A method for automatic detection of infant vision

Abstract

The invention relates to a method for automatic detection of infant vision. The method comprises the steps of S1, synchronously recording a video image for showing a Teller visual chart and a video image of a testee; S2, determining the position of a visual chart display window and a corresponding visual chart display time frame through the video image for showing the Teller visual chart, extracting a section corresponding to the visual chart display time frame from the video image of the testee, and conducting S3-S8; S3, training a strong classifier with the best human face recognition effect by means of the AdaBoost algorithm; S4, cascading the strong classifier to obtain a screening type cascade classifier; S5, decomposing the section extracted in S2 to obtain a video frame set, and then conducting S6-S7 on each video frame in the set; S6, detecting each video frame with the screening type cascade classifier, and determining a human face area of each video frame; S7, determining the position of an interest point in each human face area by means of the SURF algorithm, and then determining an SURF descriptor according to Haar wavelet response of a neighborhood of the interest point; and S8, calculating the similarity of the SURF descriptors of any two adjacent frames of images in the video frame set, concluding that matching is successful and eyes of an infant move if calculated similarity is larger than a set threshold value, and determining the vision level of the infant according to the grade of a grating corresponding to the visual chart.

Description

technical field [0001] The invention relates to the field of image processing, and more specifically, to a method for automatically detecting eyesight of infants. Background technique [0002] Compared with older children, infants under the age of 3 have no expressive language or physical abilities, lower levels of intelligence and recognition, and cannot maintain concentration for a longer period of time, making it difficult to cooperate with routine vision inspections. Therefore, in 1985, Mc Donald and others designed the Teller vision card. The Teller vision card is mainly composed of 17 grid cards and 1 blank card. It is mainly used for infants under the age of 3 who have no language expression ability and abnormal development of the nervous system. Children with cognitive or language impairments. The specific principle process of the Teller vision card to detect the eyesight of infants and young children is as follows: show different bar cards to the infants, if the ey...

AI Technical Summary

Problems solved by technology

[0003] In order to solve the defects of high misjudgment rate and unsatisfactory detection accuracy of the above infant vision detection method, the present invention provides a method of applying image processing technology to detect infant vision. The method has low misjudgment rate and high detection accuracy. Accuracy is improved compared to existing technologies

Images