将显著性检测图像后处理，绘出一个矩形框，方便使用
先通过边缘检测，然后通过opencv接口实现
结果如下：
# --coding:utf
# -8-*-
import osimport cv2 as cv
import numpy as np# canny边缘检测
def canny_demo(image):t = 50canny_output = cv.Canny(image, t, 95)return canny_outputdef image_target_detection(img, mask):# 读取图像src = cv.imread(mask)img = cv.imread(img)binary = canny_demo(src)k = np.ones((5, 5), dtype=np.uint8)binary = cv.morphologyEx(binary, cv.MORPH_DILATE, k)# kernel = cv.getStructuringElement(cv.MORPH_RECT, (5, 5))# binary = cv.morphologyEx(binary, cv.MORPH_OPEN, kernel)cv.namedWindow("input", cv.WINDOW_NORMAL | cv.WINDOW_KEEPRATIO)cv.resizeWindow("input", 400, 300)cv.imshow("input", binary)image, contours, hierarchy = cv.findContours(binary, cv.RETR_EXTERNAL, cv.CHAIN_APPROX_SIMPLE)print("contours", contours)print("contours", len(contours))print("contours", type(contours))for c in range(len(contours)):x, y, w, h = cv.boundingRect(contours[c])cv.rectangle(img, (x, y), (x + w, y + h), (0, 0, 255), 10)rect = cv.minAreaRect(contours[c]) # 最小旋转矩形cx, cy = rect[0]box = cv.boxPoints(rect) # 点集的最小外接矩形box = np.int0(box)cv.drawContours(src, [box], 0, (0, 0, 255), 10)cv.drawContours(img, [box], 0, (0, 0, 255), 10)cv.circle(src, (np.int32(cx), np.int32(cy)), 2, (255, 0, 0), 2, 8, 0)cv.drawContours(src, contours, c, (0, 0, 255), 2, 8)# 图像显示cv.namedWindow("contours_analysis", 0)cv.resizeWindow("contours_analysis", 400, 300)cv.imshow("contours_analysis", img)# cv.namedWindow("out", 0)# cv.resizeWindow("out", 400, 300)# cv.imshow("out", src)# cv.imwrite("D:/vsprojects/images/contours_analysis.png", src)cv.waitKey(0)cv.destroyAllWindows()if __name__ == '__main__':imgdir = 'E:\\data\\'maskdir = 'E:\\data\\'img_list = [imgdir + f for f in os.listdir(imgdir) if f.endswith("jpg")]for img in img_list:mask = maskdir + img.split(os.sep)[-1].split('.')[0] + '.png'image_target_detection(img, mask)
