2

backend/txm/app/image_processing.py

@@ -0,0 +1,84 @@
+from typing import List, Optional, Tuple
+
+import cv2
+import numpy as np
+
+
+def resize_keep_aspect(image: np.ndarray, target_width: int) -> np.ndarray:
+    if target_width <= 0:
+        return image
+    h, w = image.shape[:2]
+    if w == target_width:
+        return image
+    scale = target_width / float(w)
+    new_size = (target_width, int(round(h * scale)))
+    return cv2.resize(image, new_size, interpolation=cv2.INTER_AREA)
+
+
+def enhance_barcode_stripes(gray: np.ndarray, gaussian_ksize: int, sobel_ksize: int) -> np.ndarray:
+    g = gray
+    if gaussian_ksize and gaussian_ksize > 1:
+        g = cv2.GaussianBlur(g, (gaussian_ksize, gaussian_ksize), 0)
+    # 使用水平 Sobel 捕捉垂直边缘（条纹）
+    grad_x = cv2.Sobel(g, cv2.CV_32F, 1, 0, ksize=sobel_ksize)
+    grad_x = cv2.convertScaleAbs(grad_x)
+    return grad_x
+
+
+def binarize_image(img: np.ndarray, method: str = "otsu") -> np.ndarray:
+    if method == "adaptive":
+        return cv2.adaptiveThreshold(
+            img, 255, cv2.ADAPTIVE_THRESH_MEAN_C, cv2.THRESH_BINARY, 31, 10
+        )
+    # 默认 OTSU
+    _, th = cv2.threshold(img, 0, 255, cv2.THRESH_BINARY | cv2.THRESH_OTSU)
+    return th
+
+
+def morph_close(img: np.ndarray, kernel_size: int) -> np.ndarray:
+    if kernel_size <= 1:
+        return img
+    kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (kernel_size, kernel_size))
+    closed = cv2.morphologyEx(img, cv2.MORPH_CLOSE, kernel)
+    return closed
+
+
+def find_barcode_roi(binary: np.ndarray, original_shape: Tuple[int, int], min_area_ratio: float, min_wh_ratio: float) -> Optional[Tuple[np.ndarray, Tuple[int, int, int, int]]]:
+    h, w = original_shape
+    contours, _ = cv2.findContours(binary, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
+    image_area = h * w
+    candidates: List[Tuple[float, Tuple[int, int, int, int]]] = []
+    for cnt in contours:
+        x, y, cw, ch = cv2.boundingRect(cnt)
+        area = cw * ch
+        if area / image_area < min_area_ratio:
+            continue
+        wh_ratio = cw / float(ch + 1e-6)
+        if wh_ratio < min_wh_ratio:
+            continue
+        candidates.append((area, (x, y, cw, ch)))
+    if not candidates:
+        return None
+    candidates.sort(key=lambda t: t[0], reverse=True)
+    _, bbox = candidates[0]
+    x, y, cw, ch = bbox
+    roi = binary[y : y + ch, x : x + cw]
+    return roi, bbox
+
+
+def warp_barcode_region(gray: np.ndarray, bbox: Tuple[int, int, int, int], target_height: int, crop_bottom_ratio: float = 0.0) -> np.ndarray:
+    x, y, cw, ch = bbox
+    crop = gray[y : y + ch, x : x + cw]
+    # 去除底部数字区域干扰
+    if 0 < crop_bottom_ratio < 1:
+        hb = int(round(ch * (1.0 - crop_bottom_ratio)))
+        hb = max(10, min(ch, hb))
+        crop = crop[:hb, :]
+    if target_height <= 0:
+        return crop
+    scale = target_height / float(ch)
+    target_width = int(round(cw * scale))
+    warped = cv2.resize(crop, (target_width, target_height), interpolation=cv2.INTER_CUBIC)
+    return warped
+
+