最近领导让我做图片识别，把这两天的工作记录一下吧,虽然中间做的磕磕碰碰，但是一个好的开始，加油！好了不灌鸡汤了，let's show！
在做图片识别之前，需要对图片做处理，利用的是opencv（python 环境需要装）
比如我们要识别的电表的数字
下面是对该图片的做opencv处理,源代码如下：
# coding=utf-8
from __future__ import division #整数相除为浮点数
import cv2
import numpy as np
import os
img = cv2.imread('testset/img4.PNG')
#cv2.imshow('Original', img)
cv2.waitKey(0)
#cv2.imwrite('save/img4.PNG',img)
# 灰度处理
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
#cv2.imshow('Gray', gray)
cv2.waitKey(0)
#cv2.imwrite('save/gray.PNG',gray)
# 均值滤波
# median = cv2.medianBlur(gray, 3)
blur = cv2.blur(img, (4, 4))
#cv2.imshow('Blur', blur)
cv2.waitKey(0)
#cv2.imwrite('save/blur.PNG',blur)
# Canny边缘提取
canny = cv2.Canny(blur, 300, 450)
#cv2.imshow('Canny', canny)
cv2.waitKey(0)
#cv2.imwrite('save/canny.PNG',canny)
# 二值处理
#ret, thresh = cv2.threshold(canny, 90, 255, cv2.THRESH_BINARY)
#kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (3, 3))
#closed = cv2.morphologyEx(thresh, cv2.MORPH_CLOSE, kernel)
# 膨胀操作
kernel = np.uint8(np.ones((7, 7)))
dilate = cv2.dilate(canny, kernel)
# 腐蚀操作
erode = cv2.erode(dilate,(9,9))
#cv2.imshow('Dilate', erode)
cv2.waitKey(0)
#cv2.imwrite('save/dilate.PNG',dilate)
(image, cnts, _) = cv2.findContours(dilate.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
for index, c in enumerate(cnts):
rect = cv2.minAreaRect(c)
box = np.int0(cv2.boxPoints(rect))
# draw a bounding box arounded the detected number and display the image
cv2.drawContours(img, [box], -1, (0, 255, 0), 0)
Xs = [i[0] for i in box]
Ys = [i[1] for i in box]
x1 = min(Xs)
x2 = max(Xs)
y1 = min(Ys)
y2 = max(Ys)
hight = y2 - y1
width = x2 - x1
cropImg = image[y1:y1+hight, x1:x1+width]
cv2.imshow(str(i + 1), cropImg)
###### 按顺序保存图片
for j in i:
cv2.imwrite('save/%d.PNG' % i[0], cropImg)
######
cv2.waitKey(0)
#cv2.imshow('Image', img)
cv2.waitKey(0)
#cv2.imwrite('save/img.PNG',img)
#图像统一预处理成28*28
imgs=os.listdir('save')
num = len(imgs)
for index,i in enumerate(imgs):
img=cv2.imread('save/'+i,0)
#print img.shape
width=img.shape[1]
height=img.shape[0]
fx=28/width
fy=28/height
res = cv2.resize(img, None, fx=fx, fy=fy, interpolation=cv2.INTER_CUBIC) #图像缩放成28x28
cv2.imwrite('save/%d.png' % (index), res)
处理后的结果如下：需要说明一下，对图片数字的小数点，我们还没有做处理，在此先搁浅，以后写出来，后补！
下面就是我们的重头戏了，利用的是两层cnn做训练并识别图片，训练的模型是mnist的demo,在这里我们是保存了该训练的模型，talk is cheap ,show you my code!
import tensorflow as tf
import tensorflow.examples.tutorials.mnist.input_data as input_data
import os
MODEL_SAVE_PATH="model_data/"
MODEL_NAME="save_net.ckpt"
def weight_variable(shape):
initial=tf.truncated_normal(shape,stddev=0.1)
return tf.Variable(initial)
def bias_variable(shape):
initial=tf.constant(0.1,shape=shape)
return tf.Variable(initial)
def conv2d(x,W):
return tf.nn.conv2d(x,W,strides=[1,1,1,1],padding="SAME")
def max_pool_2x2(x):
return tf.nn.max_pool(x,ksize=[1,2,2,1],strides=[1,2,2,1],padding='SAME')
with tf.Session() as sess:
mnist = input_data.read_data_sets("MNIST_data", one_hot=True)
x = tf.placeholder(tf.float32, [None, 784])
w_conv1=weight_variable([5,5,1,32])
b_conv1=bias_variable([32])
x_image=tf.reshape(x,[-1,28,28,1])
y_ = tf.placeholder("float", [None, 10])
h_conv1=tf.nn.relu(conv2d(x_image,w_conv1)+b_conv1)
h_pool1=max_pool_2x2(h_conv1)
w_conv2=weight_variable([5,5,32,64])
b_conv2=bias_variable([64])
h_conv2=tf.nn.relu(conv2d(h_pool1,w_conv2)+b_conv2)
h_pool2=max_pool_2x2(h_conv2)
w_fc1=weight_variable([7*7*64,1024])
