public static void Close()
{
if (_DEnginer != IntPtr.Zero)
{
ArcWrapper.DClose(_DEnginer);
_DEnginer = IntPtr.Zero;
}
if (_DBuffer != IntPtr.Zero)
{
Marshal.FreeCoTaskMem(_DBuffer);
_DBuffer = IntPtr.Zero;
}
for (int i = 0; i < TaskNum; i++)
{
if (_REngine[i] != IntPtr.Zero)
{
ArcWrapper.RClose(_REngine[i]);
_REngine[i] = IntPtr.Zero;
}
if (_RBuffer[i] != IntPtr.Zero)
{
Marshal.FreeCoTaskMem(_RBuffer[i]);
_RBuffer[i] = IntPtr.Zero;
}
foreach (var item in _FaceLib[i].Items)
{
Marshal.FreeCoTaskMem(item.FaceModel.PFeature);
}
}
foreach (var item in CacheFaceResults.Items)
{
Marshal.FreeCoTaskMem(item.FaceModel.PFeature);
}
}
/// <summary>
/// 初始化，主要用于视频，取消人脸方向参数
/// </summary>
/// <param name="appId">虹软SDK的AppId</param>
/// <param name="dKey">虹软SDK人脸检测的Key</param>
/// <param name="rKey">虹软SDK人脸比对的Key</param>
/// <param name="orientPriority">脸部角度，毋宁说是鼻子方向，上下为0或180度，左右为90或270度</param>
/// <param name="scale">最小人脸尺寸有效值范围[2,50] 推荐值 16。该尺寸是人脸相对于所在图片的长边的占比。例如，如果用户想检测到的最小人脸尺寸是图片长度的 1/8，那么这个 scale 就应该设置为8</param>
/// <param name="maxFaceNumber">用户期望引擎最多能检测出的人脸数有效值范围[1,100]</param>
/// <param name="faceDataPath">人脸数据文件夹</param>
/// <param name="message"></param>
/// <returns></returns>
public static bool Init(out string message, string appId, string dKey, string rKey, int scale = 50, int maxFaceNumber = 10, string faceDataPath = "d:\\FeatureData")
{
if (scale < 2 || scale > 50)
{
message = "scale的值必须在2-50之间";
return(false);
}
if (maxFaceNumber < 1 || maxFaceNumber > 100)
{
message = "人脸数量必须在1-100之间";
return(false);
}
_DBuffer = Marshal.AllocCoTaskMem(20 * 1024 * 1024);
var initResult = (ErrorCode)ArcWrapper.DInit(appId, dKey, _DBuffer, 20 * 1024 * 1024, out _DEnginer, (int)ArcApi.EOrientPriority.Only0, scale, maxFaceNumber);
if (initResult != ErrorCode.Ok)
{
message = $"初始化人脸检测引擎失败，错误代码:{(int)initResult}，错误描述：{ initResult.GetType().GetMember(initResult.ToString()).FirstOrDefault()?.GetCustomAttribute<DescriptionAttribute>().Description ?? initResult.ToString()}";
return(false);
}
for (int i = 0; i < TaskNum; i++)
{
_RBuffer[i] = Marshal.AllocCoTaskMem(40 * 1024 * 1024);
initResult = (ErrorCode)ArcWrapper.RInit(appId, rKey, _RBuffer[i], 40 * 1024 * 1024, out _REngine[i]);
if (initResult != ErrorCode.Ok)
{
message = $"初始化人脸比对引擎失败，错误代码:{(int)initResult}，错误描述：{ initResult.GetType().GetMember(initResult.ToString()).FirstOrDefault()?.GetCustomAttribute<DescriptionAttribute>().Description ?? initResult.ToString()}";
return(false);
}
_FaceLib[i] = new FaceLib();
}
CacheFaceResults = new FaceResults(maxFaceNumber);
_MaxFaceNumber = maxFaceNumber;
_FaceDataPath = faceDataPath;
_FaceImagePath = Path.Combine(_FaceDataPath, "Image");
if (!Directory.Exists(faceDataPath))
{
Directory.CreateDirectory(faceDataPath);
}
if (!Directory.Exists(_FaceImagePath))
{
Directory.CreateDirectory(_FaceImagePath);
}
int index = 0;
//for (int i = 0; i < 100; i++)
foreach (var file in Directory.GetFiles(faceDataPath))
{
var info = new FileInfo(file);
var data = File.ReadAllBytes(file);
var pFeature = Marshal.AllocCoTaskMem(data.Length);
Marshal.Copy(data, 0, pFeature, data.Length);
_FaceLib[index % TaskNum].Items.Add(new FaceLib.Item()
{
OrderId = 0, ID = info.Name.Replace(info.Extension, ""), FaceModel = new FaceModel {
