如何用OpenCV检测像素RPG游戏截图中的怪物实体

1. HSV颜色范围调试技巧 #

• 别硬写固定值，用交互式工具快速锁定范围：

  import cv2
  import numpy as np
  
  def nothing(x):
      pass
  
  cv2.namedWindow("HSV Tuner")
  cv2.createTrackbar("H Lower", "HSV Tuner", 0, 179, nothing)
  cv2.createTrackbar("H Upper", "HSV Tuner", 179, 179, nothing)
  cv2.createTrackbar("S Lower", "HSV Tuner", 0, 255, nothing)
  cv2.createTrackbar("S Upper", "HSV Tuner", 255, 255, nothing)
  cv2.createTrackbar("V Lower", "HSV Tuner", 0, 255, nothing)
  cv2.createTrackbar("V Upper", "HSV Tuner", 255, 255, nothing)
  
  img = cv2.imread("你的游戏截图路径")
  hsv = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
  
  while True:
      h_low = cv2.getTrackbarPos("H Lower", "HSV Tuner")
      h_high = cv2.getTrackbarPos("H Upper", "HSV Tuner")
      s_low = cv2.getTrackbarPos("S Lower", "HSV Tuner")
      s_high = cv2.getTrackbarPos("S Upper", "HSV Tuner")
      v_low = cv2.getTrackbarPos("V Lower", "HSV Tuner")
      v_high = cv2.getTrackbarPos("V Upper", "HSV Tuner")
  
      lower = np.array([h_low, s_low, v_low])
      upper = np.array([h_high, s_high, v_high])
      mask = cv2.inRange(hsv, lower, upper)
      res = cv2.bitwise_and(img, img, mask=mask)
  
      cv2.imshow("Mask", mask)
      cv2.imshow("Result", res)
      if cv2.waitKey(1) & 0xFF == ord('q'):
          break
  
  cv2.destroyAllWindows()
  print(f"最终HSV范围：lower = [{h_low}, {s_low}, {v_low}], upper = [{h_high}, {s_high}, {v_high}]")
  

• 黄红色怪物的H通道覆盖低区间（0-30）和高区间（150-179），可分两个范围生成掩码后合并，避免漏检。

2. 轮廓过滤参数调整 #

• 先统计怪物实际像素尺寸：打开标记好的截图，用画图工具量取怪物宽高，确定最小/最大面积、宽高比范围。
• 轮廓过滤代码示例：

  contours, _ = cv2.findContours(mask, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
  for cnt in contours:
      area = cv2.contourArea(cnt)
      # 替换为你统计的最小/最大面积阈值
      if area < 50 or area > 500:
          continue
      x, y, w, h = cv2.boundingRect(cnt)
      aspect_ratio = w / h
      # 替换为怪物的宽高比范围，比如0.8-1.2对应接近正方形的怪物
      if aspect_ratio < 0.8 or aspect_ratio > 1.2:
          continue
      # 绘制检测框
      cv2.rectangle(img, (x, y), (x+w, y+h), (0,255,0), 2)
  

• 用cv2.RETR_EXTERNAL只保留外部轮廓，过滤内部小轮廓干扰。

3. 图像预处理优化 #

• 先做高斯模糊减少像素噪点：

  img_blur = cv2.GaussianBlur(img, (3,3), 0)
  hsv = cv2.cvtColor(img_blur, cv2.COLOR_BGR2HSV)
  

• 若背景灰黑色有杂色，可先二值化再合并掩码：

  gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
  _, bg_mask = cv2.threshold(gray, 50, 255, cv2.THRESH_BINARY)
  final_mask = cv2.bitwise_and(mask, bg_mask)
  

内容的提问来源于stack exchange，提问作者BeastLike Good