百度智能云平臺的人臉識別項目，旨在利用其強大的人臉識別服務(wù)實現(xiàn)自動人臉識別。選擇百度智能云的原因是其高效的API接口和穩(wěn)定的服務(wù)質(zhì)量，能夠幫助開發(fā)者快速實現(xiàn)人臉識別應(yīng)用。

本項目使用攝像頭捕捉圖像后，通過百度智能云平臺的人臉識別服務(wù)，能夠輕松識別圖像中的人臉，并將識別結(jié)果實時顯示在Qt界面上。

功能特性

1、圖片處理和人臉識別：使用百度智能云的人臉識別服務(wù)，通過API輕松識別圖像中的人臉。

2、攝像頭實時采集圖像并保存：使用Qt設(shè)計了直觀的用戶界面，控制USB攝像頭的打開、關(guān)閉以及實時顯示攝像頭捕獲的視頻流，并將采集到的視頻流保存為圖像。

環(huán)境說明

1、開發(fā)環(huán)境操作系統(tǒng)：Ubuntu18.04 64位版

2、交叉編譯工具鏈：arm-poky-linux-gnueabi-gcc 5.3.0

3、開發(fā)板使用Bootloader版本：u-boot-2016.03

4、開發(fā)板內(nèi)核版本：linux-4.1.15

5、開發(fā)板移植QT版本：qt5.6.2

圖片處理和人臉識別

百度智能云網(wǎng)址：cloud.baidu.com

本次人臉識別的方案是通過百度智能云平臺進(jìn)行實現(xiàn)的。首先進(jìn)入百度智能云網(wǎng)頁- >選擇人臉與人體- > 人臉識別。

進(jìn)入人臉識別頁面之后可通過閱讀技術(shù)文檔來學(xué)習(xí)人臉識別的使用方法。

1、創(chuàng)建人臉庫

（1）單擊“應(yīng)用列表”，單擊“查看人臉庫”。

（2）單擊“新建組”。

（3）選擇用戶組場景類型，填寫組ID，單擊“確認(rèn)”。

（4）單擊剛新建的組ID。

（5）單擊“新建用戶”，填寫用戶ID，添加圖片，單擊“確認(rèn)”。

2、在線識別人臉圖片

在本地實現(xiàn)之前可通過平臺提供的在線驗證方法進(jìn)行驗證，如下圖，需要在旁邊輸入一張人臉圖片的base64 編碼的字符串或者選擇上傳一張人臉圖片，并填寫組ID即可進(jìn)行在線識別。

3、識別本地人臉圖片

本地實現(xiàn)人臉識別的方法需要將識別代碼拷貝到本地，并需要實現(xiàn)一個將圖片轉(zhuǎn)換為base64編碼的函數(shù)。需要輸入自己的access_token（通過閱讀文檔可知怎么獲取）。

#include #include #include #include #include #include #include #include #include #include #include #include inline size_t onWriteData(void * buffer, size_t size, size_t nmemb, void * userp) { std::string * str = dynamic_cast((std::string *)userp); str->append((char *)buffer, size * nmemb); return nmemb;

} std::string getFileBase64Content(const char * path, bool urlencoded=false) { const std::string base64_chars = "ABCDEFGHIJKLMNOPQRSTUVWXYZ" "abcdefghijklmnopqrstuvwxyz" "0123456789+/"; std::string ret;

int i = 0; int j = 0; unsigned char char_array_3[3]; unsigned char char_array_4[4]; unsigned int bufferSize = 1024;

unsigned char buffer[bufferSize]; std::ifstream file_read; file_read.open(path, std::ios::binary); while (!file_read.eof()){ file_read.read((char *) buffer, bufferSize * sizeof(char)); int num = file_read.gcount(); int m = 0; while (num--){ char_array_3[i++] = buffer[m++];

if(i == 3){ char_array_4[0] = (char_array_3[0] & 0xfc) >> 2; char_array_4[1] = ((char_array_3[0] & 0x03) << 4) + ((char_array_3[1] & 0xf0) >> 4); char_array_4[2] = ((char_array_3[1] & 0x0f) << 2) + ((char_array_3[2] & 0xc0) >> 6);

