'commit'
This commit is contained in:
HuangHai
@@ -93,270 +93,6 @@ def save_image(path, img):
|
||||
logger.error(f"Error saving image {path}: {e}")
|
||||
return False
|
||||
|
||||
def detect_list_price_blocks_cv(image_path, top_ratio=0.40, bottom_ratio=0.12):
|
||||
img = read_image(image_path)
|
||||
if img is None:
|
||||
return []
|
||||
h, w = img.shape[:2]
|
||||
top_limit = int(h * top_ratio)
|
||||
bottom_limit = int(h * (1 - bottom_ratio))
|
||||
roi_bgr = img[top_limit:bottom_limit, :, :]
|
||||
if roi_bgr.size == 0:
|
||||
return []
|
||||
b, g, r = cv2.split(roi_bgr)
|
||||
mask_bgr = (r > 190) & (g > 80) & (r - g > 25) & (r - b > 25)
|
||||
mask_bgr = (mask_bgr.astype(np.uint8)) * 255
|
||||
if cv2.countNonZero(mask_bgr) < 50:
|
||||
hsv = cv2.cvtColor(roi_bgr, cv2.COLOR_BGR2HSV)
|
||||
lower_red1 = np.array([0, 100, 120])
|
||||
upper_red1 = np.array([20, 255, 255])
|
||||
lower_red2 = np.array([160, 100, 120])
|
||||
upper_red2 = np.array([180, 255, 255])
|
||||
mask1 = cv2.inRange(hsv, lower_red1, upper_red1)
|
||||
mask2 = cv2.inRange(hsv, lower_red2, upper_red2)
|
||||
mask = cv2.bitwise_or(mask1, mask2)
|
||||
else:
|
||||
mask = mask_bgr
|
||||
kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (30, 10))
|
||||
closed = cv2.morphologyEx(mask, cv2.MORPH_CLOSE, kernel)
|
||||
contours, _ = cv2.findContours(closed, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
|
||||
blocks = []
|
||||
for cnt in contours:
|
||||
x, y, cw, ch = cv2.boundingRect(cnt)
|
||||
y_global = y + top_limit
|
||||
if ch < 30 or ch > 140:
|
||||
continue
|
||||
if cw < 80 or cw > int(w * 0.8):
|
||||
continue
|
||||
center_y = y_global + ch // 2
|
||||
if center_y < int(h * 0.6) or center_y > bottom_limit:
|
||||
continue
|
||||
blocks.append([x, y_global, x + cw, y_global + ch])
|
||||
blocks.sort(key=lambda b: b[1])
|
||||
return blocks
|
||||
|
||||
def detect_price_click_point_cv(image_path):
|
||||
"""
|
||||
使用 HSV 颜色过滤定位详情页的橘红色价格区域，返回最左侧区域的中心点击点
|
||||
"""
|
||||
img = read_image(image_path)
|
||||
if img is None:
|
||||
return None
|
||||
|
||||
h, w = img.shape[:2]
|
||||
# 1. 转换为 HSV 空间
|
||||
hsv = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
|
||||
|
||||
# 2. 橘红色的 HSV 范围 (适配特来电价格颜色)
|
||||
lower_orange = np.array([0, 150, 150])
|
||||
upper_orange = np.array([20, 255, 255])
|
||||
mask = cv2.inRange(hsv, lower_orange, upper_orange)
|
||||
|
||||
# 3. 对掩码进行膨胀，连接数字
|
||||
kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (50, 20))
|
||||
dilated = cv2.dilate(mask, kernel)
|
||||
|
||||
# 4. 寻找轮廓
|
||||
contours, _ = cv2.findContours(dilated, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
|
||||
|
||||
detected_areas = []
|
||||
for cnt in contours:
|
||||
x, y, cw, ch = cv2.boundingRect(cnt)
|
||||
# 1. 过滤掉宽度过大（可能是横幅广告）或过小（可能是杂点）的区域
|
||||
# 2. 价格区域 P0 通常在屏幕的中部，且宽度约为屏幕的一半
|
||||
if 200 < y < h * 0.8 and 100 < cw < w * 0.6 and ch > 30:
|
||||
detected_areas.append([x, y, x + cw, y + ch])
|
||||
|
||||
if not detected_areas:
|
||||
# 备选：如果 HSV 失败，尝试通过轮廓大小寻找
|
||||
# 1.1556 这种大数字通常会有很明显的轮廓
|
||||
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
|
||||
thresh = cv2.adaptiveThreshold(gray, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY_INV, 11, 2)
|
||||
kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (30, 10))
|
||||
closed = cv2.morphologyEx(thresh, cv2.MORPH_CLOSE, kernel)
|
||||
