From 34c46df0a9a20b1ee1c0728d2e99ac9fab101a98 Mon Sep 17 00:00:00 2001
From: xia <shartoo518@gmail.com>
Date: Fri, 11 Oct 2019 14:58:13 +0800
Subject: [PATCH] =?UTF-8?q?=E5=AE=9E=E7=8E=B0=E4=BA=86=E5=9F=BA=E4=BA=8Elb?=
 =?UTF-8?q?p=E7=89=B9=E5=BE=81=E7=9A=84svm=E5=88=86=E7=B1=BB=E6=A8=A1?=
 =?UTF-8?q?=E5=9E=8B?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 lbp_svm.py | 116 +++++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 116 insertions(+)
 create mode 100644 lbp_svm.py

diff --git a/lbp_svm.py b/lbp_svm.py
new file mode 100644
index 0000000..2c44ede
--- /dev/null
+++ b/lbp_svm.py
@@ -0,0 +1,116 @@
+import cv2
+import os
+import numpy as np
+from skimage import feature
+from sklearn.svm import LinearSVC
+from sklearn.metrics import accuracy_score
+import pickle
+from tqdm import tqdm
+
+
+class LocalBinaryPatterns:
+  def __init__(self,num_points,radius):
+    # 存储数据点数目和半径
+    self.numPoints = num_points
+    self.radius = radius
+
+  def describe(self,image,eps=1e-7):
+    # 计算图像的LBP表征，然后使用LBP表征来构建起直方图模式
+    lbp = feature.local_binary_pattern(image,self.numPoints,self.radius,method='uniform')
+    # 直接返回的lbp特征是无法直接使用的，它与输入图像等宽高，取值范围为0到numPoitns+2
+    # 构建直方图可以用来计算 每个LBP模式出现的次数
+    (hist,_) = np.histogram(lbp.ravel(),bins = np.arange(0,self.numPoints+3), range=(0,self.numPoints +2))
+    # 归一化直方图
+    hist = hist.astype("float")
+    hist /= (hist.sum()+eps)
+    return hist
+
+
+def prepare_lbp_data(image_path):
+  desc = LocalBinaryPatterns(24,8)
+  data = []
+  pbar = tqdm(image_path)
+  for image in pbar:
+    pbar.set_description("处理图像 %s" % image)
+    im = cv2.imread(image)
+    #print("图像尺寸是\t",im.shape)
+    gray = cv2.cvtColor(im,cv2.COLOR_BGR2GRAY)
+    gray = cv2.GaussianBlur(gray, (3, 3), 0)
+    hist = desc.describe(gray)
+    data.append(hist)
+  return data
+
+def train_model(data,label):
+  model =LinearSVC(C=100.0, random_state=42)
+  model.fit(data,label)
+  return model
+
+def model_predict(model,test_img):
+  im = cv2.imread(test_img)
+  gray = cv2.cvtColor(im,cv2.COLOR_BGR2GRAY)
+  desc = LocalBinaryPatterns(24, 8)
+  hist = desc.describe(gray)
+  prediction = model.predict(hist.reshape(1,-1))
+  cv2.putText(im,prediction[0],(10,30),cv2.FONT_HERSHEY_SIMPLEX,1.0,(0,0,255),3)
+  cv2.imshow("SVC predict ",im)
+  cv2.waitKey(0)
+
+def get_train_test_data():
+  train_imgs = []
+  train_labels = []
+  test_imgs = []
+  test_labels = []
+  base_dir = "./data/"
+  for subset in os.listdir(base_dir):
+    subpath = os.path.join(base_dir, subset)
+    if subset.startswith("defect") or subset.startswith("normal"):
+      for img in os.listdir(subpath):
+        img_file = os.path.join(subpath,img)
+        train_imgs.append(img_file)
+        if subset.startswith("defect"):
+          train_labels.append(1)
+        else:
+          train_labels.append(0)
+    elif subset.startswith("test"):
+      test_paths = os.listdir(subpath)
+      for path in test_paths:
+        new_path = os.path.join(subpath,path)
+        if os.path.isdir(new_path):
+          for img in os.listdir(new_path):
+            img_file = os.path.join(new_path,img)
+            test_imgs.append(img_file)
+            if path =="defect":
+              test_labels.append(1)
+            else:
+              test_labels.append(0)
+  return train_imgs,np.array(train_labels),test_imgs,np.array(test_labels)
+
+if __name__ =="__main__":
+  train_imgs, train_labels, test_imgs, test_labels = get_train_test_data()
+  indexs = [x for x in range(0,len(train_imgs))]
+  np.random.shuffle(indexs)
+  print("获取的图片概览..")
+  print("训练集前5 ",train_imgs[:5])
+  print("训练集标签前5 ",train_labels[:5])
+  # 数据混排
+  real_train_imgs = []
+  real_train_labels =[]
+  for i in indexs:
+      real_train_imgs.append(train_imgs[i])
+      real_train_labels.append(train_labels[i])
+  print("训练集 图片总数 : %d，标签总数: %d ,测试集 图片总数 : %d，标签总数 %d  " % (
+  len(train_imgs), len(train_labels), len(test_imgs), len(test_labels)))
+
+  train_data = prepare_lbp_data(real_train_imgs)
+  model = train_model(train_data,real_train_labels)
+  save_file = "defect_normal_svm_875.bin"
+  out = open(save_file,"wb")
+  model_save = pickle.dumps(model)
+  out.write(model_save)
+  out.close()
+  input = open(save_file, 'rb')
+  real_model = input.read()
+  model = pickle.loads(real_model)
+  test_data = prepare_lbp_data(test_imgs)
+  y_pred = model.predict(test_data)
+  print("模型测试准确率\t",accuracy_score(test_labels,y_pred))
\ No newline at end of file