| 123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197 |
- import cv2
- import numpy as np
- from module.logger import logger
- def detect_safe_area(image, target_gray_value=255, tolerance=0):
- """
- 在灰度图片中查找与目标灰度值在容差范围内的纯色区块,并返回这些区块的坐标。
- :param image_path: 图片路径
- :param target_gray_value: 目标灰度值,默认为255(最亮)
- :param tolerance: 灰度差异容忍度,默认为0(即完全匹配)
- :return: 包含找到的纯色区块坐标元组的列表
- """
- # 加载图片并转换为灰度图
- # img = Image.open(image_path).convert('L') # 'L' 表示转换为灰度图
- # img_arr = np.array(img) # 转换为numpy数组以便高效处理
- # 将图片转换为灰度图
- gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
- _, img_arr = cv2.threshold(gray, 127, 255, cv2.THRESH_BINARY)
- # 初始化结果列表
- solid_regions = []
- # 遍历图片的每个像素
- for y in range(img_arr.shape[0]): # 高度
- for x in range(img_arr.shape[1]): # 宽度
- pixel_gray_value = img_arr[y, x]
- # 检查当前像素灰度值是否与目标灰度值在容差范围内匹配
- if abs(pixel_gray_value - target_gray_value) <= tolerance:
- # 检查周围是否也是同样的灰度值,以确定这是一个区块而非单个像素
- is_solid_region = True
- for dy in range(-1, 2): # 检查3x3区域
- for dx in range(-1, 2):
- if dy == 0 and dx == 0:
- continue # 跳过中心点
- ny, nx = y + dy, x + dx
- if 0 <= ny < img_arr.shape[0] and 0 <= nx < img_arr.shape[1]:
- if abs(img_arr[ny, nx] - target_gray_value) > tolerance:
- is_solid_region = False
- break
- if not is_solid_region:
- break
- if is_solid_region:
- # 记录纯色区块的左上角坐标和右下角坐标
- start_x, start_y = x, y
- while start_y > 0 and abs(img_arr[start_y - 1, x] - target_gray_value) <= tolerance:
- start_y -= 1
- while start_x > 0 and abs(img_arr[y, start_x - 1] - target_gray_value) <= tolerance:
- start_x -= 1
- end_x, end_y = x, y
- while end_y < img_arr.shape[0] - 1 and abs(img_arr[end_y + 1, x] - target_gray_value) <= tolerance:
- end_y += 1
- while end_x < img_arr.shape[1] - 1 and abs(img_arr[y, end_x + 1] - target_gray_value) <= tolerance:
- end_x += 1
- solid_regions.append(((start_x, start_y), (end_x + 1, end_y + 1)))
- return solid_regions
- def detect_safe_area2(image, safe_color_lower, safe_color_upper, num_areas=3, debug=False):
- """
- 找出当前截图里最大的纯色区域,找出来的这个区域将被用于随机点击的区域。
- :note: 注意拿到这个区域后要马上点击,不要sleep!