contours, _ = cv2.findContours(closed, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
|
||||
for cnt in contours:
|
||||
x, y, cw, ch = cv2.boundingRect(cnt)
|
||||
if 200 < y < h * 0.6 and 150 < cw < 300 and 50 < ch < 150:
|
||||
detected_areas.append([x, y, x + cw, y + ch])
|
||||
|
||||
if not detected_areas:
|
||||
return None
|
||||
|
||||
# 5. 按 X 轴排序，取最左边的区域 (即用户确认的 P0)
|
||||
# 但要排除掉可能在最左侧的导航栏返回按钮等小元素，所以前面加了宽度限制
|
||||
detected_areas.sort(key=lambda b: b[0])
|
||||
target = detected_areas[0]
|
||||
|
||||
center_x = (target[0] + target[2]) // 2
|
||||
center_y = (target[1] + target[3]) // 2
|
||||
|
||||
return [center_x, center_y]
|
||||
|
||||
def detect_cards_cv(image_path, top_ratio=0.40, bottom_ratio=0.12):
|
||||
"""
|
||||
使用计算机图形学 (OpenCV) 检测列表中的场站卡片。
|
||||
"""
|
||||
from Apps.TeLaiDian.Config.Setting import MIN_CARD_HEIGHT
|
||||
|
||||
img = read_image(image_path)
|
||||
if img is None:
|
||||
return []
|
||||
|
||||
h, w = img.shape[:2]
|
||||
|
||||
price_blocks = detect_list_price_blocks_cv(image_path, top_ratio=top_ratio, bottom_ratio=bottom_ratio)
|
||||
cards = []
|
||||
if price_blocks:
|
||||
from Apps.TeLaiDian.Config.Setting import SAFE_EXCLUDE_RATIO, BOTTOM_SAFE_EXCLUDE_RATIO
|
||||
top_limit = int(h * SAFE_EXCLUDE_RATIO)
|
||||
bottom_limit = int(h * (1 - BOTTOM_SAFE_EXCLUDE_RATIO))
|
||||
for b in price_blocks:
|
||||
px1, py1, px2, py2 = b
|
||||
center_y = (py1 + py2) // 2
|
||||
card_height = MIN_CARD_HEIGHT
|
||||
card_top = max(top_limit, center_y - int(card_height * 0.7))
|
||||
card_bottom = card_top + card_height
|
||||
if card_bottom > bottom_limit:
|
||||
card_bottom = bottom_limit
|
||||
card_top = max(top_limit, card_bottom - card_height)
|
||||
card_left = 0
|
||||
card_right = w
|
||||
cards.append([card_left, card_top, card_right, card_bottom])
|
||||
cards.sort(key=lambda b: b[1])
|
||||
return cards
|
||||
|
||||
# 转换为灰度图
|
||||
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
|
||||
|
||||
# 限制检测范围
|
||||
top_limit = int(h * top_ratio)
|
||||
bottom_limit = int(h * (1 - bottom_ratio))
|
||||
logger.info(f"CV检测卡片: h={h}, w={w}, top_limit={top_limit}, bottom_limit={bottom_limit}")
|
||||
|
||||
# 使用自适应阈值
|
||||
thresh = cv2.adaptiveThreshold(gray, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY_INV, 11, 2)
|
||||
|
||||
# 闭运算：连接断开的边缘
|
||||
# 减小水平核宽度，增加一点垂直连接，以适应卡片边缘不连续的情况
|
||||
kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (w // 4, 3))
|
||||
closed = cv2.morphologyEx(thresh, cv2.MORPH_CLOSE, kernel)
|
||||
|
||||
# 寻找轮廓
|
||||
contours, _ = cv2.findContours(closed, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
|
||||
|
||||
cards = []
|
||||
# 更严格的最小宽度要求：列表卡片通常接近整宽
|
||||
min_card_width = int(w * 0.8)
|
||||
|
||||
temp_boxes = []
|
||||
for cnt in contours:
|
||||
x, y, cw, ch = cv2.boundingRect(cnt)
|
||||
center_y = y + ch // 2
|
||||
|
||||
# 绿色营销券区域排除
|
||||
hsv = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
|
||||
roi = hsv[max(0,y):min(h,y+ch), max(0,x):min(w,x+cw)]
|
||||
green_mask = cv2.inRange(roi, np.array([35, 80, 80]), np.array([85, 255, 255]))
|
||||
green_ratio = float(cv2.countNonZero(green_mask)) / (roi.shape[0]*roi.shape[1]) if roi.size > 0 else 0.0
|
||||
|
||||
# 圆角矩形判断：面积占比与顶点数量
|
||||
rect_area = cw * ch
|
||||
cnt_area = cv2.contourArea(cnt)
|
||||
extent = cnt_area / rect_area if rect_area > 0 else 0
|
||||
approx = cv2.approxPolyDP(cnt, 0.02 * cv2.arcLength(cnt, True), True)
|
||||
