feat: 检测算法开关设置 + AI大模型分析开关设置

config.py

@@ -27,6 +27,17 @@ DEFAULT_CONFIG = {
     "auto_analyze": True,
     "refresh_interval": 5,  # 页面刷新间隔（秒）
     "display_limit": 20,    # 显示最近多少条
+    
+    # 检测算法开关
+    "use_haar_cascade": True,       # Haar Cascade 人体检测
+    "use_mediapipe_face": True,      # MediaPipe 人脸检测
+    "use_face_recognition": True,    # face_recognition 人脸识别
+    
+    # AI大模型分析开关
+    "use_vision_api": False,         # 是否使用大模型分析（默认关闭）
+    "vision_api_trigger": "person_change",  # 触发条件：person_change/motion/brightness/always
+    
+    # Vision API 配置
     "vision_api_url": os.environ.get("VISION_API_URL", "https://dashscope.aliyuncs.com/compatible-mode/v1"),
     "vision_api_key": os.environ.get("VISION_API_KEY", "sk-lm-fuP5tGU8:Hi7YU87jHyDP6Ay8Tl2j"),
     "vision_model": os.environ.get("VISION_MODEL", "qwen-vl-plus")

local_analyzer.py

@@ -19,6 +19,8 @@ import os
 # 添加项目路径
 sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
 
+from config import config_mgr
+
 try:
     from person_manager import person_manager
     HAS_PERSON_MANAGER = True
@@ -26,6 +28,12 @@ except ImportError:
     HAS_PERSON_MANAGER = False
     print("[LocalAnalyzer] PersonManager not available")
 
+try:
+    import mediapipe as mp
+    HAS_MEDIAPIPE = True
+except ImportError:
+    HAS_MEDIAPIPE = False
+
 
 class LocalAnalyzer:
     """本地视觉分析器"""
@@ -37,7 +45,7 @@ class LocalAnalyzer:
         self.prev_human_count = 0  # 记录前一帧人数
         
         # 初始化人体检测器
-        self._init_human_detector()
+        self._init_detectors()
         
         # 阈值配置
         self.config = {
@@ -53,22 +61,22 @@ class LocalAnalyzer:
         self.frame_count = 0
         self.motion_count = 0
         self.human_count = 0
+        self.person_change_count = 0  # 人员变化次数
     
-    def _init_human_detector(self):
+    def _init_detectors(self):
-        """初始化人体检测器"""
+        """初始化检测器"""
+        # Haar Cascade 人体检测
         try:
-            # 使用 OpenCV 内置的 HOG 人体检测器
-            self.hog = cv2.HOGDescriptor()
-            self.hog.setSVMDetector(cv2.HOGDescriptor_getDefaultPeopleDetector())
-            
-            # 或者使用 Haar Cascade（更快但精度稍低）
             cascade_path = cv2.data.haarcascades + 'haarcascade_fullbody.xml'
             if Path(cascade_path).exists():
                 self.human_cascade = cv2.CascadeClassifier(cascade_path)
-            
+                print("[LocalAnalyzer] Haar Cascade body detector initialized")
-            print("[LocalAnalyzer] Human detector initialized")
         except Exception as e:
-            print(f"[LocalAnalyzer] Human detector init failed: {e}")
+            print(f"[LocalAnalyzer] Haar Cascade init failed: {e}")
+        
+        # MediaPipe 人脸检测状态（由 PersonManager 管理）
+        if HAS_PERSON_MANAGER:
+            print("[LocalAnalyzer] PersonManager available (MediaPipe face detection)")
         
     def analyze(self, image_path, prev_image_path=None):
         """
@@ -118,20 +126,26 @@ class LocalAnalyzer:
             
             # 2. 人员检测与识别
             person_result = {'persons': [], 'total_count': 0, 'new_count': 0, 'known_count': 0}
+            haar_result = {'count': 0, 'positions': []}
             
