support tflite

Dockerfile_ai

@@ -43,8 +43,14 @@ RUN pip3 config set global.index-url https://mirrors.aliyun.com/pypi/simple/ \
 
 # 安装 Python 依赖
 # --break-system-packages: Debian Trixie 使用 PEP 668 保护系统 Python
+# TFLite 支持：优先尝试 tflite-runtime，失败则用 ai-edge-litert（Google 新包名）
 COPY ./analysis/requirements.txt /tmp/requirements.txt
 RUN pip3 install --no-cache-dir --break-system-packages -r /tmp/requirements.txt \
+    && (pip3 install --no-cache-dir --break-system-packages \
+            -i https://pypi.org/simple/ tflite-runtime 2>/dev/null \
+        || pip3 install --no-cache-dir --break-system-packages \
+            -i https://pypi.org/simple/ ai-edge-litert 2>/dev/null \
+        || echo "TFLite runtime not available, will use ONNX only") \
     && rm /tmp/requirements.txt \
     && rm -rf /root/.cache/pip
 
@@ -54,7 +60,9 @@ COPY --from=mwader/static-ffmpeg:6.1 /ffmpeg /usr/local/bin/ffmpeg
 COPY --from=mwader/static-ffmpeg:6.1 /ffprobe /usr/local/bin/ffprobe
 
 # 复制应用文件
-ADD ./build/linux_amd64/bin ./gowvp
+# 使用 TARGETARCH 自动识别目标架构（amd64 或 arm64）
+ARG TARGETARCH
+ADD ./build/linux_${TARGETARCH}/bin ./gowvp
 ADD ./www ./www
 ADD ./analysis ./analysis
 

Dockerfile_zlm

@@ -18,7 +18,8 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
 COPY --from=zlmediakit/zlmediakit:master /opt/media/bin /opt/media/bin
 COPY --from=mwader/static-ffmpeg:6.1 /ffmpeg /usr/local/bin/ffmpeg
 
-ADD ./build/linux_amd64/bin ./gowvp
+ARG TARGETARCH
+ADD ./build/linux_${TARGETARCH}/bin ./gowvp
 ADD ./www ./www
 
 RUN mkdir -p configs

README_CN.md

@@ -135,11 +135,13 @@ postgres 和 mysql 的格式即:
 
 > 如何关闭 AI?
 
-ai 默认是开启状态，1 秒检测 5 帧
+ai 默认是开启状态，1 秒检测 1 帧
 
 可以在 `configs/config.toml` 中修改 `disabledAI = true` 关闭 ai 检测
 
+> 国标设备在线，通道离线?
 
+属于 ipc 的问题，请检查 ipc 后台注册的 平台 sip_id 和 域是否与 gowvp/owl 一致。
 
 
 ## 文档
@@ -302,6 +304,8 @@ services:
 [@joestarzxh](https://github.com/joestarzxh)
 [@oldweipro](https://github.com/oldweipro)
 [@beixiaocai](https://github.com/beixiaocai)
+[@chencanfggz](https://github.com/chencanfggz)
+[@zhangxuan1340](https://github.com/zhangxuan1340)
 
 
 ## 许可证

analysis/Makefile

@@ -0,0 +1,19 @@
+
+## ai/init: 安装 AI 分析模块依赖（本地开发用）
+ai/init:
+	@echo "安装基础依赖..."
+	@pip install -r analysis/requirements.txt
+	@echo ""
+	@echo "TFLite 支持需要额外安装（二选一）:"
+	@echo "  Linux:        make ai/init/tflite"
+	@echo "  macOS/Windows: make ai/init/tensorflow"
+	@echo ""
+	@echo "如果使用 .tflite 模型，请根据系统选择安装"
+
+## ai/init/tflite: 安装 TFLite 支持（Linux 轻量版）
+ai/init/tflite:
+	@conda install tflite-runtime
+
+## ai/init/tensorflow: 安装 TensorFlow（macOS/Windows 完整版）
+ai/init/tensorflow:
+	@ pip install tensorflow -i https://pypi.org/simple

analysis/detect.py

@@ -1,11 +1,11 @@
 import logging
+import os
 import time
+from abc import ABC, abstractmethod
 from typing import Any
 
 import numpy as np
 import cv2
-import onnxruntime as ort
-
 
 slog = logging.getLogger("Detector")
 
@@ -94,55 +94,77 @@ COCO_LABELS = [
 ]
 
 
-class ObjectDetector:
+class ModelBackend(ABC):
+    """
+    模型推理后端抽象接口
+    不同模型格式（ONNX、TFLite）需实现此接口，确保上层调用逻辑统一
+    """
 
