Method analyze

agent/custom/reco/critter_crash.py:130–170

(
        self,
        context: Context,
        argv: CustomRecognition.AnalyzeArg,
    )

@AgentServer.custom_recognition("CCRemainMoney")
class CCRemainMoney(CustomRecognition):
    def analyze(
        self,
        context: Context,
        argv: CustomRecognition.AnalyzeArg,
    ) -> CustomRecognition.AnalyzeResult | RectType | None:

        type = parse_params(argv.custom_recognition_param, "type")["type"]

        img = argv.image
        # 定义目标颜色和颜色容差
        target_color = np.array([215, 241, 249])
        tolerance = 55  # 颜色容差，可以根据需要调整

        # 创建颜色过滤掩码
        lower_bound = np.maximum(target_color - tolerance, 0)
        upper_bound = np.minimum(target_color + tolerance, 255)

        # 创建掩码：保留在目标颜色范围内的像素
        color_mask = np.all((img >= lower_bound) & (img <= upper_bound), axis=-1)

        # 处理图像：保留目标颜色，其他颜色变成黑色
        # 创建一个全黑图像
        processed_img = np.zeros_like(img, dtype=np.uint8)
        # 保留匹配目标颜色的像素
        processed_img[color_mask] = img[color_mask]

        if type == 0:
            reco_detail = context.run_recognition("CCRemainMoney_rec", processed_img)
        elif type == 1:
            reco_detail = context.run_recognition(
                "CCRemainMoney_rec_refresh", processed_img
            )

        if is_hit(reco_detail):
            money = ocr_text(reco_detail)
            logger.debug(f"识别到剩余缪斯币: {money}")
            if int(money) >= 3:
                return CustomRecognition.AnalyzeResult(
                    box=reco_detail.box, detail=reco_detail.raw_detail
                )
        return CustomRecognition.AnalyzeResult(box=None, detail={})