第五章：实战项目案例

---

5.1 项目一：车辆检测与计数

5.1.1 项目描述

利用视频流中的车辆检测，统计通过某一区域的车辆数量。

5.1.2 环境准备

• OpenCV
• 预训练YOLOv4模型

1. 加载模型与参数设置

python

import cv2
import numpy as np

# 加载YOLOv4模型
net = cv2.dnn.readNet("yolov4.weights", "yolov4.cfg")
layer_names = net.getUnconnectedOutLayersNames()

# 加载类别名称
with open("coco.names", "r") as f:
    classes = f.read().strip().split("\n")

---

2. 目标检测与区域计数

python

cap = cv2.VideoCapture("traffic.mp4")
while cap.isOpened():
    ret, frame = cap.read()
    if not ret:
        break

    # 创建图像Blob
    blob = cv2.dnn.blobFromImage(frame, scalefactor=1/255.0, size=(416, 416), swapRB=True)
    net.setInput(blob)
    outputs = net.forward(layer_names)

    # 检测并绘制边界框
    for output in outputs:
        for detection in output:
            scores = detection[5:]
            class_id = np.argmax(scores)
            confidence = scores[class_id]

            if confidence > 0.5 and classes[class_id] == "car":
                center_x, center_y, width, height = (detection[:4] * np.array([w, h, w, h])).astype("int")
                x, y = int(center_x - width / 2), int(center_y - height / 2)
                cv2.rectangle(frame, (x, y), (x + width, y + height), (0, 255, 0), 2)

    cv2.imshow("Vehicle Detection", frame)
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break
cap.release()
cv2.destroyAllWindows()

---

5.2 项目二：图像拼接与全景图生成

5.2.1 项目描述

利用多张重叠图像，生成一幅完整的全景图。

1. 读取输入图像

python

images = [cv2.imread(f"image_{i}.jpg") for i in range(1, 4)]

2. 检测与匹配特征点

python

# 使用SIFT检测特征
sift = cv2.SIFT_create()
keypoints, descriptors = zip(*[sift.detectAndCompute(img, None) for img in images])

# 匹配特征点
bf = cv2.BFMatcher()
matches = [bf.knnMatch(descriptors[i], descriptors[i+1], k=2) for i in range(len(images) - 1)]

# 筛选匹配点
good_matches = []
for m, n in matches[0]:
    if m.distance < 0.75 * n.distance:
        good_matches.append(m)

3. 计算单应性并拼接图像

python

src_pts = np.float32([keypoints[0][m.queryIdx].pt for m in good_matches]).reshape(-1, 1, 2)
dst_pts = np.float32([keypoints[1][m.trainIdx].pt for m in good_matches]).reshape(-1, 1, 2)

H, _ = cv2.findHomography(src_pts, dst_pts, cv2.RANSAC, 5.0)
result = cv2.warpPerspective(images[0], H, (width, height))
cv2.imshow("Panorama", result)
cv2.waitKey(0)
cv2.destroyAllWindows()

---

5.3 项目三：实时手势识别

5.3.1 项目描述

通过摄像头捕获视频流，实时检测手势类型。

5.3.2 具体实现

1. 使用MediaPipe识别手部

python

import mediapipe as mp

mp_hands = mp.solutions.hands
hands = mp_hands.Hands(min_detection_confidence=0.7, min_tracking_confidence=0.7)
mp_draw = mp.solutions.drawing_utils

cap = cv2.VideoCapture(0)
while cap.isOpened():
    ret, frame = cap.read()
    if not ret:
        break

    frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
    results = hands.process(frame_rgb)

    if results.multi_hand_landmarks:
        for hand_landmarks in results.multi_hand_landmarks:
            mp_draw.draw_landmarks(frame, hand_landmarks, mp_hands.HAND_CONNECTIONS)

    cv2.imshow("Hand Gesture Recognition", frame)
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break
cap.release()
cv2.destroyAllWindows()