Handwritten numeral recognition implementation method

Abstract

The invention discloses a handwritten numeral recognition implementation method, and belongs to the field of image recognition. According to the method, the convolutional neural network is mainly deployed on a ZYNQ embedded hardware platform, and handwritten numeral recognition is realized through collaborative acceleration of software and hardware. The method comprises the following steps of: firstly, performing graying and binarization processing on an input image, performing identification frame matching on the input image and a data set picture in size, and then storing an identification frame image into a BRAM (Block Random Access Memory) storage unit; then, convolution operation, function activation and pooling operation acceleration are carried out on identification frame image data at a PL end; constructing a camera time sequence by using the pooled image data, and transmitting the camera time sequence to a DDR (Double Data Rate) of a PS end; and finally, hidden layer and output layer operation is completed at the PS end, and an identification result is transmitted to the PL end to be displayed. According to the method, reasoning operation of a part of neural networks can be accelerated, and handwritten digits in the picture can be quickly recognized.

Description

technical field [0001] The invention belongs to the field of image recognition, in particular to a method for realizing handwritten digit recognition. Background technique [0002] At present, the recognition of human handwritten characters has become a research hotspot. The technology is already widely used in tax form processing, mail sorting and computerized filing. In these applications, handwritten digit recognition algorithms are usually required to have high recognition speed and recognition accuracy, as well as high reliability and stability. Though the category of handwritten numerals only uses ten kinds, and stroke is simple, its recognition problem still exists very big difficulty. Some existing test results have shown that the correct recognition rate of numbers is not as high as the recognition rate of printed Chinese characters, and even not as high as the recognition rate of online handwritten Chinese characters. The main reason is that the fonts are not mu...

AI Technical Summary

Problems solved by technology

[0006] However, since CNN calculations involve many different types of operations, such as two-dimensional convolution operations, nonlinear activation function operations, pooling (subsampling) operations, and fully connected layer operations, and these operations often involve a large amount of data access As well as the storage of intermediate result data, it is still a challenging task to use FPGA to implement such a complex CNN with a huge amount of calculation. At the same time, a large number of floating-point calculations will cause precision problems. The realization of the convolutional neural network system applied to handwritten digit recognition has important theoretical research significance and practical value for the development of the field of image recognition

Images