Intelligent glasses for refraction and visual function rehabilitation based on AI real-time image correction

Abstract

The application discloses a kind of refraction and visual function rehabilitation intelligent glasses based on AI real-time image correction.The method comprises the following steps: collecting physiological parameters of both eyes respectively and independently;Pixel compensation parameters are generated by local chip or cloud server conversion;Pixel-level displacement, scaling, distortion correction is performed on the real scene picture, and the corrected picture is output;The compensation parameter is reduced in a fixed time period step by step;The above steps are executed cyclically.The application also protects the hardware device of the intelligent glasses that implements the method, and the acquisition module installation position, operation carrier and parameter adjustment period all have diversified implementation paths.The application is different from the existing scheme based on adjustable physical lenses, which completely replaces the physical lens light path adjustment through digital image correction, realizes instant correction and gradual rehabilitation.The application belongs to the image processing display control method, and does not involve the diagnosis and treatment of diseases.

Description

Technical Field

[0001] This invention relates to the fields of smart wearable devices, augmented reality image processing, and ophthalmic optometry correction technology, specifically to a smart glasses based on AI real-time image correction that can realize immediate correction and progressive rehabilitation treatment of refractive errors and visual function defects. Background Technology

[0002] Currently, the correction and rehabilitation programs for refractive errors and strabismus are mainly divided into three categories: traditional optical correction equipment, ordinary smart glasses, and special vision rehabilitation training equipment.

[0003] Traditional physical lens correction relies on the refraction of light by the lens to change the optical path. With fixed correction parameters, frequent lens replacements are necessary after vision changes. Prisms suffer from chromatic aberration and weight issues, offering only temporary correction without any rehabilitative effect. While adjustable physical lenses (such as fluid-filled lenses and adjustable-focus lenses) have emerged in recent years, achieving adjustable parameters, they still rely on hardware optics and suffer from drawbacks such as complex structure, high cost, and inability to achieve digital, progressive rehabilitation.

[0004] Existing AI / AR smart glasses primarily function as audio-visual interaction devices, lacking the ability to perform real-time digital image correction of real-world scenes and failing to possess adaptive rehabilitation adjustment logic. A few patents propose methods for AI pre-corrected images, but these are offline preprocessing systems that cannot respond in real-time to dynamic changes in eye parameters and do not include a rehabilitation mechanism for periodically reducing the amount of correction.

[0005] Traditional rehabilitation training devices are bulky and cannot be worn daily. Their correction parameters are fixed and they lack a closed-loop rehabilitation mechanism that allows for daily automatic fine-tuning.

[0006] In summary, existing technologies lack a wearable eyeglass that integrates real-time autonomous refraction, digital optical path correction, real-time image correction, adaptive fine-tuning, and long-term rehabilitation, while also possessing strong hardware expandability. Summary of the Invention

[0007] The technical problem to be solved by this invention is to provide a hardware device with a multi-path progressive correction method that uses digital image processing to replace physical lens optical correction, supports local / cloud dual algorithm carriers and multi-cycle parameter adjustment mode, prevents competitors from circumventing patent protection by changing the computing method or hardware structure, and realizes daily correction and progressive rehabilitation of various vision diseases.

[0008] Technical Solution: On the one hand, it protects the multi-architecture hardware product structure (as described in claim 5), and on the other hand, it protects the multi-path five-step closed-loop correction method (as described in claim 1). The core steps are: independent acquisition of optometry data for the left and right eyes → local / cloud conversion of correction parameters → digital correction of real-scene images → periodic stepwise reduction of compensation amount → cyclical wearing. It should be noted that the method provided by this invention is a display control method based on image processing, which improves the viewing effect by digitally transforming real-scene images. This method does not contain any steps for diagnosis or treatment purposes, nor does it apply any physical or chemical intervention to a living body, therefore it does not fall under the "methods for diagnosis and treatment of diseases" as stipulated in Article 25, Paragraph 1, Item (3) of the Patent Law.

