'commit'
This commit is contained in:
@@ -175,15 +175,21 @@ async def get_station_list(d, service, uploader, max_scrolls=MAX_SCROLLS):
|
||||
|
||||
if ad_type == "rabbit":
|
||||
logger.info(">>> 检测到顽固的兔子广告,执行强力双击策略...")
|
||||
# 使用 u2 的 double_click,它比两次 click + sleep 更快
|
||||
d.double_click(x, y, duration=0.05)
|
||||
|
||||
# 处理“误触背景卡片”的逻辑:既然可能进去了,就等一会儿再退出来
|
||||
logger.info(">>> 考虑到可能误触背景卡片,等待 2s 后执行返回操作...")
|
||||
await asyncio.sleep(2.0)
|
||||
d.press("back")
|
||||
await asyncio.sleep(1.0)
|
||||
logger.info(">>> 已尝试从误触页面返回。")
|
||||
# 处理“误触背景卡片”的逻辑
|
||||
logger.info(">>> 等待 3s 以确认是否进入了二级页面...")
|
||||
await asyncio.sleep(3.0)
|
||||
|
||||
# 检查是否还在列表页 (新电途列表页通常有“全城”或“综合排序”按钮)
|
||||
# 如果这些按钮不存在,说明可能进到了详情页
|
||||
is_list_page = d(text="全城").exists or d(text="综合排序").exists or d(text="搜索场站").exists
|
||||
if not is_list_page:
|
||||
logger.info(">>> 检测到当前不在列表页,尝试执行返回操作...")
|
||||
d.press("back")
|
||||
await asyncio.sleep(1.5)
|
||||
else:
|
||||
logger.info(">>> 当前仍在列表页,无需返回。")
|
||||
else:
|
||||
d.click(x, y)
|
||||
await asyncio.sleep(1.5)
|
||||
|
||||
@@ -137,6 +137,110 @@ def is_background_dimmed(image_path, threshold=80):
|
||||
return avg_brightness < threshold
|
||||
|
||||
|
||||
def detect_rabbit_ad_close(image_path, debug_dir=None):
|
||||
"""
|
||||
通过图形学算法检测“新电兔AI”广告的关闭按钮
|
||||
特征:黑色圆,中间有白色X
|
||||
:param image_path: 截图路径
|
||||
:param debug_dir: 调试图保存目录
|
||||
:return: (x, y) 归一化坐标,如果未找到返回 None
|
||||
"""
|
||||
if not os.path.exists(image_path):
|
||||
return None
|
||||
|
||||
img = read_image(image_path)
|
||||
if img is None:
|
||||
return None
|
||||
|
||||
h, w = img.shape[:2]
|
||||
|
||||
# ROI: 左侧,中下部 (x: 0-25%, y: 60-90%)
|
||||
roi_x1, roi_x2 = 0, int(w * 0.25)
|
||||
roi_y1, roi_y2 = int(h * 0.6), int(h * 0.9)
|
||||
roi = img[roi_y1:roi_y2, roi_x1:roi_x2]
|
||||
|
||||
gray = cv2.cvtColor(roi, cv2.COLOR_BGR2GRAY)
|
||||
|
||||
candidates = []
|
||||
# 尝试多个阈值以应对不同的亮度环境
|
||||
for threshold_val in [40, 60, 80, 100]:
|
||||
_, thresh = cv2.threshold(gray, threshold_val, 255, cv2.THRESH_BINARY_INV)
|
||||
|
||||
# Find contours
|
||||
contours, _ = cv2.findContours(thresh, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
|
||||
|
||||
for cnt in contours:
|
||||
area = cv2.contourArea(cnt)
|
||||
# 圆形度检查
|
||||
perimeter = cv2.arcLength(cnt, True)
|
||||
if perimeter == 0: continue
|
||||
circularity = 4 * np.pi * (area / (perimeter * perimeter))
|
||||
|
||||
# 兔子广告关闭按钮通常很小 (面积在 100-4000 之间)
|
||||
if 100 < area < 4000 and circularity > 0.4:
|
||||
# 获取该候选区域的 bounding box
|
