传统目标检测方法研究（二）

1.3 特征分类

特征分类：最后，对上一步得到的特征进行分类，通常使用如 SVM、AdaBoost的分类器。

1.3.1 支持向量机

支持向量机参考视频

1.3.2 python-opencv实现支持向量机

对线性可分数据分类

from re import S
from sklearn import datasets
from sklearn import metrics
import matplotlib.pyplot as plt
import numpy as np
from sklearn import model_selection as ms
import cv2
# 数据生成
X,y = datasets.make_classification(n_samples=100,n_features=2,
                                    n_redundant=0,n_classes=2,random_state=7816)
# 数据可视化
plt.scatter(X[:,0],X[:,1],c=y,s=100)
plt.xlabel("x values")
plt.ylabel("y values")
plt.savefig("data.png")
plt.show()
# 数据集预处理
X = X.astype(np.float32)
y = y*2 -1 # 标签 0 1 转换为 -1 1
X_train, X_test, y_train, y_test = ms.train_test_split(X,y,test_size=0.2,random_state=42)
# 创建支持向量机
svm = cv2.ml.SVM_create()
svm.setKernel(cv2.ml.SVM_LINEAR) # 使用线性SVM
svm.train(X_train,cv2.ml.ROW_SAMPLE,y_train)
# 预测目标标签
_,y_pred = svm.predict(X_test)
# 分类器打分
acc = metrics.accuracy_score(y_test,y_pred)
print(acc)
# 决策边界可视化
def plot_decision_boundary(svm,X_test,y_test):
    x_min, x_max = X_test[:,0].min() - 1, X_test[:,0].max() + 1
    y_min, y_max = X_test[:,1].min() - 1, X_test[:,1].max() + 1
    h = 0.02
    xx, yy = np.meshgrid(np.arange(x_min,x_max,h),
                        np.arange(y_min,y_max,h))
    X_hypo = np.c_[xx.ravel().astype(np.float32),
                    yy.ravel().astype(np.float32)]
    _,zz = svm.predict(X_hypo)
    zz = zz.reshape(xx.shape)
    plt.contourf(xx,yy,zz,cmap=plt.cm.coolwarm,alpha=0.8)
    plt.scatter(X_test[:,0],X_test[:,1],c=y_test,s=200)
    plt.savefig("vis_decision_boundary.png")
    plt.show()

plot_decision_boundary(svm,X_test,y_test)

数据可视化图像

决策边界可视化图像

采用不同高斯核可以实现非线性支持向量机

# 决策边界可视化
def plot_decision_boundary(svm,X_test,y_test):
    x_min, x_max = X_test[:,0].min() - 1, X_test[:,0].max() + 1
    y_min, y_max = X_test[:,1].min() - 1, X_test[:,1].max() + 1
    h = 0.02
    xx, yy = np.meshgrid(np.arange(x_min,x_max,h),
                        np.arange(y_min,y_max,h))
    X_hypo = np.c_[xx.ravel().astype(np.float32),
                    yy.ravel().astype(np.float32)]
    _,zz = svm.predict(X_hypo)
    zz = zz.reshape(xx.shape)
    plt.contourf(xx,yy,zz,cmap=plt.cm.coolwarm,alpha=0.8)
    plt.scatter(X_test[:,0],X_test[:,1],c=y_test,s=200)
if __name__ == "__main__":
    from re import S
	from sklearn import datasets
	from sklearn import metrics
	import matplotlib.pyplot as plt
	import numpy as np
	from sklearn import model_selection as ms
	import cv2
    # 数据生成
    X,y = datasets.make_classification(n_samples=100,n_features=2,
                                        n_redundant=0,n_classes=2,random_state=7816)
    # 数据可视化
    plt.scatter(X[:,0],X[:,1],c=y,s=100)
    plt.xlabel("x values")
    plt.ylabel("y values")
    plt.savefig("data.png")
    plt.show()
    # 数据集预处理
    X = X.astype(np.float32)
    y = y*2 -1 # 标签 0 1 转换为 -1 1
    X_train, X_test, y_train, y_test = ms.train_test_split(X,y,test_size=0.2,random_state=42)
    # 创建支持向量机
    kernels = [cv2.ml.SVM_LINEAR,cv2.ml.SVM_INTER,cv2.ml.SVM_SIGMOID,cv2.ml.SVM_RBF] 
    for idx, kernel in enumerate(kernels):
        svm = cv2.ml.SVM_create()
        svm.setKernel(kernel) # 使用线性SVM
        svm.train(X_train,cv2.ml.ROW_SAMPLE,y_train)
        # 预测目标标签
        _,y_pred = svm.predict(X_test)
        # 分类器打分
        acc = metrics.accuracy_score(y_test,y_pred)
        print(acc)
        plt.subplot(2,2,idx+1)
        plot_decision_boundary(svm,X_test,y_test)
        plt.title("accuracy=%.2f"%acc)
    plt.savefig("non-linear.png")
    plt.show()