OpenCV4入门124:HOG特征描述子—使用HOG进行对象检测

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
import cv2 as cv
import numpy as np

if __name__ == '__main__':
image = cv.imread("../images/elec_watch/test/scene_08.jpg")
test_img = cv.resize(image, (0, 0), fx=0.2, fy=0.2)
cv.imshow("input", test_img)
gray = cv.cvtColor(test_img, cv.COLOR_BGR2GRAY)
print(test_img.shape)
h, w = test_img.shape[:2]
svm = cv.ml.SVM_load('svm_data.dat')
sum_x = 0
sum_y = 0
count = 0
hog = cv.HOGDescriptor()
for row in range(64, h-64, 4):
for col in range(32, w-32, 4):
win_roi = gray[row-64:row+64,col-32:col+32]
hog_desc = hog.compute(win_roi, winStride=(8, 8), padding=(0, 0))
one_fv = np.zeros([len(hog_desc)], dtype=np.float32)
for i in range(len(hog_desc)):
one_fv[i] = hog_desc[i][0]
one_fv = np.reshape(one_fv, [-1, len(hog_desc)])
result = svm.predict(one_fv)[1]
if result[0][0] > 0:
sum_x += (col-32)
sum_y += (row-64)
count += 1
# cv.rectangle(test_img, (col-32, row-64), (col+32, row+64), (255, 0, 0), 2, 8, 0)
x = sum_x // count
y = sum_y // count
cv.rectangle(test_img, (x, y), (x+64, y+128), (0, 0, 255), 2, 8, 0)
cv.imshow("result", test_img)
cv.waitKey(0)
cv.destroyAllWindows()

对于已经训练好的HOG+SVM的模型,我们可以通过开窗实现对象检测,从而完成自定义对象检测。以电表检测为例,这样我们就实现HOG+SVM对象检测全流程。OpenCV中实现对每个窗口像素块预测,需要首先加载SVM模型文件,然后使用predict方法实现预测。这种方法的缺点就是开窗检测是从左到右、从上到下,是一个高耗时的操作,所以步长选择一般会选择HOG窗口默认步长的一半,这样可以减少检测框的数目,同时在predict时候会发现多个重复框,求取它们的平均值即可得到最终的检测框。