import cv2 # OpenCV库，用于图像处理
import numpy as np # 数值计算库
from collections import defaultdict # 提供带默认值的字典
import shutil # 文件操作库
import pyttsx3 # 本地语音合成库
from datetime import datetime # 日期时间处理
import logging # 日志处理库
import threading # 线程库
import queue # 队列库
import json # JSON处理库
from pathlib import Path # 路径处理库
from typing import List # 类型提示
import time # 时间处理库
import os # 操作系统相关操作
import sys # 系统相关操作
from flask import Flask, render_template # Flask框架及模板渲染
import geocoder # 地理位置获取模块
from .weather.weather_service import WeatherService # 导入天气服务模块
from .weather.config import * # 如有需要可导入配置
from .weather.api_client import * # 如有需要可导入API客户端
from .weather.exception import * # 如有需要可导入异常定义
from .utils import read_text_baidu, user_speech_recognition, record_audio_until_silence, audio_to_text, text_to_speech_chinese, load_known_faces_from_folder # 导入通用工具函数
from .voice_feature import VoiceAssistant # 导入语音助手模块
from .face_recognition import FaceRecognition # 导入人脸识别系统
import RPi.GPIO as GPIO # 导入树莓派的GPIO库，用于HC-SR501传感器
# 初始化Flask应用
app = Flask(__name__)
app.config['TEMPLATES_AUTO_RELOAD'] = True # 设置模板自动重载
# 全局运行标志，控制主循环0
running = True # 控制主循环是否继续运行
face_detected = False # 标记是否检测到人脸
wake_words = ["小", "小朋友", "朋友"] # 唤醒词列表
WAKE_WORD_THRESHOLD = 0.7 # 唤醒词识别阈值
assistant = None # 全局语音助手对象
face_detection_running = False # 标记人脸检测线程是否运行中
face_detection_success = False # 标记人脸检测是否成功
pir_detected = False # 标记是否检测到人靠近
# HC-SR501传感器设置
PIR_PIN = 17 # 假设HC-SR501连接到GPIO 17
GPIO.setmode(GPIO.BCM) # 使用BCM编号
GPIO.setup(PIR_PIN, GPIO.IN) # 设置PIR_PIN为输入
# 定义Flask路由，主页
@app.route('/')
def index():
"""渲染主界面页面"""
return render_template('index.html')
# 定义Flask路由，测试接口
@app.route('/test')
def test():
"""测试接口，验证Flask服务是否正常"""
return "Flask server is working!"
preloaded_face_data = None # 全局变量，存储预加载的人脸数据
def preload_face_data():
"""预加载人脸数据到内存，提升识别速度"""
face_system = FaceRecognition() # 创建人脸识别系统对象
image_paths_by_person = load_known_faces_from_folder("known_faces") # 加载已知人脸图片路径
for person_name, image_paths in image_paths_by_person.items(): # 遍历每个人
face_system.add_new_person(person_name, image_paths) # 添加人脸特征
return face_system # 返回人脸识别系统对象
def detect_face(timeout=60) -> bool:
"""使用预加载的人脸数据检测人脸，超时返回False"""
global preloaded_face_data
if preloaded_face_data is None: # 检查是否已预加载
print("Error: Face data not preloaded.")
return False
start_time = time.time() # 记录检测开始时间
while time.time() - start_time < timeout and running: # 在超时时间内循环检测
try:
recognized = preloaded_face_data.start_recognition() # 调用识别方法
if recognized: # 如果识别到人脸
print("Face detected")
return True
else:
print("No face detected (during activation)")
time.sleep(0.5) # 未识别到，等待0.5秒再试
except Exception as e:
print(f"Error during recognition: {e}") # 捕获异常并打印
time.sleep(0.5)
return False # 超时未识别到人脸
def get_user_location():
"""通过geocoder获取用户地理位置"""
try:
return geocoder.ip("me") # 获取本机IP对应的地理位置
except Exception as e:
print(f"Error getting location: {e}")
return None
def assistant_mode():
"""语音助手主循环，处理用户语音指令"""
global running, face_detected, assistant
if assistant is None: # 检查语音助手是否初始化
print("Error: Assistant not initialized.")
