Real-time video streaming character positioning method based on heterogeneous image computing

Abstract

A real-time video streaming character positioning method based on heterogeneous image computing includes creating a heterogeneous computing platform, and arranging a camera; installing an operating system and configuring related necessary software and driving programs; creating a standard word stock, and storing all characters in a character set to a database; entering a computing step for abstract digital image processing, and creating a parallel computing API (application program interface); capturing an image by the camera, and transforming the image into a gray-scale map; blocking the map according to fixed size, and performing histogram equalization and black-white binaryzation on blocking contrast ratio; blurring the original image, and segmenting the original image into vertical strips according to fixed width to acquire starting and ending positions of each line of each vertical strip; calculating the slope according to the acquired positions, sequencing similarities of all words in a descending order, weighing the slope and historical results, and calibrating prior 5% of the words as a search result. The real-time video streaming character positioning method has the advantages that computing quantity is much smaller than that of the prior art; high resistance capability to adverse factors such as shaking, inclination and illumination nonuniformity is achieved; computing results are accurate.

Description

technical field [0001] The invention belongs to the technical field of machine vision and relates to embedded heterogeneous computing, digital image processing and real-time user interaction technologies. Background technique [0002] In people's daily life, we always encounter scenes like the following: when browsing through complicated materials, we want to quickly find relevant materials containing specific keywords, or when a person comes to a completely unfamiliar scene, To find the urgently needed information in such an unfamiliar environment, it is extremely difficult and ineffective to use traditional methods. [0003] For example, Optical Character Recognition (OCR) technology extracts text bitmaps in videos or static images and converts them into text streams that can be recognized by computers. information. However, this method has the following disadvantages: [0004] (1) The calculation amount of OCR is very large, and now and for a long time to come, with th...

AI Technical Summary

Problems solved by technology

[0004] (1) The calculation amount of OCR is very large, and now and for a long time to come, with the help of cutting-edge hardware equipment, it is impossible to achieve practical real-time performance;

[0005] (2) The accuracy of OCR technology can only be satisfied with helping to enter texts. For searching specific texts, the unstable accuracy rate and misrecognition rate make the effect not good;

[0006] (3) OCR is almost powerless for the distortion and tilt of the image, and cannot recognize most of the text content

[0008] (1) The amount of calculation required by image matching technology is much greater than that of OCR technology, so it is even more impossible to achieve real-time pe

Images