasks on system performance to realize convolution calculation of different scales, so that high-performance Gaussian pyramid and Gaussian difference pyramid construction is realized. The method comprises a fast and slow double-clock domain design, a dynamic edge filling design, an input image part and a multiplexing design method, the hardware utilization rate is improved, reconfigurable calculation of different edge filling and different scale convolution is achieved, and the calculation amount and hardware overhead of Gaussian pyramid and Gaussian difference pyramid construction are reduced.

Description

technical field [0001] The invention belongs to the field of hardware accelerator design, and more specifically relates to a reconfigurable hardware acceleration method and system for Gaussian pyramid construction. Background technique [0002] Scale Invariant Feature Transform (SIFT) is a local feature description algorithm, which can solve the matching problem caused by translation, rotation, affine transformation and other conditions of two images, and has good invariance and strong matching ability. This algorithm converts a picture into a set with a large number of local feature vectors, and can still obtain more reliable matching results under conditions such as affine changes in the image, 3D viewing angle changes, scaling, exposure changes, and noise addition. Therefore, the algorithm is widely used in robotics, drones, autonomous driving, virtual reality, augmented reality, computer vision and other fields. [0003] With the improvement of image clarity and the rea...

AI Technical Summary

Problems solved by technology

[0005] Aiming at the defects of the prior art, the purpose of the present invention is to provide a reconfigurable hardware acceleration method and system for Gaussian pyramid construction, aiming to solve the problem of large hardware overhead, low hardware utilization, and computational complexity during the construction of a full-parallel Gaussian pyramid. A large amount and does not consider the problem of edge padding

Images