return
location = get_user_location() # 获取用户地理位置
weather = WeatherService() # 创建天气服务对象
longitude = location.lng if location else None # 获取经度
latitude = location.lat if location else None # 获取纬度
response = weather.get_weather_info(longitude, latitude) if longitude and latitude else {} # 获取天气信息
listening_duration = 60 # 监听超时时间（秒）
last_interaction_time = time.time() # 上次交互时间
read_text_baidu("你好！有什么我可以帮您？") # 语音播报欢迎语
while running and face_detected: # 只要系统运行且检测到人脸
if time.time() - last_interaction_time > listening_duration: # 超时未交互
read_text_baidu("等待唤醒...")
face_detected = False
break
text = user_speech_recognition() # 获取用户语音输入
if text: # 如果识别到内容
print(f"User said: {text}")
last_interaction_time = time.time() # 重置交互时间
if '天气' in text and response: # 天气查询
print("Weather query detected")
print("Response: " + response.get('weather_condition', ''))
read_text_baidu(f"今天的天气状况如下, 位置:{location.city if location else ''}")
read_text_baidu(f"天气：{response.get('weather_condition', '')}")
read_text_baidu(f"温度：{response.get('temperature', '')}")
read_text_baidu(f"体感温度：{response.get('feels_like', '')}")
read_text_baidu(f"湿度：{response.get('humidity', '')}")
read_text_baidu(f"风向：{response.get('wind_direction', '')}")
read_text_baidu(f"风速：{response.get('wind_speed', '')}")
read_text_baidu(f"气压：{response.get('pressure', '')}")
read_text_baidu(f"能见度：{response.get('visibility', '')}")
read_text_baidu(f"云量：{response.get('cloud_coverage', '')}")
continue
elif '几点' in text: # 时间查询
print("Time query detected")
read_text_baidu(f"现在是 {time.strftime('%H:%M')}")
continue
elif '空调' in text: # 空调控制
print("AC query detected")
read_text_baidu("好的，正在处理空调指令。")
continue
elif '拜拜' in text or '再见' in text: # 结束对话
read_text_baidu("拜拜，下次再见！")
face_detected = False
continue
else: # 其他内容交给大模型处理
print("Deepseek request")
print(f"Heard: {text}, processing with DeepSeek...")
response = assistant.chat(text) # 调用大模型
print(f"DeepSeek response: {response}")
read_text_baidu(response)
continue
else:
print("Listening for command...") # 未识别到内容，继续监听
time.sleep(1)
if face_detected: # 如果还在检测状态，提示等待唤醒
read_text_baidu("等待唤醒...")
face_detected = False
def run_face_detection():
"""运行人脸检测线程，检测成功则标记，否则语音提示失败"""
global face_detection_running, face_detection_success
face_detection_running = True # 标记检测线程已启动
if detect_face(timeout=60): # 检测人脸
face_detection_success = True # 检测成功
else:
text_to_speech_chinese("我无法识别您的面部。如果需要我，请随时叫我。") # 检测失败语音提示
face_detection_running = False # 检测线程结束
def pir_detection_loop():
"""HC-SR501传感器检测主循环，检测到人靠近后启动人脸检测和助手"""
global running, face_detected, assistant
global face_detection_running, face_detection_success, pir_detected
if assistant is None: # 检查语音助手是否初始化
print("Error: Assistant not initialized.")
