import threading from enum import Enum import cv2 import numpy as np from pydantic import BaseModel from module.base.timer import Timer from module.logger import logger from tasks.Component.GeneralBattle.config_general_battle import GreenMarkType from tasks.Component.GeneralBattle.general_battle import GeneralBattle from tasks.Dokan.utils import retry # # def green_marker_detector(func): # def wrapper(self, *args, **kwargs): # event = threading.Event() # # def detect_green_marker(image): # hsv_image = cv2.cvtColor(image, cv2.COLOR_RGB2HSV) # lower_green = np.array([67, 115, 110]) # upper_green = np.array([80, 255, 232]) # mask = cv2.inRange(hsv_image, lower_green, upper_green) # result = cv2.bitwise_and(image, image, mask=mask) # res = self.I_GREEN_MARKER.match(result) # return res # # def emit_exists(): # logger.info("Green marker detected!") # unpack_screenshot() # event.set() # # def screenshot_and_detect(): # self.just_screenshot() # exists = detect_green_marker(self.device.image) # if exists: # emit_exists() # return self.device.image # # def pack_screenshot(): # self.just_screenshot = self.screenshot # self.screenshot = screenshot_and_detect # # def unpack_screenshot(): # self.screenshot = self.just_screenshot # # def thread_run(self, detect_time=3, cb=None, cbargs=None): # logger.info("detect thread running...") # t = Timer(detect_time) # t.start() # while event.wait(detect_time): # logger.info("detect thread quit") # return # unpack_screenshot() # if cb: # logger.info(f"didn't detect green mark ,start running callback with arg:{cbargs}") # try: # cb(**cbargs) # except Exception as e: # logger.error(f"callback error:{e}") # # # wrapper # callback = kwargs.get('callback', None) # cbargs = kwargs.get('cbargs', None) # detect_time = kwargs.get('detect_time', 3.0) # # if callback: # logger.info("Green marker detector running...") # pack_screenshot() # # green mark 函数的参数列表 # kwarg_key = ['enable', 'mark_mode'] # func_kwargs = {key: kwargs[key] for key in kwarg_key if key in kwargs} # if not event.is_set(): # func(self, *args, **func_kwargs) # # if callback: # thread = threading.Thread(target=thread_run, args=(self, detect_time, callback, cbargs), daemon=True) # thread.start() # return # # return wrapper class GreenMarkState(int, Enum): # NOT_INIT = 0, INITED = 1, # 备用 MARKED = 2, DISAPPEARED = 3 class ExtendGreenMark(GeneralBattle): # 式神名称 _shikigami_name = "" # 绿标检测范围,x,y,w,h _green_mark_detect_area: list[int] = None # 式神名称范围,OCR检测出的式神名位置,x,y,w,h _green_mark_click_roi: list[int] = None # 绿标消失临界值 MAX_DISAPPEAR_COUNT = 40 MAX_DISAPPEAR_TIME = 4 # 尝试生成绿标的间隔 GREEN_MARK_INTERVAL = 1 # 绿标消失后,screenshot的次数 _disappear_count = 0 # 绿标消失的时间 _disappear_timer = Timer(MAX_DISAPPEAR_TIME) # 判断是否开始检测绿标 _state = GreenMarkState.NOT_INIT # def __init__(self, config, device): super().__init__(config, device) self.super_green_mark = self.green_mark def init_green_mark_from_cfg(self, names: str = None, green_mark_type: GreenMarkType = None): self._shikigami_name = names if self._shikigami_name is None or self._shikigami_name == "": # 使用默认的绿标模式 tmp = self.generate_green_mark_area(green_mark_type) self._green_mark_click_roi = tmp.roi_front else: # 使用式神名称进行绿标 def detect_name_position_retry(): self.device.screenshot() return self.detect_name_position(self.device.image, self._shikigami_name) # 多试几次,避免识别失败 self._green_mark_click_roi = retry(detect_name_position_retry, 3) # self._green_mark_detect_area = self.calc_green_mark_locate_area(self._green_mark_click_roi) self.start_green_mark() return # @green_marker_detector # def ex_green_mark(self, enable=True, mark_mode=GreenMarkType.GREEN_MAIN): # if not enable: # return # return self.super_green_mark(enable=enable, mark_mode=mark_mode) # # def ex_green_mark_3_retry(self, enable=True, mark_mode=GreenMarkType.GREEN_MAIN): # if not enable: # return # return self.ex_green_mark(enable=enable, mark_mode=mark_mode, detect_time=3, callback=self.super_green_mark, # cbargs={'enable': enable, 'mark_mode': mark_mode}) def generate_green_mark_area(self, mark_mode: GreenMarkType = GreenMarkType.GREEN_MAIN): result = None logger.info(mark_mode.name) match