-    def __init__(self, model_path: str = "yolo11n.onnx"):
-        self.model_path = model_path
-        self.session: ort.InferenceSession | None = None
+    @abstractmethod
+    def load(self, model_path: str) -> bool:
+        """加载模型文件，返回是否成功"""
+        pass
+
+    @abstractmethod
+    def is_ready(self) -> bool:
+        """检查模型是否已加载并可用"""
+        pass
+
+    @abstractmethod
+    def get_input_shape(self) -> tuple:
+        """获取模型输入形状，用于预处理"""
+        pass
+
+    @abstractmethod
+    def infer(self, input_tensor: np.ndarray) -> np.ndarray:
+        """执行推理，返回原始输出"""
+        pass
+
+
+class ONNXBackend(ModelBackend):
+    """
+    ONNX Runtime 推理后端
+    使用 onnxruntime 库加载和执行 ONNX 格式模型
+    """
+
+    def __init__(self):
+        self.session = None
         self.input_name: str = ""
         self.input_shape: tuple = (1, 3, 640, 640)
         self._is_ready = False
-        self.names: dict[int, str] = {i: name for i, name in enumerate(COCO_LABELS)}
 
-    def load_model(self) -> bool:
-        """加载 ONNX 模型并初始化推理会话"""
+    def load(self, model_path: str) -> bool:
         try:
-            slog.info(f"加载 ONNX 模型: {self.model_path} ...")
+            import onnxruntime as ort
+
+            slog.info(f"加载 ONNX 模型: {model_path} ...")
             start_time = time.time()
 
-            # 配置 ONNX Runtime 会话选项
             sess_options = ort.SessionOptions()
             sess_options.graph_optimization_level = (
                 ort.GraphOptimizationLevel.ORT_ENABLE_ALL
             )
-            # 限制线程数，避免在容器中占用过多 CPU
             sess_options.intra_op_num_threads = 4
             sess_options.inter_op_num_threads = 2
 
-            # 优先使用 CPU 执行提供程序
             providers = ["CPUExecutionProvider"]
-
             self.session = ort.InferenceSession(
-                self.model_path, sess_options=sess_options, providers=providers
+                model_path, sess_options=sess_options, providers=providers
             )
 
-            # 获取输入信息
             input_info = self.session.get_inputs()[0]
             self.input_name = input_info.name
             self.input_shape = tuple(input_info.shape)
 
-            # 预热模型
-            dummy_img = np.zeros((640, 640, 3), dtype=np.uint8)
-            self._preprocess(dummy_img)
-            dummy_input = self._preprocess(dummy_img)
-            self.session.run(None, {self.input_name: dummy_input})
-
             elapsed = time.time() - start_time
             slog.info(
                 f"ONNX 模型加载完成 (耗时: {elapsed:.2f}s, 输入形状: {self.input_shape})"
             )
             self._is_ready = True
             return True
+        except ImportError:
+            slog.error("未安装 onnxruntime，无法加载 ONNX 模型")
+            return False
         except Exception as e:
             slog.error(f"加载 ONNX 模型失败: {e}")
             return False
@@ -150,29 +172,335 @@ class ObjectDetector:
     def is_ready(self) -> bool:
         return self._is_ready and self.session is not None
 
