使用 OpenCV 霍夫變換-圓檢測(cè)，對(duì)周圍背景比較敏感，容易誤識(shí)別，不受控。若你也有此困惑，建議試試本文中的方法，識(shí)別效果佳，能夠很好地排除類圓矩形的干擾，話不多說直接上代碼。

代碼

一、實(shí)現(xiàn)類

import math

import cv2


class CircleDetector(object):

    '''
	Parameters
	----------
	img: ndarray
		A color image.
	threshold: int or float
        Image binary threshold.
	minRadius: int or float
		Minimum value of circle radius.
	maxRadius: int or flaot
		Maximum value of circle radius.

	Returns
	-------
	A tuple of (center(x, y), size(w, h), angle)
	'''
    def detectCircles(self, image, threshold, minRadius, maxRadius):
        circles = list()
        gray_image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
        blur_image = cv2.GaussianBlur(gray_image, (5, 5), 0)
        ret, thresh = cv2.threshold(gray_image, threshold, 255, cv2.THRESH_BINARY)
        kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (5, 5), (-1, -1))
        thresh = cv2.morphologyEx(thresh, cv2.MORPH_CLOSE, kernel, (-1, -1))
        thresh = cv2.morphologyEx(thresh, cv2.MORPH_OPEN, kernel, (-1, -1))

        contours, hierarchy = cv2.findContours(thresh, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)

        for cnt in contours:
            if len(cnt) < 5:
                continue

            area = cv2.contourArea(cnt)
            if area < (minRadius**2) * math.pi or area > (maxRadius**2) * math.pi:
                continue

            arc_length = cv2.arcLength(cnt, True)
            radius = arc_length / (2 * math.pi)

            if not (minRadius < radius and radius < maxRadius):
                continue

            ellipse = cv2.fitEllipse(cnt)
            ratio = float(ellipse[1][0]) / float(ellipse[1][1])

            if ratio > 0.9 and ratio < 1.1:
                corner = cv2.approxPolyDP(cnt, 0.02 * arc_length, True)
                cornerNum = len(corner)
                if cornerNum > 4: # 當(dāng)cornerNum=4時(shí)，識(shí)別矩形；而cornerNum>4時(shí)，識(shí)別圓
                    circles.append(ellipse)

        return circles

二、使用

import cv2

from detector.circle_detector import CircleDetector

if __name__ == '__main__':
    src = 0
    cap = cv2.VideoCapture(src)

    detector = CircleDetector()
    
    while True:
        if not cap.isOpened():
            print('相機(jī)未打開')
            break

        ret, frame = cap.read()

        if not ret:
            continue

        circles = detector.detectCircles(frame, 158, 50, 200)
        img = frame.copy()

        for circle in circles:
            cv2.circle(img, (int(circle[0][0]), int(circle[0][1])), int(20), (0, 255, 0), thickness=5)

        cv2.imshow('image', img)
        key = cv2.waitKey(int(1000/30)) & 0xFF
        if key == ord(' '):
            break

三、注意事項(xiàng)

如果遇到無法識(shí)別，或者誤識(shí)別，注意調(diào)整參數(shù)。

1. 確保你圖中所繪圓的半徑在[minRadius, maxRadius]之間。
2. 建議使用白背景黑圓作為識(shí)別目標(biāo)，否之請(qǐng)自行調(diào)整Threshold，直到識(shí)別效果達(dá)到最佳狀態(tài)(可以查看二值化圖片進(jìn)行分析)。

效果圖

文章來源地址http://www.zghlxwxcb.cn/news/detail-620902.html

到了這里，關(guān)于【OpenCV】高精度識(shí)別圓（支持復(fù)雜場(chǎng)景下的圓）的文章就介紹完了。如果您還想了解更多內(nèi)容，請(qǐng)?jiān)谟疑辖撬阉鱐OY模板網(wǎng)以前的文章或繼續(xù)瀏覽下面的相關(guān)文章，希望大家以后多多支持TOY模板網(wǎng)！