一、安装和配置opencv
sudo apt-get install build-essential checkinstall cmake pkg-config yasm
sudo apt-get install git gfortran
sudo apt-get install libjpeg8-dev libjasper-dev libpng12-dev
# If you are using Ubuntu 14.04
sudo apt-get install libtiff4-dev
# If you are using Ubuntu 16.04
sudo apt-get install libtiff5-dev
sudo apt-get install libavcodec-dev libavformat-dev libswscale-dev libdc1394-22-dev
sudo apt-get install libxine2-dev libv4l-dev
sudo apt-get install libgstreamer0.10-dev libgstreamer-plugins-base0.10-dev
sudo apt-get install qt5-default libgtk2.0-dev libtbb-dev
sudo apt-get install libatlas-base-dev
sudo apt-get install libfaac-dev libmp3lame-dev libtheora-dev
sudo apt-get install libvorbis-dev libxvidcore-dev
sudo apt-get install libopencore-amrnb-dev libopencore-amrwb-dev
sudo apt-get install x264 v4l-utils
# Optional dependencies
sudo apt-get install libprotobuf-dev protobuf-compiler
sudo apt-get install libgoogle-glog-dev libgflags-dev
sudo apt-get install libgphoto2-dev libeigen3-dev libhdf5-dev doxygen
下载解压opencv opencv-contrib
安装python3库(可选)
sudo apt-get install python-dev python-pip python3-dev python3-pip
sudo -H pip2 install -U pip numpy
sudo -H pip3 install -U pip numpy
# Install virtual environment
sudo pip2 install virtualenv virtualenvwrapper
sudo pip3 install virtualenv virtualenvwrapper
echo "# Virtual Environment Wrapper" >> ~/.bashrc
echo "source /usr/local/bin/virtualenvwrapper.sh" >> ~/.bashrc
source ~/.bashrc
pip install numpy scipy matplotlib scikit-image scikit-learn ipython
编译和安装
cd opencv
mkdir build
cd build
cmake -D CMAKE_BUILD_TYPE=RELEASE \
-D CMAKE_INSTALL_PREFIX=/usr/local \
-D INSTALL_C_EXAMPLES=ON \
-D INSTALL_PYTHON_EXAMPLES=ON \
-D WITH_TBB=ON \
-D WITH_V4L=ON \
-D WITH_QT=ON \
-D WITH_OPENGL=ON \
-D OPENCV_EXTRA_MODULES_PATH=../../opencv_contrib/modules \
-D BUILD_EXAMPLES=ON ..
# find out number of CPU cores in your machine
nproc
# substitute 4 by output of nproc
make -j4
sudo make install
sudo sh -c 'echo "/usr/local/lib" >> /etc/ld.so.conf.d/opencv.conf'
sudo ldconfig
二、收集手机图片和非手机图片,形成正样本和负样本
可以通过百度图片,爬虫等方式获取自己需要的图片
三、标记正样本中手机所在的位置
执行命令:python3 object_marker.py /home/zhou/Desktop/github_summary/haar_cascade/result.txt /home/zhou/Desktop/github_summary/haar_cascade/sample/pos_color/
object_marker.py代码如下:
#!/usr/bin/python
###############################################################################
# Name : ObjectMarker.py
# Author : Python implementation: sqshemet
# Original ObjectMarker.cpp: http://www.cs.utah.edu/~turcsans/DUC_files/HaarTraining/
# Date : 7/24/12
# Description : Object marker utility to be used with OpenCV Haar training.
# Tested on Ubuntu Linux 10.04 with OpenCV 2.1.0.
# Usage : python ObjectMarker.py outputfile inputdirectory
###############################################################################
import cv2
import sys
import os
import numpy as np
#shape = (1000,1000,3)
#image = np.zeros(shape, np.uint8)
#image2 = np.zeros(shape, np.uint8)
ratio=1
roi_x0 = 0
roi_y0 = 0
roi_x1 = 0
roi_y1 = 0
num_of_rec = 0
start_draw = False
window_name = "<Space> to save and load next, <X> to skip, <ESC> to exit."