+    def get_input_shape(self) -> tuple:
+        return self.input_shape
+
+    def infer(self, input_tensor: np.ndarray) -> np.ndarray:
+        if not self.session:
+            raise RuntimeError("ONNX 模型未加载")
+        outputs = self.session.run(None, {self.input_name: input_tensor})
+        return np.asarray(outputs[0])
+
+
+class TFLiteBackend(ModelBackend):
+    """
+    TensorFlow Lite 推理后端
+    支持两种模型格式：
+    1. YOLO 格式：单输出张量 (1, 84, 8400)
+    2. SSD 格式：多输出张量（boxes, classes, scores, num_detections）
+    """
+
+    def __init__(self):
+        self.interpreter: Any = None
+        self.input_details: list[dict[str, Any]] = []
+        self.output_details: list[dict[str, Any]] = []
+        self.input_shape: tuple = (1, 640, 640, 3)
+        self._is_ready = False
+        self._is_nhwc = True
+        self._is_ssd_format = False  # 区分 SSD 和 YOLO 格式
+        self._input_quantization: tuple[float, int] = (1.0, 0)  # scale, zero_point
+
+    def load(self, model_path: str) -> bool:
+        try:
+            Interpreter = None
+            try:
+                from tflite_runtime.interpreter import Interpreter  # type: ignore
+            except ImportError:
+                try:
+                    from ai_edge_litert.interpreter import Interpreter  # type: ignore
+                except ImportError:
+                    try:
+                        import tensorflow as tf
+
+                        Interpreter = tf.lite.Interpreter
+                    except ImportError:
+                        pass
+
+            if Interpreter is None:
+                raise ImportError("未找到 tflite_runtime、ai_edge_litert 或 tensorflow")
+
+            slog.info(f"加载 TFLite 模型: {model_path} ...")
+            start_time = time.time()
+
+            self.interpreter = Interpreter(model_path=model_path)
+            self.interpreter.allocate_tensors()
+
+            self.input_details = self.interpreter.get_input_details()
+            self.output_details = self.interpreter.get_output_details()
+
+            input_shape = self.input_details[0]["shape"]
+            self.input_shape = tuple(input_shape)
+
+            if len(self.input_shape) == 4:
+                self._is_nhwc = self.input_shape[3] == 3
+
+            # 获取输入量化参数（用于 uint8 量化模型）
+            quant_params = self.input_details[0].get("quantization_parameters", {})
+            scales = quant_params.get("scales", np.array([1.0]))
+            zero_points = quant_params.get("zero_points", np.array([0]))
+            if len(scales) > 0 and len(zero_points) > 0:
+                self._input_quantization = (float(scales[0]), int(zero_points[0]))
+
+            # 检测模型格式：SSD 模型通常有4个输出（boxes, classes, scores, num）
+            # 且输出名称包含 "TFLite_Detection_PostProcess"
+            self._is_ssd_format = len(self.output_details) >= 3 and any(
+                "Detection" in d.get("name", "") for d in self.output_details
+            )
+
+            elapsed = time.time() - start_time
+            format_name = "SSD" if self._is_ssd_format else "YOLO"
+            slog.info(
+                f"TFLite 模型加载完成 (耗时: {elapsed:.2f}s, 输入: {self.input_shape}, "
+                f"格式: {format_name}, 量化: {self._input_quantization})"
+            )
+            self._is_ready = True
+            return True
+        except ImportError:
+            slog.error("未安装 tflite_runtime 或 tensorflow，无法加载 TFLite 模型")
+            return False
+        except Exception as e:
+            slog.error(f"加载 TFLite 模型失败: {e}")
+            return False
+
+    def is_ready(self) -> bool:
+        return self._is_ready and self.interpreter is not None
+
+    def get_input_shape(self) -> tuple:
+        return self.input_shape
+
+    def is_nhwc(self) -> bool:
+        """返回模型是否使用 NHWC 格式"""
+        return self._is_nhwc
+
+    def is_ssd_format(self) -> bool:
+        """返回是否为 SSD 格式（多输出张量）"""
+        return self._is_ssd_format
+
+    def get_input_quantization(self) -> tuple[float, int]:
+        """返回输入量化参数 (scale, zero_point)"""
+        return self._input_quantization
+
+    def get_input_dtype(self) -> np.dtype:
+        """返回模型期望的输入数据类型"""
+        return self.input_details[0]["dtype"]
+
+    def infer(self, input_tensor: np.ndarray) -> np.ndarray:
+        """执行推理，返回第一个输出张量（用于 YOLO 格式）"""
+        if not self.interpreter or len(self.input_details) == 0:
+            raise RuntimeError("TFLite 模型未加载")
+
+        input_dtype = self.input_details[0]["dtype"]
+        if input_tensor.dtype != input_dtype:
+            input_tensor = input_tensor.astype(input_dtype)
+
+        self.interpreter.set_tensor(self.input_details[0]["index"], input_tensor)
+        self.interpreter.invoke()
+
+        output = self.interpreter.get_tensor(self.output_details[0]["index"])
+        return np.asarray(output)
+
+    def infer_ssd(
+        self, input_tensor: np.ndarray
+    ) -> tuple[np.ndarray, np.ndarray, np.ndarray, int]:
+        """
+        执行 SSD 格式推理，返回解析后的检测结果
+        SSD 输出格式（已内置后处理）：
+        - boxes: (1, num_boxes, 4) 归一化坐标 [y_min, x_min, y_max, x_max]
+        - classes: (1, num_boxes) 类别 ID
+        - scores: (1, num_boxes) 置信度分数
+        - num_detections: 有效检测数量
+        """
+        if not self.interpreter or len(self.input_details) == 0:
+            raise RuntimeError("TFLite 模型未加载")
+
+        input_dtype = self.input_details[0]["dtype"]
+        if input_tensor.dtype != input_dtype:
+            input_tensor = input_tensor.astype(input_dtype)
+
+        self.interpreter.set_tensor(self.input_details[0]["index"], input_tensor)
+        self.interpreter.invoke()
