pip install opencv-python numpy 

import numpy as np
import cv2
 
def order_points(pts):
    # 初始化4個頂點的坐標（左上、右上、右下、左下）
    rect = np.zeros((4, 2), dtype="float32")
    
    # 1. 計算每個點的x+y之和：左上點的和最小，右下點的和最大
    s = pts.sum(axis=1)
    rect[0] = pts[np.argmin(s)]  # 左上（tl）
    rect[2] = pts[np.argmax(s)]  # 右下（br）
    
    # 2. 計算每個點的y-x之差（np.diff默認是后減前，即y-x）：右上點的差最小，左下點的差最大
    diff = np.diff(pts, axis=1)
    rect[1] = pts[np.argmin(diff)]  # 右上（tr）
    rect[3] = pts[np.argmax(diff)]  # 左下（bl）
    
    return rect

def four_point_transform(image, pts):
    # 1. 獲取排序后的四個頂點
    rect = order_points(pts)
    (tl, tr, br, bl) = rect  # 左上、右上、右下、左下
    
    # 2. 計算目標矩形的寬度（取文檔左右兩邊的最大長度，避免變形）
    widthA = np.sqrt(((br[0] - bl[0]) ** 2) + ((br[1] - bl[1]) ** 2))  # 下邊長
    widthB = np.sqrt(((tr[0] - tl[0]) ** 2) + ((tr[1] - tl[1]) ** 2))  # 上邊長
    maxWidth = max(int(widthA), int(widthB))  # 目標寬度
    
    # 3. 計算目標矩形的高度（取文檔上下兩邊的最大長度）
    heightA = np.sqrt(((tr[0] - br[0]) ** 2) + ((tr[1] - br[1]) ** 2))  # 右邊長
    heightB = np.sqrt(((tl[0] - bl[0]) ** 2) + ((tl[1] - bl[1]) ** 2))  # 左邊長
    maxHeight = max(int(heightA), int(heightB))  # 目標高度
    
    # 4. 定義目標矩形的四個頂點
    dst = np.array([
        [0, 0],                  # 左上
        [maxWidth - 1, 0],       # 右上（減1是因為像素索引從0開始）
        [maxWidth - 1, maxHeight - 1],  # 右下
        [0, maxHeight - 1]], dtype="float32")  # 左下
    
    # 5. 計算透視矩陣M
    M = cv2.getPerspectiveTransform(rect, dst)
    
    # 6. 應用透視變換，得到矯正后的圖像
    warped = cv2.warpPerspective(image, M, (maxWidth, maxHeight))
    
    return warped

# 圖像預處理
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)  # 灰度化
gray = cv2.GaussianBlur(gray, (5, 5), 0)  # 高斯模糊（5x5卷積核，標準差0）
edged = cv2.Canny(gray, 15, 45)  # 邊緣檢測（低閾值15，高閾值45）
 
# 輪廓檢測與篩選
cnts = cv2.findContours(edged, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)[-2]  # 提取外部輪廓
cnts = sorted(cnts, key=cv2.contourArea, reverse=True)[:3]  # 按面積降序，取前3個
 
# 遍歷輪廓，判斷是否為文檔（四邊形+面積足夠大）
for c in cnts:
    peri = cv2.arcLength(c, True)  # 計算輪廓周長（True表示輪廓閉合）
    # 多邊形逼近：epsilon=0.05*peri（逼近精度，值越小越接近原輪廓）
    approx = cv2.approxPolyDP(c, 0.05 * peri, True)
    area = cv2.contourArea(approx)  # 計算逼近后多邊形的面積
    
    # 條件：面積>20000（過濾小物體）且頂點數=4（文檔為四邊形）
    if area > 20000 and len(approx) == 4:
        screenCnt = approx  # 確定文檔輪廓
        print("檢測到文檔，輪廓面積：", area)
        break

# 對透視變換后的圖像進行二值化
warped_gray = cv2.cvtColor(warped_result, cv2.COLOR_BGR2GRAY)
# 自動閾值二值化（THRESH_OTSU：自動計算最優(yōu)閾值，避免手動調參）
ref_result = cv2.threshold(warped_gray, 0, 255, cv2.THRESH_BINARY | cv2.THRESH_OTSU)[1]

