With the rapid advancement of cloud computing and artificial intelligence, the aviation industry is increasingly integrating intelligent technologies. A growing number of intelligent applications are being deployed in airborne computing systems, where workloads are typically dynamic and variable. Container technology, due to its lightweight, scalable, and flexible deployment capabilities, provides an effective execution environment for these applications. However, large-scale computational tasks within containers often require GPU acceleration. The mainstream container orchestration platform, Kubernetes, allocates GPUs exclusively to individual containers, resulting in low GPU utilization. To overcome this limitation, this paper investigates the domestic GPU programming framework OpenCL and proposes vOpenCL, a GPU virtualization solution designed for Kubernetes-based environments. vOpenCL virtualizes a single physical GPU into multiple virtual GPUs (vGPUs) through API redirection, enabling dynamic allocation and resource sharing among containers. Additionally, it employs a flexible resource allocation strategy to improve GPU utilization efficiency. Experimental results demonstrate that vOpenCL successfully implements GPU virtualization on domestic hardware, allowing containers to share memory and computational resources while ensuring effective resource isolation. This approach significantly enhances overall resource utilization and provides a practical, efficient solution for intelligent airborne computing systems.