+
+        # 按名称或索引获取各输出张量
+        boxes = None
+        classes = None
+        scores = None
+        num_detections = 0
+
+        for detail in self.output_details:
+            name = detail.get("name", "")
+            tensor = self.interpreter.get_tensor(detail["index"])
+
+            if "boxes" in name.lower() or (
+                detail["shape"][-1] == 4 and len(detail["shape"]) == 3
+            ):
+                boxes = np.asarray(tensor)
+            elif "class" in name.lower() or (
+                len(detail["shape"]) == 2
+                and detail["shape"][1] > 1
+                and boxes is not None
+            ):
+                classes = np.asarray(tensor)
+            elif "score" in name.lower() or ":2" in name:
+                scores = np.asarray(tensor)
+            elif "num" in name.lower() or (
+                len(detail["shape"]) == 1 and detail["shape"][0] == 1
+            ):
+                num_detections = int(tensor[0])
+
+        # 兜底处理：按输出顺序分配
+        if boxes is None or classes is None or scores is None:
+            outputs = [
+                self.interpreter.get_tensor(d["index"]) for d in self.output_details
+            ]
+            if len(outputs) >= 4:
+                boxes = np.asarray(outputs[0])
+                classes = np.asarray(outputs[1])
+                scores = np.asarray(outputs[2])
+                num_detections = int(outputs[3][0]) if outputs[3].size > 0 else 0
+
+        if boxes is None:
+            boxes = np.array([])
+        if classes is None:
+            classes = np.array([])
+        if scores is None:
+            scores = np.array([])
+
+        return boxes, classes, scores, num_detections
+
+
+def get_model_type(model_path: str) -> str:
+    """根据模型文件后缀判断模型类型"""
+    ext = os.path.splitext(model_path)[1].lower()
+    return "tflite" if ext == ".tflite" else "onnx"
+
+
+def create_backend(model_type: str) -> ModelBackend:
+    """
+    根据模型类型创建对应的推理后端
+    """
+    if model_type == "tflite":
+        return TFLiteBackend()
+    else:
+        return ONNXBackend()
+
+
+class ObjectDetector:
+    """
+    目标检测器 - 支持多种模型格式（ONNX、TFLite）
+    通过统一的 ModelBackend 接口实现模型无关的检测逻辑
+    """
+
+    def __init__(self, model_path: str):
+        self.model_path = model_path
+        self.model_type = get_model_type(model_path)
+        self.backend: ModelBackend | None = None
+        self.input_shape: tuple = (1, 3, 640, 640)
+        self._is_ready = False
+        self.names: dict[int, str] = {i: name for i, name in enumerate(COCO_LABELS)}
+
+    def load_model(self) -> bool:
+        """加载模型并初始化推理后端"""
+        try:
+            start_time = time.time()
+
+            # 创建对应类型的后端
+            self.backend = create_backend(self.model_type)
+
+            # 加载模型
+            if not self.backend.load(self.model_path):
+                return False
+
+            self.input_shape = self.backend.get_input_shape()
+
+            # 预热模型
+            self._warmup()
+
+            elapsed = time.time() - start_time
+            slog.info(f"模型预热完成 (总耗时: {elapsed:.2f}s)")
+            self._is_ready = True
+            return True
+        except Exception as e:
+            slog.error(f"加载模型失败: {e}")
+            return False
+
+    def _warmup(self) -> None:
+        """预热模型，减少首次推理延迟"""
+        if not self.backend:
+            return
+
+        dummy_img = np.zeros((640, 640, 3), dtype=np.uint8)
+        dummy_input = self._preprocess(dummy_img)
+        self.backend.infer(dummy_input)
+        slog.info("模型预热完成")
+
+    def is_ready(self) -> bool:
+        return self._is_ready and self.backend is not None and self.backend.is_ready()
+
+    def _get_target_size(self) -> int:
+        """获取模型期望的输入尺寸"""
+        # NCHW: (1, 3, H, W) -> shape[2]
+        # NHWC: (1, H, W, 3) -> shape[1]
+        if self._is_nhwc_format():
+            return int(self.input_shape[1])
+        return int(self.input_shape[2])
+
+    def _is_nhwc_format(self) -> bool:
+        """判断当前后端是否使用 NHWC 格式"""
+        if isinstance(self.backend, TFLiteBackend):
+            return self.backend.is_nhwc()
+        return False
+
+    def _is_ssd_format(self) -> bool:
+        """判断当前后端是否为 SSD 格式"""
+        if isinstance(self.backend, TFLiteBackend):
+            return self.backend.is_ssd_format()
+        return False
+
     def _preprocess(self, image: np.ndarray) -> np.ndarray:
         """
         预处理图像：调整大小、归一化、转换格式
-        YOLO 期望输入格式: NCHW, float32, 归一化到 [0, 1]
+        根据后端类型自动选择 NCHW 或 NHWC 格式，并处理量化输入
         """
-        # 保持宽高比的 letterbox 缩放
-        target_size = self.input_shape[2]  # 640
+        target_size = self._get_target_size()
         h, w = image.shape[:2]
+
+        # SSD 模型使用直接缩放（不保持宽高比）
+        # YOLO 模型使用 letterbox 缩放（保持宽高比）
+        if self._is_ssd_format():
+            resized = cv2.resize(
+                image, (target_size, target_size), interpolation=cv2.INTER_LINEAR
+            )
+            rgb = resized[:, :, ::-1]  # BGR -> RGB
+
+            # 检查是否需要量化为 uint8
+            if isinstance(self.backend, TFLiteBackend):
+                input_dtype = self.backend.get_input_dtype()
+                if input_dtype == np.uint8:
+                    # 直接返回 uint8 格式
+                    return np.expand_dims(rgb.astype(np.uint8), axis=0)
+
+            # float32 格式
+            rgb = rgb.astype(np.float32) / 255.0
+            return np.expand_dims(rgb, axis=0)
+
+        # YOLO letterbox 预处理
         scale = min(target_size / h, target_size / w)
         new_h, new_w = int(h * scale), int(w * scale)
 