[0009] Beneficial effects: 1. By locking in the entire protection logic at the methodological level, regardless of whether the competitor switches to cloud algorithms, monthly parameter tuning, or separate hardware, all fall within the protection scope. 2. The essential difference from existing technologies (such as adjustable physical lens solutions) is that this invention is a software-level digital image correction, rather than a hardware-level optical lens adjustment; the technical paths are completely different. 3. Compared with offline pre-correction solutions, this invention is a real-time closed-loop system that can dynamically adjust the correction amount based on real-time collected eye parameters, and includes the unique feature of periodic parameter reduction. 4. Achieves comprehensive protection across product and method dimensions, preventing circumvention of patent infringement by merely modifying local technologies. Attached Figure Description

[0010] Figure 1 : Schematic diagram of the overall hardware structure of the smart glasses of this invention; Figure 2 : System functional closed-loop logic block diagram of the present invention; Figure 3 A schematic diagram comparing the digital optical path correction principle of this invention with existing technologies; Figure 4 : Schematic diagram of the adaptive progressive fine-tuning rehabilitation process of this invention. Detailed Implementation

[0011] Example 1 (Local Daily Full-Screen Correction Solution) The glasses' local chip performs all AI calculations. A binocular physiological parameter acquisition module is fixed inside the frame, capturing eye position and refractive error for each frame. The AI ​​calculation module outputs pixel compensation parameters for each frame based on a deep neural network (e.g., for myopia, the overall real-world image is magnified by a preset ratio; for exotropia, the left and right eye images are shifted inward by preset pixel amounts). An adaptive correction adjustment unit lowers the compensation parameters by a preset amount each day. The display unit is an AR translucent display screen, outputting a uniformly corrected image across the entire screen. This method does not involve any medical diagnostic steps; it is purely image display control.

[0012] It should be noted that the visual function parameters collected by the binocular physiological parameter acquisition module include, but are not limited to, at least one of the accommodation function parameters, convergence function parameters, and three-level visual function parameters.

[0013] The accommodative function parameters include accommodative amplitude, accommodative sensitivity, positive relative accommodation, negative relative accommodation, and accommodative response. Accommodative amplitude characterizes the eye's maximum accommodative capacity; accommodative sensitivity characterizes the eye's ability to quickly switch focus between different distances; positive relative accommodation and negative relative accommodation characterize the additional accommodative force the eye can utilize when focusing on near objects and the accommodative force it can relax when focusing on near objects, respectively; and accommodative response characterizes the difference between the actual accommodative amount and the ideal target accommodative amount. These accommodative function parameters can comprehensively assess the user's accommodative function status, providing a data foundation for myopia prevention and eye strain rehabilitation.

[0014] Convergence function parameters include near and far heterophoria, AC / A ratio, near convergence point, positive relative convergence, and negative relative convergence. Near and far heterophoria represent the eye displacement at far and near distances, respectively; the AC / A ratio represents the linkage between convergence and accommodation; the near convergence point represents the closest distance at which both eyes can maintain single vision; and positive and negative relative convergence represent the maximum range at which the eye can resist inward or outward rotation, respectively. These convergence function parameters can be used to diagnose binocular vision dysfunctions such as convergence insufficiency and convergence excess.

[0015] The three levels of visual function parameters include simultaneous vision, horizontal fusion range, vertical fusion range, and stereoscopic acuity. Simultaneous vision characterizes whether both eyes can see the same target at the same time. Horizontal and vertical fusion ranges characterize the ability of both eyes to fuse two slightly different images into a single image in different directions. Stereoscopic acuity characterizes the ability to perceive depth and three-dimensional structure. These three levels of visual function parameters can be used to evaluate the effectiveness of amblyopia rehabilitation and postoperative training for strabismus.

[0016] By collecting at least one of the aforementioned visual function parameters, the optometry AI computing module can more accurately identify the type of eye defect in the user and generate personalized pixel compensation parameters, thereby achieving more comprehensive refractive correction and visual function rehabilitation.

[0017] Example 2 (Cloud-based Half-Moon Partition Correction Scheme) The glasses are only responsible for collecting data and uploading it to a cloud server via a wireless network. The cloud server runs an AI model and returns correction parameters for each frame. The system uniformly lowers the image compensation level by one level every fixed number of days (e.g., 15 days). During display, a monocular localized correction mode is used, meaning that only the image corresponding to the eye with anisometropia is corrected. This method also does not involve any treatment steps.

