Abstract: In recent years, the issue of people’s vision has become increasingly important and has emerged as a social issue that has attracted widespread attention. Multiple survey reports indicate that while the widespread use of modern digital devices, such as computers, smartphones, and televisions, has greatly enhanced the quality of people’s lives and work, it has also caused unprecedented strain and damage to their eyes. With the development of artificial intelligence, numerous intelligent devices have emerged on the market, many of which are designed to protect vision. However, existing vision protection devices often encounter problems, such as limited data analysis and inaccurate assessments of the visual environment in which they are used. For instance, some smart glasses designed for vision protection can only analyze a single light source, such as blue light, and are unable to simultaneously analyze multiple diverse light sources. To address these challenges, this study proposes a multimodal data-driven intelligent eye health analysis method, designs a portable pair of smart glasses at the hardware level, and designs a multimodal data fusion analysis method based on fuzzy logic reasoning at the software level. The glasses are designed with a variety of high-precision sensors, which can comprehensively collect ambient light data from various sources. These sensors include three types: visible light spectrum sensor, flicker detection sensor, and 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 across the color spectrum. This is especially important for detecting harmful low-frequency blue light. The flicker detection sensor is based on the light-emitting diode flicker measurement method provided by the Solid-State Lighting Systems and Technology Consortium to monitor flicker frequency. The glare sensor is equipped to measure the uniform glare rating, which is used to evaluate the overall glare effect produced by light distribution and luminance. The raw data from the three types of sensors are processed to obtain blue light radiation, the intensity of the flicker frequency, uniform glare rating, and other data. These data are blurred in this study. Blue light is represented by two fuzzy variables, flicker by three fuzzy variables, and uniform glare rating by five fuzzy variables. Fuzzy logic is used to process this data according to fuzzy rules to judge the degree of harm the current environment causes to the eyes. This study designs three sets of fuzzy rules, which are conservative, moderate, and aggressive. In addition, this study conducts comparative experiments using self-built datasets. A group of college students was recruited as experimental participants to simulate various eye use scenarios. The participants wore smart glasses prototypes and engaged in activities such as reading, playing games, using electronic devices in a dark room, watching videos, and performing office work, while experimental data were collected. Experimental results show that compared with similar methods, the proposed method achieves a 14.53%, 26.13%, and 17.72% improvement in accuracy, recall, and F₁ value, respectively. The experimental results validate the effectiveness and efficiency of the proposed method.