-        # 缩放图像
         resized = cv2.resize(image, (new_w, new_h), interpolation=cv2.INTER_LINEAR)
 
-        # 创建正方形画布并居中放置图像
         canvas = np.full((target_size, target_size, 3), 114, dtype=np.uint8)
         top = (target_size - new_h) // 2
         left = (target_size - new_w) // 2
         canvas[top : top + new_h, left : left + new_w] = resized
 
-        # BGR -> RGB, HWC -> CHW, 归一化
-        blob = canvas[:, :, ::-1].transpose(2, 0, 1).astype(np.float32) / 255.0
-        return np.expand_dims(blob, axis=0)
+        rgb = canvas[:, :, ::-1].astype(np.float32) / 255.0
+
+        if self._is_nhwc_format():
+            return np.expand_dims(rgb, axis=0)
+        else:
+            blob = rgb.transpose(2, 0, 1)
+            return np.expand_dims(blob, axis=0)
 
     def _postprocess(
         self,
@@ -214,7 +542,7 @@ class ObjectDetector:
 
         # 缩放坐标到原始图像尺寸
         orig_h, orig_w = original_shape
-        target_size = self.input_shape[2]
+        target_size = self._get_target_size()
         scale = min(target_size / orig_h, target_size / orig_w)
         pad_h = (target_size - orig_h * scale) / 2
         pad_w = (target_size - orig_w * scale) / 2
@@ -341,18 +669,17 @@ class ObjectDetector:
         self, image: np.ndarray, threshold: float, label_filter: list[str] | None = None
     ) -> list[dict[str, Any]]:
         """对单张图像执行检测"""
-        if not self.session:
+        if not self.backend or not self.backend.is_ready():
             return []
 
         # 预处理
         input_tensor = self._preprocess(image)
 
         # 推理
-        outputs = self.session.run(None, {self.input_name: input_tensor})
+        output = self.backend.infer(input_tensor)
 
         # 后处理
         original_shape = image.shape[:2]
-        output: np.ndarray = np.asarray(outputs[0])
         return self._postprocess(output, original_shape, threshold, label_filter)
 
 