mark_mode: case GreenMarkType.GREEN_LEFT1: result = self.C_GREEN_LEFT_1 logger.info("Green left 1") case GreenMarkType.GREEN_LEFT2: result = self.C_GREEN_LEFT_2 logger.info("Green left 2") case GreenMarkType.GREEN_LEFT3: result = self.C_GREEN_LEFT_3 logger.info("Green left 3") case GreenMarkType.GREEN_LEFT4: result = self.C_GREEN_LEFT_4 logger.info("Green left 4") case GreenMarkType.GREEN_LEFT5: result = self.C_GREEN_LEFT_5 logger.info("Green left 5") case GreenMarkType.GREEN_MAIN: result = self.C_GREEN_MAIN logger.info("Green main") return result def detect_name_position(self, img, names: str) -> list[int]: def similarity(txt1, txt2): txtset1 = set(txt1) txtset2 = set(txt2) intersection = txtset1.intersection(txtset2) union = txtset1.union(txtset2) return len(intersection) / len(union) name_list = names.split(',') for name in name_list: # 此处为了获取到OCR模型,随机选择一个了RuleOcr对象,可以使用任意的RuleOcr对象 res_list = self.O_EXP_50.model.detect_and_ocr(img, 0.7) # 增加日志输出,方便纠错 output = [item.ocr_text for item in res_list] logger.info(f"detect res_list:{output}") res = [item for item in res_list if similarity(item.ocr_text, name) > 0.5] res = sorted(res, key=lambda x: x.score) if len(res) <= 0: continue box = res[0].box logger.info(f"detect{res[0].ocr_text} with name:{name} in {box}") # 经实验,点击式神名称位置,可能点击不到,故此做一个修正 offset = [5, 30, -10, 0] # x,y,w,h return [box[0, 0] + offset[0], box[0, 1] + offset[1], box[2, 0] - box[0, 0] + offset[2], box[2, 1] - box[0, 1] + offset[3]] return None def calc_green_mark_locate_area(self, roi, margin=None): """ @param roi: 根据名字OCR得出的ROI,格式为x,y,w,h @type roi: list[int] @param margin: 扩展大小,格式为top,right,bottom,left,同CSS margin @type margin: @return: @rtype: """ if roi is None: return [0, 0, 1280, 720] result = [0, 0, 0, 0] if margin is None: margin = [160, 40, 20, 80] result[0] = x if (x := roi[0] - margin[3]) > 0 else 0 result[1] = y if (y := roi[1] - margin[0]) > 0 else 0 result[2] = w if (w := roi[2] + margin[1] + margin[3]) + x <= 1280 else 1280 - x result[3] = h if (h := roi[3] + margin[0] + margin[2]) + y <= 720 else 720 - y return result def detect_green_mark(self, img=None, area=None): def init_green_marker_roi(roi): self.I_GREEN_MARKER.roi_back = roi self.I_GREEN_MARKER_LEFT_TOP.roi_back = roi self.I_GREEN_MARKER_BOTTOM.roi_back = roi # 初始化绿标相关变量的roi信息 if area is not None: init_green_marker_roi(area) elif self._green_mark_detect_area is not None: init_green_marker_roi(self._green_mark_detect_area) else: init_green_marker_roi((0, 0, 1280, 720)) if img is not None: self.device.image = img def detect_green_marker_base(image): hsv_image = cv2.cvtColor(image, cv2.COLOR_RGB2HSV) lower_green = np.array([67, 115, 110]) upper_green = np.array([80, 255, 232]) mask = cv2.inRange(hsv_image, lower_green, upper_green) result = cv2.bitwise_and(image, image, mask=mask) # 伤害数字,御魂生效后弹出会遮挡绿标信息,目前的策略:尽量的延长检测到绿标的时间 res = (self.I_GREEN_MARKER_LEFT_TOP.match(result) or self.I_GREEN_MARKER_BOTTOM.match(result) or self.I_GREEN_MARKER.match(result)) return res if detect_green_marker_base(self.device.image): return True return False def green_mark_screenshot(self, callback=None): """ 替代原有screenshot @return: @rtype: """ self.screenshot() # 若未初始化,直接返回,不检测绿标状态 if self._state == GreenMarkState.NOT_INIT or self._state is None: return self.device.image # 检测式神名区域 if self._green_mark_click_roi is None and (self._shikigami_name is not None and self._shikigami_name != ""): self._green_mark_click_roi = self.detect_name_position(self.device.image, self._shikigami_name) if self._green_mark_click_roi is not None: self._green_mark_detect_area = self.calc_green_mark_locate_area(self._green_mark_click_roi) # 如果检测到绿标,直接返回 if self.detect_green_mark(self.device.image): self._disappear_count = 0 self._disappear_timer.reset() if self._state != GreenMarkState.MARKED: logger.info("green mark appear") self._state = GreenMarkState.MARKED return self.device.image if self._state != GreenMarkState.DISAPPEARED: logger.info("green mark DISappear") self._state = GreenMarkState.DISAPPEARED self._disappear_count += 1 self._disappear_timer.start() if self.need_green_mark(): if