def on_mouse(event, x, y, flag, params):
global start_draw
global roi_x0
global roi_y0
global roi_x1
global roi_y1
if (event == cv2.EVENT_LBUTTONDOWN):
if (not start_draw):
roi_x0 = x
roi_y0 = y
start_draw = True
else:
roi_x1 = x
roi_y1 = y
print("%d,%d" %(x, y))
start_draw = False
elif (event == cv2.EVENT_MOUSEMOVE and start_draw):
#Redraw ROI selection
image2 = image.copy()
#print("%d,%d" %(x, y))
#print(image2.shape)
# x is ratio
#cv2.rectangle(image2, (roi_x0, roi_y0), (x,roi_y0+int(ratio*(x-roi_x0))),
#(255,0,255),2)
cv2.rectangle(image2, (roi_x0, roi_y0), (x,y),
(255),5)
cv2.imshow(window_name, image2)
def main():
# Might want to divide this a bit more.
global image
iKey = 0
if (len(sys.argv) != 3):
sys.stderr.write("%s output_info.txt raw/data/directory\n"
% sys.argv[0])
return -1
input_directory = sys.argv[2]
output_file = sys.argv[1]
#Get a file listing of all files within the input directory
try:
files = os.listdir(input_directory)
except OSError:
sys.stderr.write("Failed to open dirctory %s\n"
% input_directory)
return -1
files.sort()
sys.stderr.write("ObjectMarker: Input Directory: %s Output File %s\n"
% (input_directory, output_file))
# init GUI
cv2.namedWindow(window_name, 0)
cv2.setMouseCallback(window_name, on_mouse, None)
sys.stderr.write("Opening directory...")
# init output of rectangles to the info file
#os.chdir(input_directory)
sys.stderr.write("done.\n")
str_prefix = input_directory
try:
output = open(output_file,'a+')
except IOError:
sys.stderr.write("Failed to open file %s.\n" % output_file)
return -1
for file in files:
str_postfix = ""
num_of_rec = 0
img = str_prefix + file
sys.stderr.write("Loading image %s...\n" % img)
try:
image = cv2.imread(img,cv2.IMREAD_UNCHANGED)
except IOError:
sys.stderr.write("Failed to load image %s.\n" % img)
return -1
# Work on current image
cv2.imshow(window_name, image)
# Need to figure out waitkey returns.
# <ESC> = 27 exit program
# <Space> = 32 add rectangle to current image
# <x> = 120 skip image keycode
iKey = cv2.waitKey(0) % 255
# This is ugly, but is actually a simplification of the C++.
if iKey == 27:
cv2.destroyWindow(window_name)
return 0
elif iKey == 32:
print(1)
num_of_rec += 1
# Currently two draw directions possible:
# from top left to bottom right or vice versa
if (roi_x0<roi_x1 and roi_y0<roi_y1):
str_postfix += " %d %d %d %d\n" % (roi_x0,
roi_y0, (roi_x1-roi_x0), (roi_y1-roi_y0))
elif (roi_x0>roi_x1 and roi_y0>roi_y1):
str_postfix += " %d %d %d %d\n" % (roi_x1,
roi_y1, (roi_x0-roi_x1), (roi_y0-roi_y1))
output.write(img + " " + str(num_of_rec) + str_postfix)
elif iKey == 120:
sys.stderr.write("Skipped %s.\n" % img)
if __name__ == '__main__':
main()
四、训练haar cascade分类器
1.执行命令opencv_createsamples -info result.txt -vec positive.vec,生存vec样本文件
2.执行训练过程opencv_traincascade -data data -vec positive.vec -bg bg.txt -numStages 10 -numPos 50 -numNeg 100
五、验证haar cascade分类效果
import cv2
import numpy as np
mobile_cascade = cv2.CascadeClassifier('cascade.xml')
cap=cv2.VideoCapture(3)
i=0
while True:
ret,img=cap.read()
gray=cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
faces=mobile_cascade.detectMultiScale(gray,1.3,5)
i+=1
for (x,y,w,h) in faces:
cv2.rectangle(img,(x,y),(x+w,y+h),(255,0,0),2)
cv2.imshow('img',img)
k=cv2.waitKey(10) & 0xFF
if i==30:
break
cap.release()
cv2.destroyWindow('img')