analysis/detect_test.py

@@ -0,0 +1,150 @@
+#!/usr/bin/env python3
+"""
+测试脚本 - 用于诊断 TFLite 和 ONNX 模型的检测问题
+使用方法: python detect_test.py <image_path> [model_path]
+"""
+
+import sys
+import os
+import cv2
+import numpy as np
+
+# 添加当前目录到路径
+sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
+
+from detect import ObjectDetector, TFLiteBackend, ONNXBackend
+
+
+def draw_detections(image: np.ndarray, detections: list) -> np.ndarray:
+    """在图像上绘制检测结果"""
+    img = image.copy()
+    for det in detections:
+        box = det["box"]
+        label = det["label"]
+        conf = det["confidence"]
+
+        x1, y1 = box["x_min"], box["y_min"]
+        x2, y2 = box["x_max"], box["y_max"]
+
+        # 绘制边框
+        cv2.rectangle(img, (x1, y1), (x2, y2), (0, 255, 0), 2)
+
+        # 绘制标签
+        text = f"{label}: {conf:.2%}"
+        (tw, th), _ = cv2.getTextSize(text, cv2.FONT_HERSHEY_SIMPLEX, 0.6, 1)
+        cv2.rectangle(img, (x1, y1 - th - 10), (x1 + tw, y1), (0, 255, 0), -1)
+        cv2.putText(
+            img, text, (x1, y1 - 5), cv2.FONT_HERSHEY_SIMPLEX, 0.6, (0, 0, 0), 1
+        )
+
+    return img
+
+
+def test_model(model_path: str, image_path: str):
+    """测试单个模型"""
+    print(f"\n{'='*60}")
+    print(f"测试模型: {model_path}")
+    print(f"测试图片: {image_path}")
+    print(f"{'='*60}")
+
+    # 读取图像
+    image = cv2.imread(image_path)
+    if image is None:
+        print(f"错误: 无法读取图像 {image_path}")
+        return
+
+    print(f"图像尺寸: {image.shape}")
+
+    # 创建检测器
+    detector = ObjectDetector(model_path)
+    if not detector.load_model():
+        print("错误: 模型加载失败")
+        return
+
+    print(f"模型类型: {detector.model_type}")
+    print(f"输入形状: {detector.input_shape}")
+
+    # 检查后端类型
+    if isinstance(detector.backend, TFLiteBackend):
+        print(f"TFLite 格式: {'NHWC' if detector.backend.is_nhwc() else 'NCHW'}")
+
+        # 调试: 查看输入输出详情
+        print(f"输入详情: {detector.backend.input_details}")
+        print(f"输出详情: {detector.backend.output_details}")
+
+    # 执行检测
+    print("\n执行检测...")
+
+    # 调试: 手动执行预处理并检查
+    input_tensor = detector._preprocess(image)
+    print(f"预处理后张量形状: {input_tensor.shape}")
+    print(f"预处理后张量范围: [{input_tensor.min():.4f}, {input_tensor.max():.4f}]")
+    print(f"预处理后张量类型: {input_tensor.dtype}")
+
+    # 执行推理
+    if detector.backend is None:
+        print("错误: 后端未初始化")
+        return
+    output = detector.backend.infer(input_tensor)
+    print(f"原始输出形状: {output.shape}")
+    print(f"原始输出范围: [{output.min():.4f}, {output.max():.4f}]")
+
+    # 检查输出格式
+    if len(output.shape) == 3:
+        print(
+            f"输出维度分析: batch={output.shape[0]}, dim1={output.shape[1]}, dim2={output.shape[2]}"
+        )
+        # 如果是 (1, 8400, 84) 格式，不需要转置
+        # 如果是 (1, 84, 8400) 格式，需要转置
+        if output.shape[1] == 84 and output.shape[2] == 8400:
+            print("输出格式: (1, 84, 8400) - YOLO 标准格式")
+        elif output.shape[1] == 8400 and output.shape[2] == 84:
+            print("输出格式: (1, 8400, 84) - 需要调整后处理逻辑")
+
+    # 使用标准检测流程
+    detections, inference_time = detector.detect(image, threshold=0.25)
+
+    print(f"\n检测结果 (阈值=0.25):")
+    print(f"推理耗时: {inference_time:.2f} ms")
+    print(f"检测到 {len(detections)} 个目标:")
+
+    for i, det in enumerate(detections):
+        print(f"  {i+1}. {det['label']}: {det['confidence']:.2%} at {det['box']}")
+        # 检查置信度是否异常
+        if det["confidence"] > 1.0:
+            print(f"     ⚠️ 警告: 置信度超过100%! 原始值: {det['confidence']}")
+
+    # 保存结果图像
+    model_name = os.path.splitext(os.path.basename(model_path))[0]
+    output_path = f"/Users/xugo/Desktop/gowvp/gb28181/analysis/result_{model_name}.jpg"
+    result_img = draw_detections(image, detections)
+    cv2.imwrite(output_path, result_img)
+    print(f"\n结果图像已保存: {output_path}")
+
+
+def main():
+    # 默认图像路径
+    image_path = "/Users/xugo/Desktop/gowvp/gb28181/out.png"
+
+    # 命令行参数
+    if len(sys.argv) > 1:
+        image_path = sys.argv[1]
+
+    # 模型路径
+    tflite_model = "/Users/xugo/Desktop/gowvp/gb28181/configs/owl.tflite"
+    onnx_model = "/Users/xugo/Desktop/gowvp/gb28181/analysis/owl.onnx"
+
+    if len(sys.argv) > 2:
+        # 只测试指定模型
+        test_model(sys.argv[2], image_path)
+    else:
+        # 测试所有可用模型
+        if os.path.exists(onnx_model):
+            test_model(onnx_model, image_path)
+
+        if os.path.exists(tflite_model):
+            test_model(tflite_model, image_path)
+
+
+if __name__ == "__main__":
+    main()

analysis/main.py

@@ -25,6 +25,14 @@ from detect import MotionDetector, ObjectDetector
 from frame_capture import FrameCapture
 import cv2
 
+# 模型文件搜索候选路径（按优先级排序）
+MODEL_SEARCH_PATHS = [
+    ("../configs/owl.tflite", "tflite"),
+    ("../configs/owl.onnx", "onnx"),
+    ("./owl.tflite", "tflite"),
+    ("./owl.onnx", "onnx"),
+]
+
 # 导入生成的 proto 代码
 # 这些模块必须存在才能启动 gRPC 服务
 import analysis_pb2
@@ -480,10 +488,31 @@ def serve(port, model_path):
         server.stop(0)
 
