# This Python file uses the following encoding: utf-8 # @author runhey # github https://github.com/runhey import types import numpy as np from module.atom.image import RuleImage class RuleGif: # 大部分实现同RuleImage 的接口 @staticmethod def attach_to(image_obj: RuleImage, targets: list[RuleImage]) -> None: """ 把GIF的多帧匹配行为原地挂载到一个RuleImage实例上 与replace_img的原地换属性同思路: 不换对象、不换引用, 持有该对象引用的Page/寻路会立即生效 :param image_obj: 被挂载的RuleImage实例 :param targets: 多帧RuleImage, 顺序即匹配优先级 """ # GIF状态 image_obj.targets = targets image_obj.roi_front = [0, 0, 0, 0] image_obj.roi_back = targets[0].roi_back image_obj.appear_target = targets[0] image_obj.name = targets[0].name image_obj._match_init = False # 方法级替换: 把RuleGif的所有公开方法绑定到该实例上 for name, attr in RuleGif.__dict__.items(): if name.startswith('_'): continue if isinstance(attr, types.FunctionType): setattr(image_obj, name, types.MethodType(attr, image_obj)) @staticmethod def detach_from(image_obj: RuleImage) -> None: """Remove attached GIF behavior while preserving the image object's identity.""" if 'targets' not in image_obj.__dict__: return for name, attr in RuleGif.__dict__.items(): method = image_obj.__dict__.get(name) if isinstance(method, types.MethodType) and method.__func__ is attr: delattr(image_obj, name) for name in ('targets', 'appear_target', 'name'): image_obj.__dict__.pop(name, None) @property def name(self) -> str: return self.targets[0].name def __init__(self, targets: list[RuleImage]): self.targets = targets self.roi_front: list = [0, 0, 0, 0] self.roi_back: list = targets[0].roi_back self._match_init = False self.appear_target = targets[0] def pre_process(self, image): return image def search(self, image, roi: list = None, threshold: float = None) -> tuple: """ :param image: :param roi: :param threshold: :return: bool 第一项是否为出现, 第二项为匹配的RuleImage """ image = self.pre_process(image) # threshold = self.targets[0].threshold if threshold is None else threshold roi = self.targets[0].roi_back if roi is None else roi for target in self.targets: target.roi_back = roi if target.match(image, threshold): self.roi_front = target.roi_front self.appear_target = target return True, target return False, None def search_with_multi_scale(self, image, roi=None, threshold=None, scales=None, scale_range=(0.8, 1.1, 0.05)): image = self.pre_process(image) threshold = self.targets[0].threshold if threshold is None else threshold roi = self.targets[0].roi_back if roi is None else roi for target in self.targets: target.roi_back = roi if target.match_multi_scale(image, threshold=threshold, scales=scales, scale_range=scale_range): self.roi_front = target.roi_front self.appear_target = target return True, target return False, None def match_multi_scale(self, image, threshold: float = None, scales: list = None, scale_range: tuple = None) -> bool: return self.search_with_multi_scale(image, threshold=threshold, scales=scales, scale_range=scale_range)[0] def match(self, image, threshold: float = None) -> bool: return self.search(image, threshold=threshold)[0] def coord(self) -> tuple: x, y, w, h = self.roi_front return x + np.random.randint(0, w), y + np.random.randint(0, h) def front_center(self) -> tuple: x, y, w, h = self.roi_front return int(x + w//2), int(y + h//2)