Abstract: In recent years, the national vision crisis has escalated into a prominent social issue, with a growing number of reports highlighting the harmful effects of modern digital devices on eye health. While the popularity of electronic devices has significantly improved the quality of life and work, at the same time devices such as computers, smartphones and televisions have put unprecedented stress and damage on an individual's visual health. With the development of artificial intelligence, numerous smart glasses have emerged in the market, many of which are designed to protect vision. However, existing vision protection devices on the market face issues such as limited data analysis and inaccurate visual environment assessments. For instance, some smart glasses for vision protection can only analyze a single light source, such as blue light, and are unable to simultaneously analyze diverse light sources. In order to solve these problems, this paper designs a portable smart glasses and proposes a multimodal data-driven intelligent eye health analysis method. The designed glasses integrate several high-precision sensors to comprehensively monitor and analyze eye health. These include a visible light spectrum sensor, a flicker detection sensor, and a glare sensor. The visible light spectrum sensor allows the glasses to perform a precise spectral analysis of ambient light, decompose visible light into different wavelengths, and capture spectral information of different colors. This is especially important for detecting harmful low-frequency blue light. The flicker detection sensor is optimized based on the LED flicker measurement method provided by the Solid-State Lighting Systems and Technology Consortium, which monitors flicker frequency. The Glare Sensor is equipped with a sensor to measure the Uniform Glare Rating (UGR), which allows for the evaluation of the overall glare effect produced by light distribution and luminance. Light source data from the three modes is collected and processed by the three sensors and input to a data analysis system, which receives processed spectral data, glare data, and UGR values, which reflect the wavelength (color component) of the ambient light, the intensity of the flicker frequency, and the overall glare effect. The input variables are then converted into fuzzy sets, with blue light radiation represented by two fuzzy variables, flicker frequency by three fuzzy variables and UGR by five fuzzy variables. The system utilizes the Mamdani method incorporating fuzzy rules, and the defuzzification process employs the center-of-mass method to determine a discrete output value, which ultimately maps the fuzzy inputs to the fuzzy outputs. This study is also tested through comparative simulation experiments. In order to ensure external validity, authenticity, and reliability of the study, participants were recruited to simulate poor eye-use habits, including prolonged computer use, working in poor lighting, and exposure to bright light. Data were collected using a 5-second sampling period, yielding 4,574 sampling periods of raw data. Participants wore the smart glasses to collect the data, which was then preprocessed to remove outliers. The processed data, including blue light radiation intensity, flicker frequency, and glare index, were input into a fuzzy logic inference system. Ultimately, this study improved the accuracy by 15.3%, recall by 62.4%, and f1 value by 25.2% compared to the existing methods. The experimental results validate the effectiveness and efficiency of the proposed method.