 
+def discover_model(model_arg: str) -> str:
+    """
+    自动发现可用模型文件
+    优先级：../configs/owl.tflite > ../configs/owl.onnx > ./owl.tflite > ./owl.onnx > 命令行参数
+    """
+    script_dir = os.path.dirname(os.path.abspath(__file__))
+
+    for rel_path, _ in MODEL_SEARCH_PATHS:
+        full_path = os.path.normpath(os.path.join(script_dir, rel_path))
+        if os.path.exists(full_path):
+            slog.info(f"发现模型文件: {full_path}")
+            return full_path
+
+    # 回退到命令行参数指定的模型
+    if os.path.isabs(model_arg):
+        return model_arg
+
+    # 相对路径基于脚本目录解析
+    return os.path.normpath(os.path.join(script_dir, model_arg))
+
+
 def main():
     parser = argparse.ArgumentParser()
     parser.add_argument("--port", type=int, default=50051)
-    parser.add_argument("--model", type=str, default="yolo11n.onnx")
+    parser.add_argument("--model", type=str, default="owl.onnx")
     parser.add_argument(
         "--callback-url",
         type=str,
@@ -503,11 +532,14 @@ def main():
     GLOBAL_CONFIG["callback_url"] = args.callback_url
     GLOBAL_CONFIG["callback_secret"] = args.callback_secret
 
+    # 自动发现模型文件
+    model_path = discover_model(args.model)
+
     slog.debug(
-        f"log level: {args.log_level}, model: {args.model}, callback url: {args.callback_url}, callback secret: {args.callback_secret}"
+        f"log level: {args.log_level}, model: {model_path}, callback url: {args.callback_url}, callback secret: {args.callback_secret}"
     )
 
-    serve(args.port, args.model)
+    serve(args.port, model_path)
 
 
 if __name__ == "__main__":

analysis/requirements.txt

@@ -1,4 +1,4 @@
-# ONNX Runtime - 替代 PyTorch，大幅减小镜像体积
+# ONNX Runtime - 用于加载 .onnx 模型
 onnxruntime>=1.17.0
 
 # gRPC 框架

changelog

@@ -1,3 +1,11 @@
+# 2026-01-11
+
+修复 webrtc 播放地址
+支持 ai 检测，支持 onnx 和 tflite
+支持自定义模型，写入 configs/owl.onnx 或 configs/owl.tflite
+支持修改国标信息免重启生效
+修复 rtmp port 偶先值为 0
+
 # 2025-06-14
 重构国标播放逻辑, play 接口仅返回播放地址，通过 webhook 来通知拉流
 可能会导致首播慢一些

internal/core/sms/driver_zlm.go

@@ -20,9 +20,9 @@ func (d *ZLMDriver) GetStreamLiveAddr(ctx context.Context, ms *MediaServer, http
 	var out StreamLiveAddr
 	out.Label = "ZLM"
 	wsPrefix := strings.Replace(strings.Replace(httpPrefix, "https", "wss", 1), "http", "ws", 1)
-	out.WSFLV = fmt.Sprintf("%s/proxy/sms/%s.live.flv", wsPrefix, stream)
-	out.HTTPFLV = fmt.Sprintf("%s/proxy/sms/%s.live.flv", httpPrefix, stream)
-	out.HLS = fmt.Sprintf("%s/proxy/sms/%s/hls.fmp4.m3u8", httpPrefix, stream)
+	out.WSFLV = fmt.Sprintf("%s/proxy/sms/%s/%s.live.flv", wsPrefix, app, stream)
+	out.HTTPFLV = fmt.Sprintf("%s/proxy/sms/%s/%s.live.flv", httpPrefix, app, stream)
+	out.HLS = fmt.Sprintf("%s/proxy/sms/%s/%s/hls.fmp4.m3u8", httpPrefix, app, stream)
 	rtcPrefix := strings.Replace(strings.Replace(httpPrefix, "https", "webrtc", 1), "http", "webrtc", 1)
 	out.WebRTC = fmt.Sprintf("%s/proxy/sms/index/api/webrtc?app=%s&stream=%s&type=play", rtcPrefix, app, stream)
 	out.RTMP = fmt.Sprintf("rtmp://%s:%d/%s", host, ms.Ports.RTMP, stream)
@@ -81,9 +81,18 @@ func (d *ZLMDriver) Connect(ctx context.Context, ms *MediaServer) error {
 func (d *ZLMDriver) Setup(ctx context.Context, ms *MediaServer, webhookURL string) error {
 	engine := d.withConfig(ms)
 
