We present a construction vehicle tracking algorithm designed for complex construction site environments. Combining detection and tracking, our approach incorporates an enhanced Convolutional Block Attention Module (CBAM) into the YOLOv7 backbone network, addressing challenges in intricate construction settings and prioritizing valuable target objects. Additionally, our proposed static status detection module enhances the tracking performance of DeepSORT. Experiments on the expanded ACID and TDCV datasets show that our algorithm achieves a 96.35% map@0.5, surpassing YOLOv7. Additionally, it demonstrates superior tracking capabilities on the tracking dataset.