# 自定義顯示函數（不自動關閉窗口，需按q退出）
def cv_show(name, img):
    cv2.imshow(name, img)
 
# 初始化攝像頭（0表示默認攝像頭）
cap = cv2.VideoCapture(0)
 
# 檢查攝像頭是否正常打開
if not cap.isOpened():
    print("無法打開攝像頭")
    exit()
 
# 實時采集與處理循環(huán)
while True:
    ret, image = cap.read()  # 讀取一幀圖像（ret：是否讀取成功，image：圖像數據）
    orig = image.copy()  # 保存原始圖像副本
    
    if not ret:
        print("無法讀取攝像頭幀")
        break
    
    # 1. 顯示原始圖像
    cv_show("Original", image)
    
    # 2. 圖像預處理與輪廓檢測（此處省略，見模塊3）
    # ...（預處理、輪廓檢測代碼）...
    
    # 3. 若檢測到文檔，顯示結果
    if flag == 1:
        # 繪制文檔輪廓
        image_with_doc = cv2.drawContours(orig.copy(), [screenCnt], 0, (0, 255, 0), 2)
        cv_show("Document Detection", image_with_doc)
        
        # 透視變換
        warped_result = four_point_transform(orig, screenCnt.reshape(4, 2))
        cv_show("Warped", warped_result)
        
        # 二值化
        ref_result = cv2.threshold(cv2.cvtColor(warped_result, cv2.COLOR_BGR2GRAY), 
                                   0, 255, cv2.THRESH_BINARY | cv2.THRESH_OTSU)[1]
        cv_show("Binarized", ref_result)
    
    # 按下'q'鍵退出循環(huán)（waitKey(1)：等待1ms，返回按鍵ASCII碼）
    if cv2.waitKey(1) == ord('q'):
        break
 
# 釋放攝像頭資源，關閉所有窗口
cap.release()
cv2.destroyAllWindows()

import numpy as np
import cv2
 
 
def order_points(pts):
    rect = np.zeros((4, 2), dtype="float32")
    s = pts.sum(axis=1)
    rect[0] = pts[np.argmin(s)]
    rect[2] = pts[np.argmax(s)]
    diff = np.diff(pts, axis=1)
    rect[1] = pts[np.argmin(diff)]
    rect[3] = pts[np.argmax(diff)]
    return rect
 
 
def four_point_transform(image, pts):
    rect = order_points(pts)
    (tl, tr, br, bl) = rect
    widthA = np.sqrt(((br[0] - bl[0]) ** 2) + ((br[1] - bl[1]) ** 2))
    widthB = np.sqrt(((tr[0] - tl[0]) ** 2) + ((tr[1] - tl[1]) ** 2))
    maxWidth = max(int(widthA), int(widthB))
    heightA = np.sqrt(((tr[0] - br[0]) ** 2) + ((tr[1] - br[1]) ** 2))
    heightB = np.sqrt(((tl[0] - bl[0]) ** 2) + ((tl[1] - bl[1]) ** 2))
    maxHeight = max(int(heightA), int(heightB))
    dst = np.array([
        [0, 0],
        [maxWidth - 1, 0],
        [maxWidth - 1, maxHeight - 1],
        [0, maxHeight - 1]], dtype="float32")
    M = cv2.getPerspectiveTransform(rect, dst)
    warped = cv2.warpPerspective(image, M, (maxWidth, maxHeight))
    return warped
 
 
def cv_show(name, img):
    cv2.imshow(name, img)
 
 
if __name__ == "__main__":
    cap = cv2.VideoCapture(0)
    if not cap.isOpened():
        print("Cannot open camera")
        exit()
 
    while True:
        flag = 0
        ret, image = cap.read()
        orig = image.copy()
        if not ret:
            print("不能讀取攝像頭")
            break
 
        # 顯示原始圖像
        cv_show("Original", image)
 
        # 圖像預處理
        gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
        gray = cv2.GaussianBlur(gray, (5, 5), 0)
        edged = cv2.Canny(gray, 15, 45)
        cv_show("Edge Detection", edged)
 
        # 輪廓檢測與篩選
        cnts = cv2.findContours(edged, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)[-2]
        cnts = sorted(cnts, key=cv2.contourArea, reverse=True)[:3]
        image_contours = cv2.drawContours(image.copy(), cnts, -1, (0, 255, 0), 2)
        cv_show("Contours", image_contours)
 
        # 識別文檔輪廓
        for c in cnts:
            peri = cv2.arcLength(c, True)
            approx = cv2.approxPolyDP(c, 0.05 * peri, True)
            area = cv2.contourArea(approx)
 
            if area > 20000 and len(approx) == 4:
                screenCnt = approx
                flag = 1
                print(f"檢測到文檔，周長：{peri:.2f}，面積：{area:.2f}")
 
                # 顯示文檔檢測結果
                image_with_doc = cv2.drawContours(orig.copy(), [screenCnt], 0, (0, 255, 0), 2)
                cv_show("Document Detection", image_with_doc)
 
                # 透視變換
                warped_result = four_point_transform(orig, screenCnt.reshape(4, 2))
                cv_show("Warped", warped_result)
 
                # 二值化
                warped_gray = cv2.cvtColor(warped_result, cv2.COLOR_BGR2GRAY)
                ref_result = cv2