b_fc1=bias_variable([1024])
h_pool2_flat=tf.reshape(h_pool2,[-1,7*7*64])
h_fc1=tf.nn.relu(tf.matmul(h_pool2_flat,w_fc1)+b_fc1)
keep_prob=tf.placeholder("float")
h_fc1_drop=tf.nn.dropout(h_fc1,keep_prob)
w_fc2=weight_variable([1024,10])
b_fc2=bias_variable([10])
y_conv=tf.nn.softmax(tf.matmul(h_fc1_drop,w_fc2)+b_fc2)
cross_entropy=-tf.reduce_sum(y_*tf.log(y_conv))
train_step=tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
saver = tf.train.Saver()
correct_prediction=tf.equal(tf.argmax(y_conv,1),tf.argmax(y_,1))
accuracy=tf.reduce_mean(tf.cast(correct_prediction,"float"))
sess.run(tf.global_variables_initializer())
for i in range(2000):
batch=mnist.train.next_batch(50)
if i%100==0:
train_accuracy=accuracy.eval(feed_dict={x:batch[0],y_:batch[1],keep_prob:1.0})
print("step %d,training accuracy %g" % (i,train_accuracy))
train_step.run(feed_dict={x:batch[0],y_:batch[1],keep_prob:0.5})
print("test accuracy %g" % accuracy.eval(feed_dict={x:mnist.test.images,y_:mnist.test.labels,keep_prob:1.0}))
saver.save(sess, os.path.join(MODEL_SAVE_PATH, MODEL_NAME), write_meta_graph=False)
接下来就是利用训练的模型来做识别了，plz see
# coding:utf-8
import tensorflow as tf
import numpy as np
import cv2
#初始化单个卷积核上的参数
def weight_variable(shape):
initial = tf.truncated_normal(shape, stddev=0.1)
return tf.Variable(initial)
#初始化单个卷积核上的偏置值
def bias_variable(shape):
initial = tf.constant(0.1, shape=shape)
return tf.Variable(initial)
#输入特征x，用卷积核W进行卷积运算，strides为卷积核移动步长，
#padding表示是否需要补齐边缘像素使输出图像大小不变
def conv2d(x, W):
return tf.nn.conv2d(x, W, strides=[1, 1, 1, 1], padding='SAME')
#对x进行最大池化操作，ksize进行池化的范围，
def max_pool_2x2(x):
return tf.nn.max_pool(x, ksize=[1, 2, 2, 1],strides=[1, 2, 2, 1], padding='SAME')
#
# 定义会话
with tf.Session() as sess:
#声明输入图片数据，类别
x = tf.placeholder(tf.float32,[None,784])
x_img = tf.reshape(x , [-1,28,28,1])
W_conv1 = weight_variable([5, 5, 1, 32])
b_conv1 = bias_variable([32])
#进行卷积操作，并添加relu激活函数
h_conv1 = tf.nn.relu(conv2d(x_img,W_conv1) + b_conv1)
#进行最大池化
h_pool1 = max_pool_2x2(h_conv1)
W_conv2 = weight_variable([5,5,32,64])
b_conv2 = bias_variable([64])
# 同理第二层卷积层
h_conv2 = tf.nn.relu(conv2d(h_pool1, W_conv2) + b_conv2)
h_pool2 = max_pool_2x2(h_conv2)
W_fc1 = weight_variable([7*7*64,1024])
b_fc1 = bias_variable([1024])
#将卷积的产出展开
h_pool2_flat = tf.reshape(h_pool2,[-1,7*7*64])
#神经网络计算，并添加relu激活函数
h_fc1 = tf.nn.relu(tf.matmul(h_pool2_flat,W_fc1) + b_fc1)
keep_prob = tf.placeholder(tf.float32)
h_fc1_drop = tf.nn.dropout(h_fc1, keep_prob)
W_fc2 = weight_variable([1024,10])
b_fc2 = bias_variable([10])
# 引用mnist训练好的保存的模型
saver = tf.train.Saver(write_version=tf.train.SaverDef.V1)
saver.restore(sess, 'model_data/save_net.ckpt')
#输出层，使用softmax进行多分类
y_conv=tf.nn.softmax(tf.matmul(h_fc1, W_fc2) + b_fc2)
im = cv2.imread('save/img4_4.png', cv2.IMREAD_GRAYSCALE)
im = cv2.resize(im, (28, 28), interpolation=cv2.INTER_CUBIC)
img = cv2.GaussianBlur(im, (3, 3), 0)
# 图片预处理
# 数据从0~255转为-0.5~0.5
img_gray = (im - (255 / 2.0)) / 255
# img_gray = (im)/255
# for i in range(28):
# for j in range(28):
# if img_gray[i][j]<=0.5:
# img_gray[i][j]=0
# else:
# img_gray[i][j]=1
cv2.imshow('out',img_gray)
cv2.waitKey(0)
x_img = np.reshape(img_gray, [-1, 784])
output = sess.run(y_conv , feed_dict = {x:x_img})
print('the y_con : ', '\n',output)
print('the predict is : ', np.argmax(output))
结果如下：
这里的数字识别大致过程差不多就这样，虽然表面看起来很完美，但是还有些数字没有识别正确，我举的例子数字是都识别出来了，但是其他的数字还有点问题，这里在随后我解决了，再做补充吧。