[0018] Example 3 (Split-type mobile phone linkage solution) The glasses and a mobile app connect wirelessly. The app embeds an AI model that receives real-world images and physiological parameters from the glasses, accelerates calculations using the phone's computing power, and returns corrective parameters. The parameter adjustment cycle is a preset number of days (e.g., 7 days), and a monocular zone correction mode can be used. The app can further offload some computations to the cloud.

[0019] Example 4 (Monthly Gradual Rehabilitation Program) All calculations are performed locally, and the parameter adjustment cycle is relatively long (e.g., 30 days). The data storage unit records monthly changes in ocular signs, and the adaptive adjustment unit can adjust the adjustment range according to the trend of change (e.g., appropriately increase the adjustment range when signs improve significantly, and decrease the adjustment range when improvement is slow).

Claims

1. A smart glasses for refractive and visual function rehabilitation based on AI real-time image correction, applied to wearable glasses devices, characterized in that, Includes the following steps: S1: By using the acquisition components mounted on the frame of the wearable glasses, independently acquire eye position data, refractive power data, astigmatism data, anisometropia values, and visual function parameters including at least one of the following: accommodation function parameters, convergence function parameters, and third-level visual function parameters for the left and right eyes. S2: Based on the local chip of the device or a remote cloud server, the corresponding real-scene pixel compensation parameters are generated according to the data collected in step S1. The pixel compensation parameters include displacement, scaling ratio and distortion correction coefficient. S3: Acquire real-world images of the outside world, and perform pixel-level displacement, scaling, and distortion correction on the real-world images of the outside world based on the compensation parameters generated in step S2, generate a corrected image and output it to the display unit for the user to view. S4: Calculate the cumulative wearing time of the device, and reduce the compensation parameters generated in step S2 in a stepwise manner at a preset fixed time period; S5: Repeat steps S1 to S4.

2. The correction method according to claim 1, characterized in that: The adjustment function parameters include at least one of adjustment amplitude, adjustment sensitivity, positive relative adjustment, negative relative adjustment, and adjustment response.

3. The correction method according to claim 1, characterized in that: The set function parameters include at least one of the following: near and far hemianopia degree, AC / A ratio, set near point, positive relative set, and negative relative set.

4. The correction method according to claim 1, characterized in that: The three-level visual function parameters include at least one of simultaneous vision, horizontal fusion range, vertical fusion range, and stereoscopic sharpness.

5. The correction method according to claim 1, characterized in that: The conditions that this method is suitable for treating include myopia, hyperopia, astigmatism, esotropia, exotropia, vertical strabismus, anisometropia, and binocular vision dysfunction.

6. The correction method according to claim 1, characterized in that: The fixed time period can be either adjusted daily or by gradient adjustment over multiple days.

7. The correction method according to claim 1, characterized in that: The image correction in step S3 includes a full-screen unified correction mode and a single-eye local partition correction mode.

8. A smart glasses hardware device for implementing the correction method according to any one of claims 1 to 7, characterized in that, include: Picture frame carrier; A binocular physiological parameter acquisition module is fixed to the inside of the frame carrier, the nose pad, or the front end of the temple, and is used to independently acquire the physiological parameters of both eyes. A real-scene image acquisition unit is fixed to the front of the frame carrier; A light-transmitting imaging display unit is fixed on the frame carrier; Data storage unit; as well as The image processing main control unit is electrically connected to the binocular physiological parameter acquisition module, the real-scene image acquisition unit, the light transmission imaging display unit, and the data storage unit, respectively. The image processing main control unit is selected from any one of the computing carriers, including a local integrated chip, a cloud server, or an external mobile terminal, and is electrically connected to the acquisition module and the display unit via wired or wireless means. The image processing main control unit integrates or connects to an optometry AI computing module and an adaptive correction dynamic adjustment unit. The adaptive correction dynamic adjustment unit includes a timing unit for recording the cumulative wearing time and a storage unit for storing the decreasing parameters. The optometry AI computing module includes a deep neural network model trained offline, which takes ocular physiological parameters as input and pixel compensation parameters as output.

Images