-            if HAS_PERSON_MANAGER:
+            # 获取检测算法配置
-                print(f"[LocalAnalyzer] Using PersonManager for face detection...")
+            use_haar = config_mgr.get('use_haar_cascade', True)
+            use_mediapipe = config_mgr.get('use_mediapipe_face', True)
+            
+            # 方法1：MediaPipe 人脸检测 + 人员识别（优先）
+            if HAS_PERSON_MANAGER and use_mediapipe:
+                print(f"[LocalAnalyzer] Using MediaPipe face detection...")
                 person_result = person_manager.analyze_image(image_path, save_new_person=True)
                 
                 metrics['person_count'] = person_result['total_count']
                 metrics['new_persons'] = person_result['new_count']
                 metrics['known_persons'] = person_result['known_count']
                 
-                # 记录人员变化
+                prev_person_count = self.prev_human_count
-                prev_person_count = self.prev_human_count  # 用之前的变量名
                 person_count_change = person_result['total_count'] - prev_person_count
                 metrics['person_count_change'] = person_count_change
                 
+                # 记录人员事件
                 for person in person_result['persons']:
                     if person['is_new']:
                         events.append({
@@ -141,10 +155,11 @@ class LocalAnalyzer:
                             'source': 'local'
                         })
                         self.human_count += 1
+                        self.person_change_count += 1
                     else:
                         events.append({
                             'event_type': '人物活动',
-                            'description': f'已知人员: {person["name"]}，已访问 {person_manager.persons.get(person["person_id"], {}).get("visit_count", 1)} 次',
+                            'description': f'已知人员: {person["name"]}',
                             'confidence': '高',
                             'source': 'local'
                         })
@@ -157,6 +172,7 @@ class LocalAnalyzer:
                         'confidence': '高',
                         'source': 'local'
                     })
+                    self.person_change_count += 1
                 elif person_count_change < 0:
                     events.append({
                         'event_type': '人员进出',
@@ -164,42 +180,44 @@ class LocalAnalyzer:
                         'confidence': '高',
                         'source': 'local'
                     })
+                    self.person_change_count += 1
                 
-                # 更新前一帧人数
                 self.prev_human_count = person_result['total_count']
             
-            else:
+            # 方法2：Haar Cascade 人体检测（备用或并行）
-                # 使用传统人体检测（备用）
+            if use_haar and self.human_cascade is not None:
-                human_result = self._detect_human(current_frame)
+                print(f"[LocalAnalyzer] Using Haar Cascade body detection...")
-                metrics['human_count'] = human_result['count']
+                haar_result = self._detect_human(current_frame)
                 
-                human_count_change = human_result['count'] - self.prev_human_count
+                # 如果 MediaPipe 没检测到，用 Haar Cascade 结果
-                metrics['human_count_change'] = human_count_change
+                if not (HAS_PERSON_MANAGER and use_mediapipe):
+                    metrics['human_count'] = haar_result['count']
                     
-                if human_count_change > 0:
+                    human_count_change = haar_result['count'] - self.prev_human_count
-                    events.append({
+                    metrics['human_count_change'] = human_count_change
-                        'event_type': '人物活动',
-                        'description': f'检测到 {human_count_change} 人进入，当前共 {human_result["count"]} 人',
-                        'confidence': '高',
-                        'source': 'local'
-                    })
-                    self.human_count += human_count_change
-                elif human_count_change < 0:
-                    events.append({
-                        'event_type': '人物活动',
-                        'description': f'检测到 {abs(human_count_change)} 人离开，当前剩 {human_result["count"]} 人',
-                        'confidence': '高',
-                        'source': 'local'
-                    })
-                elif human_result['count'] > 0:
-                    events.append({
-                        'event_type': '人物活动',
-                        'description': f'检测到 {human_result["count"]} 个人（无变化）',
-                        'confidence': '低',
-                        'source': 'local'
-                    })
                     
-                self.prev_human_count = human_result['count']
+                    if human_count_change > 0:
+                        events.append({
+                            'event_type': '人员进出',
+                            'description': f'检测到 {human_count_change} 人进入（Haar）',
+                            'confidence': '中',
+                            'source': 'local'
+                        })
+                        self.human_count += human_count_change
+                        self.person_change_count += 1
+                    elif human_count_change < 0:
+                        events.append({
+                            'event_type': '人员进出',
+                            'description': f'检测到 {abs(human_count_change)} 人离开（Haar）',
+                            'confidence': '中',
+                            'source': 'local'
+                        })
+                        self.person_change_count += 1
+                    
+                    self.prev_human_count = haar_result['count']
+                else:
+                    # 并行运行时，记录 Haar 结果但不作为主要判断
+                    metrics['haar_count'] = haar_result['count']
             
