The continuous rainy and snowy weather in the south can lead to snow and ice accumulation on roads, thereby reducing vehicle grip, increasing braking distance, and easily causing traffic accidents, resulting in a decrease in road capacity and posing many challenges to the normal operation of cities and people's daily life and travel. We addressed a series of issues with existing snow detection methods, such as low detection accuracy and poor real-time performance. By incorporating the CBAM dual attention mechanism module, we designed the VGG16-Unet semantic segmentation model framework, which can accurately identify road snow and perform snow segmentation on collected road paintings, maximizing the accuracy of snow detection. The experimental results show that the detection accuracy of CVUnet reaches over 95%, which can be well applied to the task of snow segmentation.