self.green_mark_with_area(interval=self.GREEN_MARK_INTERVAL): self._disappear_timer.reset() self._disappear_count = 0 if callback is not None: callback() return self.device.image def need_green_mark(self): result = self._disappear_count > self.MAX_DISAPPEAR_COUNT if result: logger.info("MAX DISAPPEAR COUNT reached,Need Green Mark") return True result |= self._disappear_timer.reached() if result: logger.info("MAX DISAPPEAR TIME reached,Need Green Mark") return True return result def green_mark_with_area(self, area=None, interval=1.5): """ 点击区域尝试添加绿标,不保证添加成功 @param area: 格式 x,y,w,h @type area: @return: @rtype: """ if area is None: area = self._green_mark_click_roi if area is None: logger.warn("Green Mark with area None,give up") return False self.C_GREEN_MARK_AREA.roi_front = area if self.click(self.C_GREEN_MARK_AREA, interval=interval): # NOTE 防止式神死亡导致的连续点击,程序退出 self.device.click_record_clear() logger.info(f"green mark with area:{area}") return True return False def set_shikigami_name(self, name): self._shikigami_name = name def start_green_mark(self): self._state = GreenMarkState.INITED self._disappear_timer.clear() self._disappear_count = 0 def stop_green_mark(self): self._state = GreenMarkState.NOT_INIT def set_disappear_count(self, count): # 设置消失次数,用于(主动触发执行绿标) self._disappear_count = count def change_green_mark(obj): obj.old_green_mark = obj.green_mark obj.green_mark = obj.ex_green_mark_3_retry if __name__ == '__main__': # 读取图片,然后转换为HSV格式 ,使用cv2 matchtemplate 比较 import cv2 import numpy as np def detect_green_marker_with_template(image_path, template_path): # 读取图片 image = cv2.imread(image_path) if image is None: logger.error(f"Failed to read image from {image_path}") return None # 读取模板图片 template = cv2.imread(template_path) if template is None: logger.error(f"Failed to read template from {template_path}") return None # 转换为HSV格式 image = cv2.cvtColor(image, cv2.COLOR_BGR2HLS) template = cv2.cvtColor(template, cv2.COLOR_BGR2HLS) # # image_h, image_s, image_v = cv2.split(image) # template_h, template_s, template_v = cv2.split(template) # image_s_norm = cv2.normalize(image_s, None, 0, 255, cv2.NORM_MINMAX) # image_v_norm = cv2.normalize(image_v, None, 0, 255, cv2.NORM_MINMAX) # template_s_norm = cv2.normalize(template_s, None, 0, 255, cv2.NORM_MINMAX) # template_v_norm = cv2.normalize(template_v, None, 0, 255, cv2.NORM_MINMAX) # clahe = cv2.createCLAHE(clipLimit=3.0, tileGridSize=(8, 8)) # image = clahe.apply(image_v) # template = clahe.apply(template_v) # 合并增强后的 HSV 通道 # image_enhanced_hsv = cv2.merge([image_h, image_s, image_v]) # template_enhanced_hsv = cv2.merge([template_h, template_s, template_v]) # 合并归一化后的 HSV 通道 # image = cv2.merge([image_h, image_s_norm, image_v_norm]) # template = cv2.merge([template_h, template_s_norm, template_v_norm]) # 使用matchTemplate进行比较 result = cv2.matchTemplate(image, template, cv2.TM_CCOEFF_NORMED) min_val, max_val, min_loc, max_loc = cv2.minMaxLoc(result) # 获取匹配结果的位置和相似度 top_left = max_loc h, w = template.shape[:2] bottom_right = (top_left[0] + w, top_left[1] + h) # 绘制矩形框标记匹配区域(可选) cv2.rectangle(image, top_left, bottom_right, (0, 255, 0), 2) # 返回匹配结果 return { "position": top_left, "similarity": max_val, "image": image } # 示例调用 result = detect_green_marker_with_template("E:\\111.png", "E:\\same.png") if result: logger.info(f"Match found at position: {result['position']} with similarity: {result['similarity']}") else: logger.info("No match found.") result = detect_green_marker_with_template("E:\\111.png", "E:\\diff.png") if result: logger.info(f"Match found at position: {result['position']} with similarity: {result['similarity']}") else: logger.info("No match found.") result = detect_green_marker_with_template("E:\\222.png", "E:\\gb_prepare_highlight.png") if result: logger.info(f"Match found at position: {result['position']} with similarity: {result['similarity']}") else: logger.info("No match found.") result = detect_green_marker_with_template("E:\\222.png", "E:\\gb_prepare_dark.png") if result: logger.info(f"Match found at position: {result['position']} with similarity: {result['similarity']}") else: logger.info("No match found.")