1.按短边筛选

 原始轮廓图：

import cv2
import numpy as np

# 读取轮廓图
contour_image = cv2.imread('..\\IMGS\\pp_edge.png', cv2.IMREAD_GRAYSCALE)

# 使用cv2.findContours()函数获取所有轮廓
contours, _ = cv2.findContours(contour_image, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)

# 设定的直径阈值
threshold_diameter = 25.0

# 遍历每个轮廓
new_image = np.zeros_like(contour_image)
for contour in contours:
    rect = cv2.minAreaRect(contour)  # 获取最小外接矩形
    diameter = rect[1][0] if rect[1][0] <= rect[1][1] else rect[1][1]    # 计算轮廓的最短尺寸，并获取直径

    # 如果轮廓的直径小于设定的阈值，则忽略该轮廓
    if diameter >= threshold_diameter:
        cv2.drawContours(new_image, [contour], -1, (255, 255, 255))
        print(diameter, end=', ')


cv2.imshow('Filtered Contours', new_image)   # 绘制保留的轮廓
cv2.waitKey(0)
cv2.destroyAllWindows()

 输出结果以及与原图对比：

                                     

2.按长边筛选

 

import cv2
import numpy as np

# 读取轮廓图
contour_image = cv2.imread('..\\IMGS\\pp_edge2.png', cv2.IMREAD_GRAYSCALE)

# 使用cv2.findContours()函数获取所有轮廓
contours, _ = cv2.findContours(contour_image, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)

# 设定的直径阈值
threshold_diameter = 25.0

# 遍历每个轮廓
new_image = np.zeros_like(contour_image)
for contour in contours:
    diameter = cv2.minEnclosingCircle(contour)[1] * 2   # cv2.minEnclosingCircle():计算一个轮廓的最小包围圆,返回值是圆心坐标和半径

    # 如果轮廓的直径小于设定的阈值，则忽略该轮廓
    if diameter >= threshold_diameter:
        cv2.drawContours(new_image, [contour], -1, (255, 255, 255))
        print(diameter, end=', ')


cv2.imshow('Filtered Contours', new_image)   # 绘制保留的轮廓
cv2.waitKey(0)
cv2.destroyAllWindows()

结果图：