char_array_4[3] = char_array_3[2] & 0x3f; for(i = 0; (i <4) ; i++) ret += base64_chars[char_array_4[i]]; i = 0; } } } file_read.close(); if(i){ for(j = i; j < 3; j++) char_array_3[j] = '\0'; char_array_4[0] = (char_array_3[0] & 0xfc) >> 2; char_array_4[1] = ((char_array_3[0] & 0x03) << 4) + ((char_array_3[1] & 0xf0) >> 4); char_array_4[2] = ((char_array_3[1] & 0x0f) << 2) + ((char_array_3[2] & 0xc0) >> 6); char_array_4[3] = char_array_3[2] & 0x3f; for(j = 0; (j < i + 1); j++) ret += base64_chars[char_array_4[j]]; while((i++ < 3)) ret += '='; }

if (urlencoded) ret = curl_escape(ret.c_str(), ret.length()); return ret; } std::string performCurlRequest(const char *pic_path, const std::string &token) { std::string result; char *web_curl = nullptr; CURL *curl = curl_easy_init(); CURLcode res = CURLE_OK;

if (!asprintf(&web_curl, "https://aip.baidubce.com/rest/2.0/face/v3/search?access_token=%s", token.c_str())) { perror("asprintf error"); }

if (curl) { curl_easy_setopt(curl, CURLOPT_CUSTOMREQUEST, "POST");

// std::string url = "https://aip.baidubce.com/rest/2.0/face/v3/search?access_token=" + token; curl_easy_setopt(curl, CURLOPT_URL, web_curl); curl_easy_setopt(curl, CURLOPT_FOLLOWLOCATION, 1L); curl_easy_setopt(curl, CURLOPT_DEFAULT_PROTOCOL, "HTTPS"); curl_easy_setopt(curl, CURLOPT_SSL_VERIFYPEER, 0L); curl_easy_setopt(curl, CURLOPT_SSL_VERIFYHOST, 0L);

struct curl_slist *headers = NULL; headers = curl_slist_append(headers, "Content-Type: application/json");

curl_easy_setopt(curl, CURLOPT_HTTPHEADER, headers); std::string base64_image = getFileBase64Content(pic_path, true); std::string post_data = "image=" + base64_image + "&group_id_list=one&image_type=BASE64";

curl_easy_setopt(curl, CURLOPT_POSTFIELDS, post_data.c_str()); curl_easy_setopt(curl, CURLOPT_WRITEDATA, &result); curl_easy_setopt(curl, CURLOPT_WRITEFUNCTION, onWriteData);

if(curl_easy_perform(curl) != CURLE_OK) fprintf(stderr, "Curl request failed: %s\n", curl_easy_strerror(res)); }

curl_easy_cleanup(curl); free(web_curl); return result; } int main(int argc, char *argv[])

{ std::string result; std::string name; std::string token = "24.2bc619cf9c09c32ce5af202ccc98c0c9.2592000.1724918062.282335-100710397"; result = performCurlRequest("/home/root/num/1.jpg", token); std::string json = result; std::regex pattern(""user_id":"(.*?)""); std::smatch match; if (std::regex_search(json, match, pattern)) { name = match[1].str(); std::cout << "read name is: " << name << std::endl; } return 0; }

1）依賴庫編譯

編譯人臉識別的應(yīng)用需要依賴Curl庫、OpenSSL庫、OpenCv庫、JsonCPP庫。詳細(xì)的依賴庫安裝步驟請參考以下鏈接：

bbs.elfboard.com/forum.php?mod=viewthread&tid=496&extra=page%3D1

bbs.elfboard.com/forum.php?mod=viewthread&tid=495&extra=page%3D1

bbs.elfboard.com/forum.php?mod=viewthread&tid=497&extra=page%3D1

bbs.elfboard.com/forum.php?mod=viewthread&tid=498&extra=page%3D1

2）應(yīng)用編譯

elf@ubuntu:~/work$ . /opt/fsl-imx-x11/4.1.15-2.0.0/environment-setup-cortexa7hf-neon-poky-linux-gnueabi elf@ubuntu:~/work$ $CC demoFace.cpp -o demoFace -I /home/elf/work/opencv-3.4.1/install/include/ -I /home/elf/work/curl-7.71.1/install/include/ -I /home/elf/work/jsoncpp-1.9.5/install/include/ -L /home/elf/work/opencv-3.4.1/install/lib/ -L /home/elf/work/curl-7.71.1/install/lib/ -L /home/elf/work/jsoncpp-1.9.5/install/lib/ -lopencv_highgui -lopencv_core -lopencv_imgproc -lopencv_objdetect -lopencv_videoio -lopencv_imgcodecs -std=c++11 -lcurl -lcrypto -ljsoncpp -lstdc++