             # 3. 亮度检测
             brightness_result = self._detect_brightness_change(current_gray, prev_image_path)
@@ -367,28 +385,44 @@ class LocalAnalyzer:
     def _should_call_model(self, metrics, events):
         """判断是否需要调用大模型"""
         
-        # 条件1：人数变化（最重要）
+        # 检查大模型是否启用
-        current_person_count = metrics.get('person_count', metrics.get('human_count', 0))
+        use_vision_api = config_mgr.get('use_vision_api', False)
-        person_count_change = metrics.get('person_count_change', metrics.get('human_count_change', 0))
+        if not use_vision_api:
+            print("[LocalAnalyzer] Vision API disabled, skipping model call")
+            return False
         
-        # 更新前一帧人数（如果还没更新）
+        # 获取触发条件配置
-        if abs(person_count_change) >= self.config['human_count_change_threshold']:
+        trigger_condition = config_mgr.get('vision_api_trigger', 'person_change')
-            print(f"[LocalAnalyzer] Person count changed: {current_person_count - person_count_change} -> {current_person_count}, triggering model")
-            return True
         
-        # 条件2：检测到新人
+        # 条件1：人员变化（person_change 模式）
-        if metrics.get('new_persons', 0) > 0:
+        if trigger_condition in ['person_change', 'always']:
-            print(f"[LocalAnalyzer] New person detected: {metrics.get('new_persons', 0)}, triggering model")
+            current_person_count = metrics.get('person_count', metrics.get('human_count', 0))
-            return True
+            person_count_change = metrics.get('person_count_change', metrics.get('human_count_change', 0))
             
-        # 条件3：运动面积超过阈值（排除有人但不动的情况）
+            if abs(person_count_change) >= self.config['human_count_change_threshold']:
-        if metrics.get('motion_ratio', 0) > self.config['trigger_model_threshold'] * 2:
+                print(f"[LocalAnalyzer] Person change detected: {person_count_change}, triggering model")
-            print(f"[LocalAnalyzer] Large motion detected: {metrics.get('motion_ratio', 0):.2%}")
+                return True
-            return True
             
-        # 条件4：亮度大幅变化（灯开关等）
+            # 检测到新人
-        if abs(metrics.get('brightness_change', 0)) > self.config['brightness_change_threshold'] * 2:
+            if metrics.get('new_persons', 0) > 0:
-            print(f"[LocalAnalyzer] Brightness changed: {metrics.get('brightness_change', 0)}")
+                print(f"[LocalAnalyzer] New person detected: {metrics.get('new_persons', 0)}, triggering model")
+                return True
+        
+        # 条件2：运动检测（motion 模式）
+        if trigger_condition in ['motion', 'always']:
+            if metrics.get('motion_ratio', 0) > self.config['trigger_model_threshold'] * 2:
+                print(f"[LocalAnalyzer] Large motion detected: {metrics.get('motion_ratio', 0):.2%}")
+                return True
+        
+        # 条件3：亮度变化（brightness 模式）
+        if trigger_condition in ['brightness', 'always']:
+            if abs(metrics.get('brightness_change', 0)) > self.config['brightness_change_threshold'] * 2:
+                print(f"[LocalAnalyzer] Brightness changed: {metrics.get('brightness_change', 0)}")
+                return True
+        
+        # always 模式：总是调用
+        if trigger_condition == 'always':
+            print("[LocalAnalyzer] Always trigger mode enabled")
             return True
         
