Vision detection algorithm based on portable infrared camera

Abstract

The invention relates to a vision detection algorithm based on a portable infrared camera, and belongs to a deep learning image processing technology. According to the optometry method, firstly, multiple facial images under lamplight at different angles are obtained by a photorefraction ocular screening device, then pupil detection is rapidly and accurately realized by adopting deep learning Faster R-CNN, pupil images obtained by detection are further researched with a threshold segmentation method, characteristic parameters of a crescent-shaped area can be obtained, and diopter is calculated according to the characteristic parameters. According to the algorithm, a fast and accurate recognition technology of deep learning and an advanced photorefraction ocular screening device technology are combined, and a result of a large number of experiments indicates that the method is high in speed, good in stability and good in real-time performance, can effectively improve accuracy measurement of shortsightedness and realizes intelligent detection of eyesight.

Description

technical field [0001] The invention relates to a vision detection algorithm of a portable infrared camera, which belongs to the deep learning image processing technology. Background technique [0002] In the field of vision screening, the previous methods lack simple and effective detection methods, and the general screening rate and treatment rate are generally low. In 2000, the eccentric photographic optometry improved by Li Xiangning, Chen Jiabi, etc., although the influence of the width of the light source and the size of the camera aperture can reduce the systematic error in the calculation, but the optometry method has a working blind spot, the method is old and the accuracy is poor. The measurement effect is not ideal. In 2002, Zhang Tao, Zou Liner and others proposed a digital photorefractor. Due to the single optical path, optometry has high requirements for the instrument and the method is outdated, and professional optometrists are required to use it. In 2009, ...

AI Technical Summary

Problems solved by technology

[0002] In the field of vision screening, the previous methods lack simple and effective detection methods, and the general screening rate and treatment rate are generally low

In 2000, the eccentric photographic optometry improved by Li Xiangning, Chen Jiabi, etc., although the influence of the width of the light source and the size of the camera aperture can reduce the systematic error in the calculation, but the optometry method has a working blind spot, the method is old and the accuracy is poor. Measuring results are not ideal

In 2002, Zhang Tao, Zou Liner and others proposed a digital photorefractor. Due to the single optical path, optometry has high requirements for the instrument and the method is outdated, and professional optometrists are required to use it.

Moreover, due to the aberration of the eye, it will also cause a certain measurement error, which will affect the measurement accuracy of the diopter.

In 2013, Luo Yuhao et al. used OpenCV to realize the limited effective area of ​​pupil detection in infrared images, which is only applicable to the situation where the pupil position is relatively fixed, and it is difficult to deal with the situation where the relative displacement of the pupil is large

This method can use the image information of the container standard to obtain better segmentation results, but at the same time, it needs to pre-segment a large number of images, which is impossible to achieve in most cases

Liang Xiaoling et al.’s research on the myopia and strabismus screening method based on image processing used the Hough transform method to realize the detection of the pupil diameter and the pupil center point to realize the detection of the pupil image, but the data collected by this method is relatively simple, and the The stickiness is low, and there is no new breakthrough in the data acquisition of traditional instruments

[0006] The new optometry instrument, such as Japan's Nidec AR-20 handheld computer optometry instrument, is more portable, and its price is about 150,000 yuan. Although it is cheaper than the large optometry instrument, it is still expensive, so it can only be used in hospitals or hospitals. other medical sectors, and not acceptable to the general public

Images