||||
x, y, w_cnt, h_cnt = cv2.boundingRect(cnt)
|
||||
|
||||
# 在这个黑色圆内部,检查是否有亮色的 'X'
|
||||
padding = 2
|
||||
inner_roi = gray[max(0, y-padding):min(roi.shape[0], y+h_cnt+padding),
|
||||
max(0, x-padding):min(roi.shape[1], x+w_cnt+padding)]
|
||||
|
||||
# 找亮色物体 (X)
|
||||
_, inner_thresh = cv2.threshold(inner_roi, 180, 255, cv2.THRESH_BINARY)
|
||||
inner_contours, _ = cv2.findContours(inner_thresh, cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)
|
||||
|
||||
has_x = False
|
||||
for i_cnt in inner_contours:
|
||||
i_area = cv2.contourArea(i_cnt)
|
||||
# X 应该比圆小很多
|
||||
if 10 < i_area < area * 0.5:
|
||||
has_x = True
|
||||
break
|
||||
|
||||
if has_x or circularity > 0.7:
|
||||
M = cv2.moments(cnt)
|
||||
if M["m00"] != 0:
|
||||
cX = int(M["m10"] / M["m00"]) + roi_x1
|
||||
cY = int(M["m01"] / M["m00"]) + roi_y1
|
||||
|
||||
norm_x = int(cX / w * 1000)
|
||||
norm_y = int(cY / h * 1000)
|
||||
|
||||
# 避免重复
|
||||
if not any(abs(cX - c[0]) < 15 and abs(cY - c[1]) < 15 for c in candidates):
|
||||
candidates.append((cX, cY, area, norm_x, norm_y, has_x))
|
||||
|
||||
if not candidates:
|
||||
return None
|
||||
|
||||
# 评分逻辑
|
||||
def score_candidate(c):
|
||||
# c = (cx, cy, area, nx, ny, has_x)
|
||||
has_x = c[5]
|
||||
# 基础分:如果有 X,大幅加分
|
||||
score = 1000 if has_x else 0
|
||||
# 距离分:越靠近预期的 (93, 830) 分越高
|
||||
dist = np.sqrt((c[3] - 93)**2 + (c[4] - 830)**2)
|
||||
score -= dist * 2
|
||||
# 面积分:理想面积在 500-1500 之间
|
||||
if 500 < c[2] < 1500: score += 200
|
||||
return score
|
||||
|
||||
candidates.sort(key=score_candidate, reverse=True)
|
||||
best = candidates[0]
|
||||
|
||||
logger.info(f"CV detected rabbit ad close button at Norm({best[3]}, {best[4]}) with score {score_candidate(best):.2f}")
|
||||
|
||||
if debug_dir:
|
||||
os.makedirs(debug_dir, exist_ok=True)
|
||||
debug_img = img.copy()
|
||||
cv2.circle(debug_img, (best[0], best[1]), 10, (0, 0, 255), -1) # 使用红点表示识别结果
|
||||
save_image(os.path.join(debug_dir, "debug_rabbit_ad_cv.jpg"), debug_img)
|
||||
|
||||
return [best[3], best[4]]
|
||||
|
||||
|
||||
def detect_black_agree_button(image_path, debug_dir=None):
|
||||
"""
|
||||
通过计算机图形学检测黑色的"同意"按钮 (Image 1 场景)
|
||||
|
||||
@@ -68,15 +68,20 @@ async def check_and_close_ad(d):
|
||||
# 针对“新电兔AI”兔子广告的强力关闭策略
|
||||
if ad_type == "rabbit":
|
||||
logger.info(">>> 检测到顽固的兔子广告,执行强力双击策略...")
|
||||
# 使用 u2 的 double_click,它比两次 click + sleep 更快
|
||||
d.double_click(x, y, duration=0.05)
|
||||
|
||||
# 处理“误触背景卡片”的逻辑
|
||||
logger.info(">>> 考虑到可能误触背景卡片,等待 2s 后执行返回操作...")