         return False

web/static/app.js

@@ -505,6 +505,17 @@ function loadSettingsForm() {
             document.getElementById('setting-auto-analyze').checked = config.auto_analyze !== false;
             document.getElementById('setting-display-limit').value = config.display_limit || 20;
             document.getElementById('setting-refresh-interval').value = config.refresh_interval || 5;
+            
+            // Detection algorithm settings
+            document.getElementById('setting-use-haar').checked = config.use_haar_cascade !== false;
+            document.getElementById('setting-use-mediapipe').checked = config.use_mediapipe_face !== false;
+            document.getElementById('setting-use-face-rec').checked = config.use_face_recognition !== false;
+            
+            // Vision API settings
+            document.getElementById('setting-use-vision-api').checked = config.use_vision_api === true;
+            document.getElementById('setting-vision-trigger').value = config.vision_api_trigger || 'person_change';
+            
+            // API config
             document.getElementById('setting-api-url').value = config.vision_api_url || '';
             document.getElementById('setting-api-key').value = config.vision_api_key || '';
             document.getElementById('setting-model').value = config.vision_model || '';
@@ -520,6 +531,17 @@ function saveSettings() {
         auto_analyze: document.getElementById('setting-auto-analyze').checked,
         display_limit: parseInt(document.getElementById('setting-display-limit').value),
         refresh_interval: parseInt(document.getElementById('setting-refresh-interval').value),
+        
+        // Detection algorithms
+        use_haar_cascade: document.getElementById('setting-use-haar').checked,
+        use_mediapipe_face: document.getElementById('setting-use-mediapipe').checked,
+        use_face_recognition: document.getElementById('setting-use-face-rec').checked,
+        
+        // Vision API
+        use_vision_api: document.getElementById('setting-use-vision-api').checked,
+        vision_api_trigger: document.getElementById('setting-vision-trigger').value,
+        
+        // API config
         vision_api_url: document.getElementById('setting-api-url').value,
         vision_api_key: document.getElementById('setting-api-key').value,
         vision_model: document.getElementById('setting-model').value
@@ -533,14 +555,13 @@ function saveSettings() {
         .then(function(res) { return res.json(); })
         .then(function(data) {
             if (data.success) {
-                alert('Settings saved!');
+                showToast('Settings saved!', 1500);
                 closeSettingsModal();
-                // Reload config to apply new refresh interval
                 loadConfig();
                 refreshAll();
             }
         })
-        .catch(function(e) { alert('Error: ' + e.message); });
+        .catch(function(e) { showToast('Error: ' + e.message, 3000); });
 }
 
 function browseImagesDir() {

web/static/index.html

@@ -137,6 +137,43 @@
                     </div>
                 </div>
                 
+                <div class="settings-section">
+                    <h4>检测算法设置</h4>
+                    <div class="setting-item">
+                        <label>Haar Cascade 人体检测:</label>
+                        <input type="checkbox" id="setting-use-haar" checked>
+                        <span class="setting-desc">传统人体检测（备用）</span>
+                    </div>
+                    <div class="setting-item">
+                        <label>MediaPipe 人脸检测:</label>
+                        <input type="checkbox" id="setting-use-mediapipe" checked>
+                        <span class="setting-desc">高精度人脸检测</span>
+                    </div>
+                    <div class="setting-item">
+                        <label>Face Recognition:</label>
+                        <input type="checkbox" id="setting-use-face-rec" checked>
+                        <span class="setting-desc">人脸识别（识别同一人）</span>
+                    </div>
+                </div>
+                
+                <div class="settings-section">
+                    <h4>AI大模型分析</h4>
+                    <div class="setting-item">
+                        <label>启用大模型分析:</label>
+                        <input type="checkbox" id="setting-use-vision-api">
+                        <span class="setting-desc">调用 Vision API 详细分析</span>
+                    </div>
+                    <div class="setting-item">
+                        <label>触发条件:</label>
+                        <select id="setting-vision-trigger">
+                            <option value="person_change">人员变化时</option>
+                            <option value="motion">运动检测时</option>
+                            <option value="brightness">亮度变化时</option>
+                            <option value="always">总是分析</option>
+                        </select>
+                    </div>
+                </div>
+                
                 <div class="settings-section">
                     <h4>API 设置</h4>
                     <div class="setting-item">

web/static/style.css

@@ -649,6 +649,16 @@ button {
     gap: 10px;
 }
 
+.person-actions {
+    display: flex;
+    gap: 10px;
+}
+
+.setting-desc {
+    color: #888;
+    font-size: 12px;
+}
+
 @media (max-width: 768px) {
     .control-panel {
         flex-direction: column;