編譯完成將文件通過scp拷貝到ELF 1開發(fā)板運行即可，這樣就可以將本地的圖片通過HTTPS發(fā)送到百度智能云進(jìn)行識別，并將識別結(jié)果返回。

攝像頭實時采集圖像并保存

1、程序設(shè)計

在前面一個章節(jié)實現(xiàn)了對本地人臉圖片的識別，下面來介紹如何通過攝像頭進(jìn)行人臉識別，采用USB攝像頭進(jìn)行識別，程序設(shè)計如下圖所示。

主函數(shù)的實現(xiàn)main.cpp

int main(int argc, char *argv[]) { QApplication a(argc, argv); Camera w; w.setWindowFlags(w.windowFlags()& ~Qt::WindowMaximizeButtonHint& ~Qt::WindowMinimizeButtonHint ); w.showMaximized(); w.show(); return a.exec(); }

設(shè)置UI

ui->setupUi(this); timer = new QTimer; QDesktopWidget* desktopWidget = QApplication::desktop(); QRect screenRect = desktopWidget->screenGeometry(); qDebug("screen.width = %d , screen.height = %d",screenRect.width(),screenRect.height()); this->imageWidget = new ImageWidget(this); this->imageWidget->setBackgroundRole(QPalette::Dark); this->imageWidget->setSizePolicy(QSizePolicy::Ignored, QSizePolicy::Ignored); this->imageWidget->setObjectName(QString::fromUtf8("imageWidget")); if(screenRect.width()==800) { ui->pbt_start->setGeometry(60,300,70,50); ui->pbt_stop->setGeometry(190,300,70,50); this->imageWidget->setGeometry(QRect(5, 30, 350, 250)); } else if(screenRect.width()>800) { ui->pbt_start->setGeometry(80,400,70,70); ui->pbt_stop->setGeometry(260,400,70,70); this->imageWidget->setGeometry(QRect(6, 37, 500, 330)); }

打開攝像頭設(shè)備

void deviceOpen(void) { fd = open(deviceName, O_RDWR | O_NONBLOCK, 0); if (-1 == fd) { QMessageBox::about(NULL, "About", "camera open error"); exit(EXIT_FAILURE); } }

初始化攝像頭設(shè)備

void deviceInit(void) { struct v4l2_capability cap; struct v4l2_cropcap cropcap; struct v4l2_crop crop; struct v4l2_format fmt; struct v4l2_streamparm sparm; unsigned int min; if (-1 == xioctl(fd, VIDIOC_QUERYCAP, &cap)) { if (EINVAL == errno) { QMessageBox::about(NULL,"Information"," no V4L2 device"); exit(EXIT_FAILURE); } else { errno_exit("VIDIOC_QUERYCAP"); } } if (!(cap.capabilities & V4L2_CAP_VIDEO_CAPTURE)) { QMessageBox::about(NULL,"Information"," no video capture device");

exit(EXIT_FAILURE); } struct v4l2_input input; input.index = 0; if ( ioctl(fd, VIDIOC_ENUMINPUT, &input) != 0) { QMessageBox::about(NULL,"Information","set input error"); exit(0); } if ((ioctl(fd, VIDIOC_S_INPUT, &input)) < 0) { QMessageBox::about(NULL,"Information","set s_input error"); exit(0); } CLEAR(cropcap); cropcap.type = V4L2_BUF_TYPE_VIDEO_CAPTURE; if (0 == xioctl(fd, VIDIOC_CROPCAP, &cropcap)) { crop.type = V4L2_BUF_TYPE_VIDEO_CAPTURE;

crop.c.top = 0; crop.c.left = 0; crop.c.height = 720; crop.c.width = 1280; if (-1 == xioctl(fd, VIDIOC_S_CROP, &crop)) { switch (errno) { case EINVAL: break; default: break; } } } CLEAR (fmt); // v4l2_format fmt.type = V4L2_BUF_TYPE_VIDEO_CAPTURE; fmt.fmt.pix.width = width; fmt.fmt.pix.height = height; fmt.fmt.pix.pixelformat = V4L2_PIX_FMT_YUYV;

fmt.fmt.pix.field = V4L2_FIELD_ANY; if (-1 == xioctl(fd, VIDIOC_S_FMT, &fmt)) errno_exit("VIDIOC_S_FMT");