Size = FeatureSize, PFeature = pFeature
}
});
index++;
}
message = "初始化成功";
return(true);
}
/// <summary>
/// 人脸比对
/// </summary>
/// <param name="bitmap">输入图片</param>
public static void FaceMatch(Bitmap bitmap)
{
var bmpData = bitmap.LockBits(new Rectangle(0, 0, bitmap.Width, bitmap.Height), ImageLockMode.ReadOnly, PixelFormat.Format24bppRgb);
var imageData = new ImageData
{
PixelArrayFormat = 513,//Rgb24,
Width = bitmap.Width,
Height = bitmap.Height,
Pitch = new int[4] {
bmpData.Stride, 0, 0, 0
},
ppu8Plane = new IntPtr[4] {
bmpData.Scan0, IntPtr.Zero, IntPtr.Zero, IntPtr.Zero
}
};
var ret = (ErrorCode)ArcWrapper.Detection(_DEnginer, ref imageData, out var pDetectResult);
if (ret != ErrorCode.Ok)
{
bitmap.UnlockBits(bmpData);
return;
}
var detectResult = Marshal.PtrToStructure <DetectResult>(pDetectResult);
CacheFaceResults.FaceNumber = detectResult.FaceCout;
for (int i = detectResult.FaceCout; i < _MaxFaceNumber; i++)
{
CacheFaceResults[i].ID = "";
}
if (detectResult.FaceCout == 0)
{
bitmap.UnlockBits(bmpData);
return;
}
if (detectResult.FaceCout == 1)
{
CacheFaceResults[0].FFI.FaceRect = Marshal.PtrToStructure <FaceRect>(detectResult.PFaceRect);
if (ArcWrapper.ExtractFeature(_REngine[0], ref imageData, ref CacheFaceResults[0].FFI, out var fm) == (int)ErrorCode.Ok)
{
ArcWrapper.CopyMemory(CacheFaceResults.Items[0].FaceModel.PFeature, fm.PFeature, FeatureSize);
}
}
else
{
Task[] ts = new Task[TaskNum < detectResult.FaceCout ? TaskNum : detectResult.FaceCout];
int faceOffset = -1;
for (int i = 0; i < ts.Length; i++)
{
var rEngine = _REngine[i];
ts[i] = Task.Factory.StartNew(() =>
{
int faceIndex = 0;
while ((faceIndex = Interlocked.Increment(ref faceOffset)) < detectResult.FaceCout)
{
CacheFaceResults[faceIndex].FFI.FaceRect = Marshal.PtrToStructure <FaceRect>(IntPtr.Add(detectResult.PFaceRect, faceIndex * Marshal.SizeOf <FaceRect>()));
if (ArcWrapper.ExtractFeature(rEngine, ref imageData, ref CacheFaceResults[faceIndex].FFI, out var fm) == (int)ErrorCode.Ok)
{
ArcWrapper.CopyMemory(CacheFaceResults.Items[faceIndex].FaceModel.PFeature, fm.PFeature, FeatureSize);
}
}
});
}
Task.WaitAll(ts);
}
bitmap.UnlockBits(bmpData);
var tsr = new Task[TaskNum];
List <int> noMatchFaces = Enumerable.Range(0, detectResult.FaceCout).ToList();
for (int i = 0; i < TaskNum; i++)
{
var taskIndex = i;
tsr[i] = Task.Factory.StartNew(() =>
{
var rEngine = _REngine[taskIndex];
foreach (var item in _FaceLib[taskIndex].Items.OrderByDescending(ii => ii.OrderId))
{
_RWL.EnterReadLock();
if (noMatchFaces.Count == 0)
{
_RWL.ExitReadLock();
break;
}
var faceIndexs = noMatchFaces.ToList();
_RWL.ExitReadLock();
foreach (var faceIndex in faceIndexs)
{
ArcWrapper.Match(rEngine, ref CacheFaceResults.Items[faceIndex].FaceModel, ref item.FaceModel, out float score);
if (score > 0.55)
{
_RWL.EnterWriteLock();
noMatchFaces.Remove(faceIndex);
_RWL.ExitWriteLock();
CacheFaceResults[faceIndex].ID = item.ID;
CacheFaceResults[faceIndex].Score = (float)Math.Round(score * 100, 2);
item.OrderId = DateTime.Now.Ticks;
}
}
}
});
}
Task.WaitAll(tsr);
foreach (var faceIndex in noMatchFaces)
{
CacheFaceResults[faceIndex].ID = "";
CacheFaceResults[faceIndex].Score = 0;
}
}
