<p>Real-time semantic segmentation in urban traffic scenes must preserve fine structures while capturing long-range context under strict latency constraints. We propose HCTNet, a hybrid CNN–Mamba framework that performs single-branch CNN inference and leverages a training-only Mamba auxiliary branch to inject global context during optimization. The method introduces a lightweight Convolutional State Module (CSM) to enlarge the effective receptive field within the CNN backbone and a Feature Alignment Module (FAM) to align multi-scale representations from the CNN and Mamba branches via spatial/channel projections and gated fusion. A shared decoder is used for all streams during training to enforce a common prediction space; at test time, only the CNN path with the shared decoder is executed to retain real-time efficiency. On Cityscapes, against strong real-time baselines under a unified evaluation protocol, HCTNet attains 81.0% mean intersection-over-union (mIoU) at 60.5 frames per second (FPS), and achieves 80.3% mIoU at 98.6 FPS under a reduced input scale. On ApolloScape under a 19-class mapping protocol, HCTNet obtains 73.8% mIoU, demonstrating strong cross-dataset generalization. Qualitative results show sharper boundaries and more coherent predictions for small and distant objects. Ablation studies indicate that the gains arise from the complementary effects of CSM, the training-time Mamba branch, multi-scale alignment through FAM, and the shared decoder, while keeping inference cost unchanged.</p>