Strabismus detection method based on a convolution neural network

Abstract

The embodiment of the invention discloses a strabismus detection method based on a convolution neural network, which comprises the following steps: strabismus images are collected and a strabismus image database composed of strabismus images is established; A convolution neural network is trained to locate the eye region by using a common face database, and the learning parameters of the convolution neural network are determined. A convolutional neural network for slant-eye recognition is trained by using the oriental face data set OFD and slant-eye image database, and the learning parametersof the neural network are determined. The eye region of the strabismus image in the strabismus image library is located by using the trained convolution neural network for locating the eye region. thetrained convolution neural network is used for cross-eye recognition to recognize the cropped eye region image, the result of cross-eye detection is output. The invention fully utilizes the feature extraction and feature learning ability of the convolution neural network, can efficiently and accurately judge whether a person suffers from squint or not, and is helpful for doctors to diagnose and treat squint diseases.

Description

technical field [0001] The invention relates to the fields of image processing and pattern recognition, in particular to a squint detection method based on a convolutional neural network. Background technique [0002] Strabismus is an eye disease that usually occurs in childhood. It is usually caused by problems with the eye nerves, brain or extraocular muscles. Squint has a serious impact on a person's life. Generally, if patients with strabismus do not receive reasonable treatment, they will worsen and develop into amblyopia, and once it degenerates, it will lead to blindness. At the same time, strabismus seriously affects the appearance, and may cause withdrawn, low self-esteem and abnormal psychology in patients with squinted eyes. [0003] It can be seen from the above that squint detection is very important for the prevention and treatment of squint. At present, the detection of the subject's squint is mainly carried out manually. Generally, trained clinicians often...

AI Technical Summary

Problems solved by technology

However, when each eye of the subject has a small difference, even experienced clinicians can easily ignore it, resulting in misjudgment and misdiagnosis, etc., and the efficiency of manual detection is not high

[0004] In existing strabismus detection methods, there are also reports on the detection of strabismus diseases based on traditional image processing methods, but the traditional image processing method has the following disadvantages: the requirements for squint image conditions are harsh, and light must be placed on the pupil Slant eye recognition can be performed only when reflection produces a focal point; using traditional image processing methods, the process is more complicated and the efficiency is low; the accuracy of squint eye recognition is low and the reliability is poor

Images