系统部署FastDeploy，长时间运行 提示opencv错误 0xc0000005

环境

【FastDeploy版本】： fastdeploy-win-gpu-1.0.6
【系统平台】: Windows x64(Windows10)
【硬件】： Nvidia GPU 1660Ti， CUDA 11.2 CUDNN 8.3
【编译语言】： C++

系统开启GPU推理，模型仅创建一次，定时循环3分钟GPU推理一次,。长时间运行，系统崩溃，查看window10系统错误日志，显示 错误模块名称: opencv_world3416.dll，异常代码: 0xc0000005.

错误模块名称: opencv_world3416.dll，版本: 3.4.16.0，时间戳: 0x6160c391
异常代码: 0xc0000005
错误偏移量: 0x000000000192295f
错误进程 ID: 0x1efc 

• 代码如下：

fastdeploy::vision::detection::YOLOv3 *model = nullptr;

void CreateModel(){
	string model_file =  "/model/yolov3_darknet53_270e_voc/model.pdmodel";
	string params_file ="/model/yolov3_darknet53_270e_voc7/model.pdiparams";
	string config_file = "/model/yolov3_darknet53_270e_voc/infer_cfg.yml";

	auto option = fastdeploy::RuntimeOption();
	option.UseGpu();
	if (model == nullptr)
	{
		model = new fastdeploy::vision::detection::YOLOv3(model_file, params_file, config_file, option);
		if (!model ->Initialized())
		{
			std::cerr << "Failed to initialize." << std::endl;
			return;
		}
        }
}

void ProcessData(const std::string& image_file){
  auto im = cv::imread(image_file);
  fastdeploy::vision::DetectionResult res;
  if (!model.Predict(&im, &res)) {
    std::cerr << "Failed to predict." << std::endl;
    return;
  }
  std::cout << res.Str() << std::endl;

  auto vis_im = fastdeploy::vision::VisDetection(im, res);
  cv::imwrite("vis_result.jpg", vis_im);
  std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl;
}

// 析构函数
delete model;

Start by reproducing the long-running GPU inference loop described in CreateModel and ProcessData on Windows 10 with fastdeploy-win-gpu-1.0.6, CUDA 11.2, and OpenCV opencv_world3416.dll. Check the Windows event log and the model, image, visualization, and destruction lifecycle; done means identifying the cause of the 0xc0000005 crash and confirming a stable long-running run.