return
while running: # 主循环
if GPIO.input(PIR_PIN): # 检测到人靠近
print("PIR sensor detected motion")
pir_detected = True
if not face_detection_running: # 如果没有人脸检测线程在运行
text_to_speech_chinese("检测到人靠近，请靠近并扫描您的面部以继续互动。这是为了您的安全。")
threading.Thread(target=run_face_detection).start() # 启动人脸检测线程
else:
pir_detected = False
if face_detection_success: # 如果人脸检测成功，启动助手线程
face_detected = True
face_detection_success = False # 重置标志
assistant_thread = threading.Thread(target=assistant_mode)
assistant_thread.start()
time.sleep(0.1) # 防止CPU占用过高
def wake_word_detection_loop():
"""唤醒词检测主循环，检测到唤醒词后启动人脸检测和助手"""
global running, face_detected, assistant
global face_detection_running, face_detection_success
if assistant is None: # 检查语音助手是否初始化
print("Error: Assistant not initialized.")
return
while running: # 主循环
if face_detected: # 如果已检测到人脸，等待
time.sleep(2)
continue
if face_detection_success: # 如果人脸检测成功，启动助手线程
face_detected = True
face_detection_success = False # 重置标志
assistant_thread = threading.Thread(target=assistant_mode)
assistant_thread.start()
continue
if not face_detection_running: # 如果没有人脸检测线程在运行
print("Listening for wake word...") # 等待唤醒词
script = audio_to_text() # 语音转文本
if script:
for wake_word in wake_words: # 检查是否包含唤醒词
if wake_word in script:
text_to_speech_chinese("唤醒成功，请靠近并扫描您的面部以继续互动。这是为了您的安全。")
threading.Thread(target=run_face_detection).start() # 启动人脸检测线程
break
else:
print("Wake word not detected.") # 未检测到唤醒词
else:
print("No speech detected.") # 未检测到语音
time.sleep(1) # 防止CPU占用过高
def launch_gui():
"""启动图形界面，优先PyQt5，失败则尝试Kivy，再失败则控制台模式"""
try:
from PyQt5.QtWebEngineWidgets import QWebEngineView
from PyQt5.QtWidgets import QApplication
from PyQt5.QtCore import QUrl
import sys
app = QApplication(sys.argv)
web = QWebEngineView()
web.load(QUrl("http://localhost:8080"))
web.show()
sys.exit(app.exec_())
except ImportError:
try:
from kivy.app import App
from kivy.uix.label import Label
class SimpleApp(App):
def build(self):
return Label(text='Display Error, your Smart Mirror is Running\nhttp://localhost:8080')
SimpleApp().run()
except ImportError:
print("GUI frameworks not available - running in console mode")
print("Access the mirror at http://localhost:8080")
import time
while True:
time.sleep(1)
def main() -> None:
"""主入口，初始化各模块并启动主线程"""
global running, assistant, preloaded_face_data
print("Smart Mirror started.")
assistant = VoiceAssistant(
baidu_app_id=os.getenv("BAIDU_APP_ID"),
baidu_api_key=os.getenv("BAIDU_API_KEY"),
baidu_secret_key=os.getenv("BAIDU_SECRET_KEY"),
deepseek_api_key=os.getenv("DEEPSEEK_API_KEY"),
)
preloaded_face_data = preload_face_data()
gui_thread = threading.Thread(target=launch_gui) # 启动GUI线程
pir_thread = threading.Thread(target=pir_detection_loop,daemon=True) # 启动
voice_thread = threading.Thread(target=wake_word_detection_loop, daemon=True)
pir_thread.start()
voice_thread.start()
gui_thread.start()
try:
while True: # 主线程保持运行
time.sleep(1)
except KeyboardInterrupt:
print("Exiting...")
running = False # 终止主循环
pir_thread.join()
voice_thread.join()
GPIO.cleanup() # 清理GPIO设置
if __name__ == "__main__":
main() # 运行主入口这段代码是否能完成人体靠近时检测到并进行人脸识别，同时不和语音的识别冲突
本文深入探讨了人工智能算法标注的基本原则、流程与实践技巧，强调了高质量数据集对于训练高效AI模型的重要性。从数据预处理到标注规范，再到质量控制，全面解析了AI算法标注的各个环节。
扫一扫
1489
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
