This paper tackles the challenges of spectrum scarcity and network efficiency in the 5G IoV environment using dynamic spectrum sharing and network slicing, supported by deep learning. It aims to improve spectrum use and network resource allocation, addressing limited resources and uneven network load. The choice of 5G NR is due to its higher bandwidth and lower latency, crucial for IoV applications. A deep learning model monitors and predicts real-time spectrum usage, followed by a dynamic sharing mechanism adjusting allocation based on demand and network status. Network slicing then distributes traffic according to application requirements, ensuring QoS. Tested in a simulated environment, results show a 2% increase in spectrum utilization to 62% for 100 vehicles, latency reduced from 58 to 55 ms, and QoS satisfaction up from 78 to 80%. The proposed technology effectively resolves resource and service quality issues, providing a foundation for future intelligent IoV systems.