- :param image: 截图图像
- :return: 最大纯色区域的坐标
- """
- # 将图片从BGR转换到HSV色彩空间
- hsv = cv2.cvtColor(image, cv2.COLOR_BGR2HSV)
- # 定义颜色范围
- # 注意:这里的颜色范围需要根据实际图片中的安全区域颜色进行调整
- mask = cv2.inRange(hsv, safe_color_lower, safe_color_upper)
- # 应用形态学操作来消除噪点和填补小的孔洞
- kernel = np.ones((5, 5), np.uint8)
- mask = cv2.dilate(mask, kernel, iterations=1)
- mask = cv2.erode(mask, kernel, iterations=1)
- # 查找轮廓
- contours, _ = cv2.findContours(mask, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
- # 计算每个轮廓的面积并排序
- # areas = [(cv2.contourArea(contour), contour) for contour in contours]
- areas = []
- for contour in contours:
- area = cv2.contourArea(contour)
- print(f"area={area}")
- areas.append((area, contour))
- # areas.sort(reverse=True)
- # 按面积降序排序
- areas_sorted = sorted(areas, key=lambda x: x[0], reverse=True)
- if debug:
- # 绘制面积排名前3的区域
- for area, contour in areas_sorted[:num_areas]:
- x1, y1, w1, h1 = cv2.boundingRect(contour)
- # 使用蓝色边框,线宽为4
- cv2.rectangle(image, (x1, y1), (x1 + w1, y1 + h1), (255, 0, 0), 4)
- # 使用红色边框,线宽为3
- cv2.drawContours(image, [contour], -1, (0, 0, 255), 3)
- # 假设最大的轮廓是安全区域
- if contours:
- safe_area = max(contours, key=cv2.contourArea)
- x, y, w, h = cv2.boundingRect(safe_area)
- if debug:
- # 绘制安全区域
- cv2.drawContours(image, [safe_area], -1, (0, 255, 0), 3)
- return (x, y, w, h)
- return (0, 0, 0, 0)
- def test_ocr_locate_dokan_target():
- import cv2
- from module.atom.ocr import RuleOcr
- logger.info("========================================>>test 2")
- O_DOKAN_MAP = RuleOcr(roi=(270, 130, 740, 460), area=(270, 130, 740, 460), mode="Full", method="Default",
- keyword="万", name="dokan_map")
- image = cv2.imread("g:/yys/oas/tests/1.png")
- pos = O_DOKAN_MAP.ocr_full(image)
- print(f"image1: {pos}")
- image = cv2.imread("g:/yys/oas/tests/2.png")
- pos = O_DOKAN_MAP.ocr_full(image)
- print(f"image2: {pos}")
- image = cv2.imread("g:/yys/oas/tests/3.png")
- pos = O_DOKAN_MAP.ocr_full(image)
- print(f"image3: {pos}")
- image = cv2.imread("g:/yys/oas/tests/4.png")
- pos = O_DOKAN_MAP.ocr_full(image)
- print(f"image4: {pos}")
- if pos == (0, 0, 0, 0):
- logger.info("failed to find {self.O_DOKAN_MAP.keyword}")
- else:
- # 取中间
- x = pos[0] + pos[2] / 2
- # 往上偏移20
- y = pos[1] - 20
- logger.info(f"ocr detect result pos={pos}, try click pos, x={x}, y={y}")
- def test_anti_detect_random_click():
- import cv2
- # image_list = ['g:/yys/oas/tests/1.png', 'g:/yys/oas/tests/2.png', 'g:/yys/oas/tests/3.png', 'g:/yys/oas/tests/4.png', 'g:/yys/oas/tests/5.png']
- image_list = ['g:/yys/oas/tests/a.png', 'g:/yys/oas/tests/b.png', 'g:/yys/oas/tests/c.png']
- for item in image_list:
- print(f"item: {item}")
- image = cv2.imread(item)
- # 注意:这里的颜色范围需要根据实际图片进行调整
- # http://kb.rg4.net/docs/omassistant/omassistant-1fnirpstkhdnh
- # 0<=h<20, 红色
- # 30<=h<45,黄色
- # 45<=h<90,绿色
- # 90<=h<125,青色
- # 125<=h<150,蓝色
- # 150<=h<175,紫色
- # 175<=h<200,粉红色
- # 200<=h<220,砖红色
- # 220<=h<255,品红色
- # 示例使用:假设安全区域是绿色的
- safe_color_lower = np.array([45, 25, 25]) # HSV颜色空间的绿色下界
- safe_color_upper = np.array([90, 255, 255]) # HSV颜色空间的绿色上界
- # 读取图片
- image = cv2.imread('c.png')
- if image is None:
- print("Error: Image not found.")
- else:
- # 调用函数
- pos = detect_safe_area2(image, safe_color_lower, safe_color_upper, 3, True)
- print(f"pos={pos}")
- # 显示结果
- cv2.imshow('Safe Area', image)
- cv2.waitKey(0)
- cv2.destroyAllWindows()
|