OpenCV：如何使用cv2.CascadeClassifier获得功能？

Question

I know that haarcascade.xml describes some features of some object we want to detect. Eg face has a nose, eyes and it is ellipse or auto-plates has a line of digits and letters with specific size and proportion.

If we use:

cascade = cv2.CascadeClassifier('haarcascade.xml')
objects = cascade.detectMultiScale(gray, 1.1, 3)

we will get a list of box parameters (x,y,w,h)

Is there any way to get all features parameters?

eg if haarcascade describes face I want to get nose coordinates (without using additional haarcascade) or if haarcascade describes auto-plates I want to get coordinates of each symbol.

Answer 1

Have a look at this paper Rapid Object Detection using a Boosted Cascade of Simple Features . It explains the implementation of the cascade. As you can see they use Haar Features , so you won't be able to extract that information. But there is a special cascade for detecting eyes . Maybe this helps you. You can find all other pretrained cascades here . If you want to extract the nose position you may need to train your own cascade or search for an .xml file for that.