/* Note VIDIOC_S_FMT may change width and height.*/ if (width != fmt.fmt.pix.width) { width = fmt.fmt.pix.width; //fprintf(stderr,"Image width set to %i by device %s.\n",width,deviceName); } if (height != fmt.fmt.pix.height) { height = fmt.fmt.pix.height; //fprintf(stderr,"Image height set to %i by device %s.\n",height,deviceName); } /*Buggy driver paranoia. */ min = fmt.fmt.pix.width * 2; if (fmt.fmt.pix.bytesperline < min) fmt.fmt.pix.bytesperline = min; min = fmt.fmt.pix.bytesperline * fmt.fmt.pix.height; if (fmt.fmt.pix.sizeimage < min) fmt.fmt.pix.sizeimage = min; CLEAR (sparm); sparm.type = V4L2_BUF_TYPE_VIDEO_CAPTURE; sparm.parm.capture.capturemode = 0; sparm.parm.capture.timeperframe.numerator = 1; sparm.parm.capture.timeperframe.denominator = 30; if(xioctl(fd,VIDIOC_S_PARM,&sparm) < 0){ errno_exit("cam s parm"); // exit(1); } mmapInit(); }

開啟視頻流捕獲

void captureStart(void) { unsigned int i; enum v4l2_buf_type type; for (i = 0; i < n_buffers; ++i) { struct v4l2_buffer buf; CLEAR (buf); buf.type = V4L2_BUF_TYPE_VIDEO_CAPTURE; buf.memory = V4L2_MEMORY_MMAP; buf.index = i; if (-1 == xioctl(fd, VIDIOC_QBUF, &buf)) errno_exit("VIDIOC_QBUF"); } type = V4L2_BUF_TYPE_VIDEO_CAPTURE; if (-1 == xioctl(fd, VIDIOC_STREAMON, &type)) errno_exit("VIDIOC_STREAMON"); }

超時處理

void Camera::up_date() { unsigned char image_buf[921600+54]; frameRead(image_buf); this->imageWidget->setPixmap(image_buf); }

2、應(yīng)用編譯及測試

1）編譯

elf@ubuntu:~/work/camera-demo$ . /opt/fsl-imx-x11/4.1.15-2.0.0/environment-setup-cortexa7hf-neon-poky-linux-gnueabi elf@ubuntu:~/work/camera-demo$ qmake elf@ubuntu:~/work/camera-demo$ make

2）拷貝camera-demo到ELF 1開發(fā)板的/home/root路徑下，運行測試。

root@ELF1:~# cp /run/media/sda1/camera-demo ./ root@ELF1:~# chmod 777 camera-demo root@ELF1:~# export DISPLAY=:0.0 root@ELF1:~# ./camera-demo

點擊start按鈕之后，使用ls num路徑下查看會有攝像頭拍攝的圖片。液晶屏上會實時預(yù)覽攝像頭拍到的圖像，如下圖所示：

在這里就可以和前面人臉識別結(jié)合起來了，比如攝像頭里面的畫面是一張人臉信息，通過截取攝像頭中的實時畫面到本地，然后上傳到百度智能云進(jìn)行識別，至此就完成了通過攝像頭進(jìn)行人臉識別的過程。

項目測試

在此基礎(chǔ)上再次完善應(yīng)用，識別人臉的應(yīng)用將識別到的人臉信息保存到文本中，基于攝像頭的應(yīng)用讀取文檔中的人臉信息顯示在Qt界面中。

1、確保開發(fā)板已連接USB攝像頭和屏幕

2、設(shè)置Wi-Fi連接

root@ELF1:~# elf1_cmd_wifi.sh -i 8723 -s 賬號 -p 密碼

3、執(zhí)行應(yīng)用

root@ELF1:~# ./camera-demo & root@ELF1:~# ./demoFace

單擊“start”按鈕，識別結(jié)果如下圖所示。