|
||||
ok_width = cw >= min_card_width
|
||||
ok_height = ch > MIN_CARD_HEIGHT * 0.8
|
||||
ok_vertical = center_y >= int(h * 0.58) and y > top_limit and center_y < bottom_limit
|
||||
ok_shape = (len(approx) >= 6 or extent > 0.85)
|
||||
ok_color = green_ratio < 0.25
|
||||
|
||||
if ok_width and ok_height and ok_vertical and ok_shape and ok_color:
|
||||
logger.info(f"✅ 找到候选卡片: y={y}, h={ch}, w={cw}, extent={extent:.2f}, verts={len(approx)}, green={green_ratio:.2f}")
|
||||
temp_boxes.append((x, y, cw, ch))
|
||||
else:
|
||||
if ok_width:
|
||||
logger.debug(f"❌ 排除候选: y={y}, h={ch}, w={cw}, extent={extent:.2f}, verts={len(approx)}, green={green_ratio:.2f} (limit: {top_limit}-{bottom_limit})")
|
||||
|
||||
if not temp_boxes:
|
||||
logger.warning(f"⚠️ CV 未能在指定范围 ({top_limit}-{bottom_limit}) 内检测到任何场站卡片")
|
||||
|
||||
# 按 Y 轴排序
|
||||
temp_boxes.sort(key=lambda b: b[1])
|
||||
|
||||
# 再次过滤和去重，并根据高度拆分为多张卡片
|
||||
for i, box in enumerate(temp_boxes):
|
||||
x, y, cw, ch = box
|
||||
|
||||
# 检查是否与已有的框重叠（按顶部 y 去重）
|
||||
is_duplicate = False
|
||||
for v in cards:
|
||||
if abs(y - v[1]) < 100:
|
||||
is_duplicate = True
|
||||
break
|
||||
if is_duplicate:
|
||||
continue
|
||||
|
||||
padding = 2
|
||||
x1 = max(0, x + padding)
|
||||
x2 = min(w, x + cw - padding)
|
||||
y1 = max(top_limit, y + padding)
|
||||
y2 = min(bottom_limit, y + ch - padding)
|
||||
|
||||
if y2 <= y1:
|
||||
continue
|
||||
|
||||
effective_h = y2 - y1
|
||||
estimated_count = max(1, int(round(effective_h / float(MIN_CARD_HEIGHT))))
|
||||
|
||||
if estimated_count <= 1:
|
||||
cards.append([x1, y1, x2, y2])
|
||||
else:
|
||||
seg_h = effective_h / float(estimated_count)
|
||||
for k in range(estimated_count):
|
||||
sy1 = int(round(y1 + k * seg_h))
|
||||
sy2 = int(round(y1 + (k + 1) * seg_h))
|
||||
if sy2 <= sy1:
|
||||
continue
|
||||
if sy2 - sy1 < MIN_CARD_HEIGHT * 0.6:
|
||||
continue
|
||||
cards.append([x1, sy1, x2, sy2])
|
||||
|
||||
return cards
|
||||
|
||||
def detect_wide_rounded_card_cv(image_path, min_width_ratio=0.8, min_y_ratio=0.5):
|
||||
img = read_image(image_path)
|
||||
if img is None:
|
||||
return []
|
||||
h, w = img.shape[:2]
|
||||
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
|
||||
blur = cv2.GaussianBlur(gray, (7, 7), 0)
|
||||
edges = cv2.Canny(blur, 60, 180)
|
||||
kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (15, 15))
|
||||
closed = cv2.morphologyEx(edges, cv2.MORPH_CLOSE, kernel)
|
||||
contours, _ = cv2.findContours(closed, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
|
||||
results = []
|
||||
for cnt in contours:
|
||||
x, y, cw, ch = cv2.boundingRect(cnt)
|
||||
center_y = y + ch // 2
|
||||
if cw < int(w * min_width_ratio):
|
||||
continue
|
||||
if center_y < int(h * min_y_ratio):
|
||||
continue
|
||||
rect_area = cw * ch
|
||||
cnt_area = cv2.contourArea(cnt)
|
||||
if rect_area <= 0:
|
||||
continue
|
||||
extent = cnt_area / rect_area
|
||||
approx = cv2.approxPolyDP(cnt, 0.02 * cv2.arcLength(cnt, True), True)
|
||||
if extent < 0.7:
|
||||
continue
|
||||
if len(approx) >= 6 or (len(approx) == 4 and extent > 0.85):
|
||||
results.append([x, y, x + cw, y + ch])
|
||||
results.sort(key=lambda b: b[1])
|
||||
return results
|
||||
|
||||
def draw_rectangles(image_path, bboxes=None, click_points=None):
|
||||
"""
|
||||
使用 OpenCV 在图片上绘制矩形框和点击点，生成 _vl.jpg 和 _flag.jpg
|
||||
|
||||
Reference in New Issue
Block a user