VS2015 + opencv4.2.0获取堡盟工业相机的实时图像（USB版）

堡盟相机的SDK不夸张的说应该是最好懂最容易上手的相机SDK，哪些接口对用户开放，哪些是内部封装都很清楚，还提供了比较完备的示例。
1.堡盟相机SDK简介
提供了C,C#,C++编程接口；
提供比较详尽的SDK解读（见图）

有针对opencv的示例。
2.环境配置
1）头文件和静态库配置
打开一个示例代码，发现包含了如下文件

#include <iostream>
#include <iomanip>
#include <string>
#include <sstream>
#include <vector>
#include "bgapi2_genicam/bgapi2_genicam.hpp"

#if _WIN32
#   include <direct.h>
#else
#   include <sys/stat.h>
#   include <stdlib.h>
#endif

#if USE_OPENCV
#   include "opencv2/opencv.hpp"
#   include "opencv2/highgui/highgui.hpp"

#if     USE_OCL_COMPONENT == 3
#           include "opencv2/core/ocl.hpp"
#elif   USE_OCL_COMPONENT == 2
#           include "opencv2/ocl/ocl.hpp"
#endif  // USE_OCL_COMPONENT
#endif  // USE_OPENCV

a.首先需要配置“bgapi2_genicam/bgapi2_genicam.hpp”的路径，右键项目->属性->VC++目录->包含目录，添加路径“C:\Program Files\Baumer GAPI SDK (64 Bit)\Components\Dev\C++\Inc”；然后点击库目录，添加库目录“C:\Program Files\Baumer GAPI SDK (64 Bit)\Components\Dev\C++\Lib”
点击链接器->输入->附加依赖项：添加bgapi2_genicam.lib。
b.按照网上通用方法配置opencv4.2.0。
2）动态库配置
将路径“C:\Program Files\Baumer GAPI SDK (64 Bit)\Components\Bin”下所有文件拷贝到工程debug目录。
3）需要添加#include "stdafx.h"的项目记着把这个填上。
4）解决方案配置选debug，平台选x64
3.关键代码
源代码多达1000多行，有兴趣的看https://blog.csdn.net/smalbig/article/details/104658550，这里列出关键代码。

			pDataStream->StartAcquisitionContinuous();
			while(true)
			{
				//WAIT FOR IMAGE
				BGAPI2::Buffer* pBufferFilled = pDataStream->GetFilledBuffer(uGetFilledBufferTimeout); //timeout <n> msec
				BGAPI2::ImageProcessor* pImageProcessor = new BGAPI2::ImageProcessor();
				void* pBufferDst = NULL;
				BGAPI2::Image* pImage = NULL;
				void* pRawBuffer = pBufferFilled->GetMemPtr();
				const int width = (int)(pBufferFilled->GetWidth());
				const int height = (int)(pBufferFilled->GetHeight());

				if (pBufferFilled == NULL)
				{
					std::cout << "Error: Buffer Timeout after " << uGetFilledBufferTimeout << "msec" << std::endl;
				}
				else
				{
					if (pBufferFilled->GetIsIncomplete() == true)
					{
						std::cout << "Error: Image is incomplete" << std::endl;
					}
					else
					{
						const bo_uint w = static_cast<bo_uint>(pBufferFilled->GetWidth());
						const bo_uint h = static_cast<bo_uint>(pBufferFilled->GetHeight());
						void* const pMemSrc = pBufferFilled->GetMemPtr();
						BGAPI2::String sPixelFormatSrc = pBufferFilled->GetPixelFormat();
						const bo_uint64 memSizeSrc = pBufferFilled->GetMemSize();

						std::cout << " Image " << std::setw(5) << pBufferFilled->GetFrameID() << " received in memory address " << std::hex << pMemSrc << std::dec
							<< " [" << sPixelFormatSrc << "]" << std::endl;

						if (pImageSrc == NULL) {
							pImageSrc = pImageProcessor->CreateImage(w, h, sPixelFormatSrc, pMemSrc, memSizeSrc);
						}
						else
						{
							pImageSrc->Init(w, h, sPixelFormatSrc, pMemSrc, memSizeSrc);
						}

						if (pImageSrc != NULL)
						{
							bo_uint64 bufferLength = pImageSrc->GetTransformBufferLength(sPixelFormatDst);

							if (bufferLength > memSizeDst)
							{
								if (pMemDst != NULL) {
									delete[] pMemDst;
									pMemDst = NULL;
									memSizeDst = 0;
								}

								pMemDst = new bo_uchar[static_cast<size_t>(bufferLength)];
								memSizeDst = bufferLength;
							}

							std::cout << " convert to " << sPixelFormatDst << std::endl;
							pImageProcessor->TransformImageToBuffer(pImageSrc, sPixelFormatDst, pMemDst, bufferLength);
							cv::Mat img(height, width, CV_8UC3, pMemDst, cv::Mat::AUTO_STEP);
							cv::namedWindow("这波怎么说bbbbb", cv::WINDOW_AUTOSIZE);
							cv::imshow("这波怎么说bbbbb", img);

							cv::cvtColor(img, img, cv::COLOR_BGR2GRAY);
							cv::Mat dst;

							cv::threshold(img, dst, 47, 255, cv::THRESH_BINARY);
							cv::namedWindow("这波怎么说aaaaa", cv::WINDOW_AUTOSIZE);
							cv::imshow("这波怎么说aaaaa", dst);

							//pBuffer2->QueueBuffer();

							cv::waitKey(20);
						}
					}

					// queue buffer again
					pBufferFilled->QueueBuffer();
				}
			}

前面看不明白没关系，可以去堡盟官网下SDK慢慢看，都是相机sdk自带的一些预处理代码。跟opencv有关的是这几句：

cv::Mat img(height, width, CV_8UC3, pMemDst, cv::Mat::AUTO_STEP);
					cv::namedWindow("这波怎么说bbbbb", cv::WINDOW_AUTOSIZE);
					cv::imshow("这波怎么说bbbbb", img);

					cv::cvtColor(img, img, cv::COLOR_BGR2GRAY);
					cv::Mat dst;

					cv::threshold(img, dst, 47, 255, cv::THRESH_BINARY);
					cv::namedWindow("这波怎么说aaaaa", cv::WINDOW_AUTOSIZE);
					cv::imshow("这波怎么说aaaaa", dst);

可以看到pMemDst是void*，可以通过

Mat img(height, width, CV_8UC3, pMemDst,cv::Mat::AUTO_STEP);

构造出一个Mat对象，然后就可以对这个Mat对象为所欲为啦！！！
上面构造代码中height在前，width在后，这是我修改后的代码，经测试没问题；堡盟的源代码是width在前，height在后，测试后出现玄幻的图像重叠，可能是对opencv4.2.0不够支持的缘故。