|
||||
await asyncio.sleep(2.0)
|
||||
d.press("back")
|
||||
await asyncio.sleep(1.0)
|
||||
logger.info(">>> 已尝试从误触页面返回。")
|
||||
logger.info(">>> 等待 3s 以确认是否进入了二级页面...")
|
||||
await asyncio.sleep(3.0)
|
||||
|
||||
# 检查是否还在列表页
|
||||
is_list_page = d(text="全城").exists or d(text="综合排序").exists or d(text="搜索场站").exists
|
||||
if not is_list_page:
|
||||
logger.info(">>> 检测到当前不在列表页,尝试执行返回操作...")
|
||||
d.press("back")
|
||||
await asyncio.sleep(1.5)
|
||||
else:
|
||||
logger.info(">>> 当前仍在列表页,无需返回。")
|
||||
else:
|
||||
d.click(x, y)
|
||||
|
||||
|
||||
@@ -8,8 +8,11 @@ import aiohttp
|
||||
import logging
|
||||
import base64
|
||||
from openai import AsyncOpenAI, BadRequestError
|
||||
from Apps.XinDianTu.Kit import (
|
||||
read_image, save_image, detect_rabbit_ad_close
|
||||
)
|
||||
from Config.Config import (
|
||||
ALY_LLM_API_KEY, VL_MODEL_NAME, VL_MODEL_NAME_AD
|
||||
ALY_LLM_API_KEY, VL_MODEL_NAME, VL_MODEL_NAME_AD, TEMP_IMAGE_DIR
|
||||
)
|
||||
from Apps.XinDianTu.Config.Setting import (
|
||||
SAFE_EXCLUDE_RATIO, FALLBACK_WIDTH, FALLBACK_HEIGHT,
|
||||
@@ -77,11 +80,28 @@ class ReadImageKit:
|
||||
@classmethod
|
||||
async def detect_ad_popup(cls, image_path: str, device_info=None):
|
||||
"""
|
||||
使用 VL 模型检测是否存在弹窗广告,并返回关闭按钮坐标
|
||||
使用 VL 模型 + 图形学算法检测是否存在弹窗广告,并返回关闭按钮坐标
|
||||
"""
|
||||
if not os.path.exists(image_path):
|
||||
return None
|
||||
|
||||
# 1. 优先使用图形学算法检测“新电兔AI”广告,因为它的特征非常明显
|
||||
cv_rabbit_point = detect_rabbit_ad_close(image_path, debug_dir=TEMP_IMAGE_DIR)
|
||||
if cv_rabbit_point:
|
||||
logger.info(f">>> 图形学算法精准捕捉到兔子广告关闭按钮: {cv_rabbit_point}")
|
||||
norm_point = cv_rabbit_point
|
||||
ad_type = "rabbit"
|
||||
|
||||
# 转换坐标并返回
|
||||
if not device_info:
|
||||
device_info = cls._FALLBACK_DEVICE_INFO
|
||||
w = device_info.get("displayWidth", FALLBACK_WIDTH)
|
||||
h = device_info.get("displayHeight", FALLBACK_HEIGHT)
|
||||
x = int(norm_point[0] / 1000 * w)
|
||||
y = int(norm_point[1] / 1000 * h)
|
||||
return {"x": x, "y": y, "ad_type": ad_type}
|
||||
|
||||
# 2. 如果图形学没找到,再请求大模型进行全量检测
|
||||
with open(image_path, "rb") as image_file:
|
||||
encoded_image = base64.b64encode(image_file.read()).decode("utf-8")
|
||||
img_input = {"url": f"data:image/jpeg;base64,{encoded_image}"}
|
||||
@@ -91,26 +111,21 @@ class ReadImageKit:
|
||||
"请仔细检查这张图片中是否存在**弹窗广告**或**悬浮广告**。\n"
|
||||
"广告可能有以下几种形式:\n"
|
||||
"1. **屏幕中央的大型弹窗广告**:通常遮挡了页面内容,内容多为优惠券、活动推广等。\n"
|
||||
"2. **'新电兔AI'专属悬浮广告**(高频):\n"
|
||||
" - **位置范围**:严格锁定在**屏幕左侧边缘**(水平坐标约 0-250),**垂直高度处于中下部**(垂直坐标约 600-900)。\n"
|
||||
" - **视觉特征**:一个卡通兔子头,或带有'新电兔AI'字样的气泡图标。\n"
|
||||
" - **关闭按钮**:务必寻找位于该图标**正下方**的一个**黑色圆形小按钮(内含白色X)**。注意:关闭按钮通常与图标本体分离,悬浮在图标下方约 20-50 像素处。请精确返回这个黑色圆形按钮的中心坐标。\n"
|
||||
"2. **'新电兔AI'专属悬浮广告**(高频):一个卡通兔子头,带有'新电兔AI'字样,关闭按钮是一个黑色圆圈内含白色X。\n"
|
||||
"3. **底部的横幅广告**:带有明显的关闭按钮。\n\n"
|
||||
"如果发现了上述任何一种广告,请务必找到该广告的**关闭按钮**(通常是一个'X'符号,或者'关闭'字样,或者位于广告下方/右上角的圆圈叉号)。\n"
|
||||
"请返回该关闭按钮的**中心坐标**。\n\n"
|
||||
"请返回关闭按钮的中心坐标。\n\n"
|
||||
"请以纯 JSON 格式输出:\n"
|
||||
"{\n"
|
||||
" \"has_ad\": true/false,\n"
|
||||
" \"ad_type\": \"rabbit\", // \"rabbit\" (新电兔AI/兔子) | \"center\" (中央弹窗) | \"bottom\" (底部横幅) | \"other\"\n"
|
||||
" \"ad_type\": \"rabbit\" | \"center\" | \"bottom\" | \"other\",\n"
|
||||
" \"close_point\": [x, y] // 0-1000 归一化坐标\n"
|
||||
"}\n"
|
||||
"如果没有广告,请返回 `{\"has_ad\": false}`。\n"
|
||||
"只输出 JSON,不要包含任何其他文字。"
|
||||
"如果没有广告,请返回 `{\"has_ad\": false}`。"
|
||||
)
|
||||
|
||||
try:
|
||||
resp = await cls._client.chat.completions.create(
|
||||
model=VL_MODEL_NAME, # 使用默认 VL 模型即可,或者使用 AD 专用
|
||||
model=VL_MODEL_NAME,
|
||||
timeout=30,
|
||||
messages=[
|
||||
{
|
||||
@@ -128,10 +143,9 @@ class ReadImageKit:
|
||||
|
||||
if result.get("has_ad") and result.get("close_point"):
|
||||
if result.get("ad_type") == "rabbit":
|
||||
logger.info("大模型找到了兔子广告!")
|
||||
# 强制使用固定坐标 (94, 830) - 归一化 0-1000
|
||||
# 根据多次 CV 测试结果 (94, 828) 和 (93, 830) 的综合评估
|
||||
logger.info(">>> 采用强制固定坐标策略 (94, 830) 以确保点击准确性")
|
||||
logger.info("大模型找到了兔子广告,但 CV 没找到。")
|
||||
# 如果大模型找到了但 CV 没找到,说明可能 CV 阈值没对上,此时采用固定坐标作为兜底
|
||||
logger.info(">>> 采用兜底固定坐标策略 (94, 830)")
|
||||
norm_point = [94, 830]
|
||||
else:
|
||||
norm_point = result["close_point"]
|
||||
@@ -144,7 +158,6 @@ class ReadImageKit:
|
||||
w = device_info.get("displayWidth", FALLBACK_WIDTH)
|
||||
h = device_info.get("displayHeight", FALLBACK_HEIGHT)
|
||||
|
||||
# 假设模型返回的是 0-1000 的坐标
|
||||
x = int(norm_point[0] / 1000 * w)
|
||||
y = int(norm_point[1] / 1000 * h)
|
||||
|
||||
|
||||
Binary file not shown.
Binary file not shown.
Reference in New Issue
Block a user