using System;
|
using System.Collections.Generic;
|
using System.Linq;
|
using System.Text;
|
using System.Threading.Tasks;
|
using System.Runtime.InteropServices;
|
using System.IO;
|
using OpenCvSharp;
|
|
// 提取人脸特征值
|
namespace FaceAI
|
{
|
// 人脸特征值结构体
|
[StructLayout(LayoutKind.Sequential, Pack = 1)]
|
public struct BDFaceFeature
|
{
|
public int size;
|
// 人脸的特征值，提取出来后是128个float的浮点值
|
[MarshalAs(UnmanagedType.ByValArray, SizeConst = 128)]
|
public float[] data;// = new float[128];
|
}
|
// 人脸特征值结构体
|
[StructLayout(LayoutKind.Sequential, Pack = 1)]
|
public struct BDDepthFeature
|
{
|
public int size;
|
// 人脸的深度特征值，提取出来后是256个float的浮点值
|
[MarshalAs(UnmanagedType.ByValArray, SizeConst = 256)]
|
public float[] data;// = new float[256];
|
}
|
class FaceFeature
|
{
|
// 提取人脸特征值
|
[DllImport("BaiduFaceApi.dll", EntryPoint = "face_feature", CharSet = CharSet.Ansi
|
, CallingConvention = CallingConvention.Cdecl)]
|
// 返回num为特征值的人数,type为0表示提取生活照的特征值，1表示nir的特征值
|
public static extern int face_feature(IntPtr feature, IntPtr box, IntPtr mat, int type);
|
|
|
|
public void test_face_feature()
|
{
|
long t_begin = TimeUtil.get_time_stamp();
|
BDFaceFeature[] fea1 = test_get_face_feature_by_path("../images/2.jpg");
|
long t_end = TimeUtil.get_time_stamp();
|
Console.WriteLine("get feature time cost is:" + (t_end - t_begin));
|
}
|
|
// 获取特征值，传入图片路径
|
public BDFaceFeature[] test_get_face_feature_by_path(string img_path)
|
{
|
// 特征值的长度，128个float值
|
int dim_count = 128;
|
Mat mat = Cv2.ImRead(img_path);
|
// 假设提取的人数，需要比实际的人数多，因为需要提取分配内存
|
int faceNum = 5;
|
BDFaceFeature[] feaList = new BDFaceFeature[faceNum];
|
for(int i = 0; i < faceNum; i++)
|
{
|
feaList[i].data = new float[dim_count];
|
feaList[i].size = 0;
|
}
|
|
if (mat.Empty())
|
{
|
return null;
|
}
|
|
int sizeFeature = Marshal.SizeOf(typeof(BDFaceFeature));
|
IntPtr ptFea = Marshal.AllocHGlobal(sizeFeature * faceNum);
|
|
// 构造返回的人脸框数据
|
BDFaceBBox[] info = new BDFaceBBox[faceNum];
|
int sizeBox = Marshal.SizeOf(typeof(BDFaceBBox));
|
IntPtr ptBox = Marshal.AllocHGlobal(sizeBox * faceNum);
|
// 返回num为特征值的人数,type为0表示提取生活照的特征值，1表示证件照的特征值，2表示nir的特征值
|
int type = 0;
|
int num = face_feature(ptFea, ptBox, mat.CvPtr, type);
|
if (num <= 0)
|
{
|
return null;
|
}
|
Console.WriteLine("face num is:{0}", num);
|
// 请确保faceNum大于num, faceNum为c#期望的检测人数，需要预先分配内存，否则无法获取
|
if (num > faceNum)
|
{
|
num = faceNum;
|
}
|
for (int index = 0; index < num; index++)
|
{
|
IntPtr ptrF = new IntPtr();
|
IntPtr ptrB = new IntPtr();
|
if (8 == IntPtr.Size)
|
{
|
ptrF = (IntPtr)(ptFea.ToInt64() + sizeFeature * index);
|
ptrB = (IntPtr)(ptBox.ToInt64() + sizeBox * index);
|
}
|
else if (4 == IntPtr.Size)
|
{
|
ptrF = (IntPtr)(ptFea.ToInt32() + sizeFeature * index);
|
ptrB = (IntPtr)(ptBox.ToInt32() + sizeBox * index);
|
}
|
feaList[index] = (BDFaceFeature)Marshal.PtrToStructure(ptrF, typeof(BDFaceFeature));
|
Console.WriteLine("feaList[index].size is:{0}", feaList[index].size);
|
|
|
for (int k = 0; k < feaList[index].size; k++)
|
{
|
Console.WriteLine("feature is {0}:", feaList[index].data[k]);
|
}
|
|
info[index] = (BDFaceBBox)Marshal.PtrToStructure(ptrB, typeof(BDFaceBBox));
|
|
// 索引值
|
Console.WriteLine("detect score is:{0}", info[index].index);
|
// 置信度
|
Console.WriteLine("detect score is:{0}", info[index].score);
|
// 人脸宽度
|
Console.WriteLine("detect mWidth is:{0}", info[index].width);
|
// 中心点X,Y坐标
|
Console.WriteLine("detect mCenter_x is:{0}", info[index].center_x);
|
Console.WriteLine("detect mCenter_y is:{0}", info[index].center_y);
|
}
|
// 绘制人脸框
|
FaceDraw.draw_rects(ref mat, faceNum, info);
|
|
//mat.ImWrite("detect.jpg");
|
mat.Release();
|
return feaList;
|
}
|
|
}
|
}
|