+	// 拼接 IP 但是不要空格
+	ips := make([]string, 0, 2)
+	for _, ip := range []string{ms.SDPIP, ms.IP} {
+		if ip != "" {
+			ips = append(ips, ip)
+		}
+	}
+	_ = ips
 	// 构造配置请求
 	req := zlm.SetServerConfigRequest{
-		RtcExternIP:          new(ms.IP),
+		RtcExternIP: new(strings.Join(ips, ",")),
+
 		GeneralMediaServerID: new(ms.ID),
 		HookEnable:           new("1"),
 		HookOnFlowReport:     new(""),

internal/web/api/ai_webhook.go

@@ -136,7 +136,8 @@ func (a AIWebhookAPI) onEvents(c *gin.Context, in *AIDetectionInput) (AIWebhookO
 			Score:     float32(det.Confidence),
 			Zones:     string(zonesJSON),
 			ImagePath: imagePath,
-			Model:     "yolo11n",
+			// TODO: 模型名称可以根据模型自定义
+			Model: "default",
 		}
 
 		if _, err := a.eventCore.AddEvent(ctx, eventInput); err != nil {
@@ -259,7 +260,7 @@ func (a *AIWebhookAPI) StartAIDetection(ctx context.Context, ch *ipc.Channel, rt
 		CameraId:       ch.ID,
 		CameraName:     ch.Name,
 		RtspUrl:        rtspURL,
-		DetectFps:      5,
+		DetectFps:      1,
 		Labels:         labels,
 		Threshold:      0.75,
 		RoiPoints:      roiPoints,

internal/web/api/api.go

@@ -62,10 +62,11 @@ func setupRouter(r *gin.Engine, uc *Usecase) {
 		AllowMethods: []string{"GET", "POST", "PUT", "PATCH", "DELETE", "HEAD", "OPTIONS"},
 		AllowHeaders: []string{
 			"Accept", "Content-Length", "Content-Type", "Range", "Accept-Language",
-			"Origin", "Authorization",
+			"Origin", "Authorization", "Referer", "User-Agent",
 			"Accept-Encoding",
 			"Cache-Control", "Pragma", "X-Requested-With",
 			"Sec-Fetch-Mode", "Sec-Fetch-Site", "Sec-Fetch-Dest",
+			"Sec-Ch-Ua", "Sec-Ch-Ua-Mobile", "Sec-Ch-Ua-Platform",
 			"Dnt", "X-Forwarded-For", "X-Forwarded-Proto", "X-Forwarded-Host",
 			"X-Real-IP", "X-Request-ID", "X-Request-Start", "X-Request-Time",
 		},
@@ -105,6 +106,7 @@ func setupRouter(r *gin.Engine, uc *Usecase) {
 	registerPushAPI(r, uc.MediaAPI, auth)
 	registerGB28181(r, uc.GB28181API, auth)
 	registerProxy(r, uc.ProxyAPI, auth)
+	uc.ConfigAPI.uc = uc
 	registerConfig(r, uc.ConfigAPI, auth)
 	registerSms(r, uc.SMSAPI, auth)
 	RegisterUser(r, uc.UserAPI, auth)
@@ -327,6 +329,7 @@ func (uc *Usecase) proxySMS(c *gin.Context) {
 	fullAddr, _ := url.Parse(addr)
 	c.Request.URL.Path = ""
 	proxy := httputil.NewSingleHostReverseProxy(fullAddr)
+
 	proxy.Director = func(req *http.Request) {
 		// 设置请求的URL
 		req.URL.Scheme = "http"

internal/web/api/config.go

@@ -18,6 +18,7 @@ import (
 type ConfigAPI struct {
 	configCore config.Core
 	conf       *conf.Bootstrap
+	uc         *Usecase
 }
 
 func NewConfigAPI(db *gorm.DB, conf *conf.Bootstrap) ConfigAPI {
@@ -83,5 +84,7 @@ func (a ConfigAPI) editSIP(_ *gin.Context, in *conf.SIP) (gin.H, error) {
 	if err := conf.WriteConfig(a.conf, a.conf.ConfigPath); err != nil {
 		return nil, reason.ErrServer.SetMsg(err.Error())
 	}
+	a.uc.SipServer.SetConfig()
+
 	return gin.H{"msg": "ok"}, nil
 }

internal/web/api/ipc.go

@@ -281,7 +281,7 @@ func (a IPCAPI) play(c *gin.Context, _ *struct{}) (*playOutput, error) {
 		host = h
 	}
 
-	item := a.uc.SMSAPI.smsCore.GetStreamLiveAddr(svr, prefix, host, app, stream)
+	item := a.uc.SMSAPI.smsCore.GetStreamLiveAddr(svr, prefix, host, app, appStream)
 	out := playOutput{
 		App:    app,
 		Stream: appStream,

internal/web/api/push.go

@@ -43,13 +43,13 @@ func (a PushAPI) findStreamPush(c *gin.Context, in *push.FindStreamPushInput) (*
 		return nil, err
 	}
 
-	cacheFn := hook.UseCache(func(s string) (*sms.MediaServer, error) {
+	cacheFn := func(s string) (*sms.MediaServer, error) {
 		v, err := a.smsCore.GetMediaServer(c.Request.Context(), s)
 		if err != nil {
 			slog.ErrorContext(c.Request.Context(), "GetMediaServer", "err", err)
 		}
 		return v, err
-	})
+	}
 
 	out := make([]*push.FindStreamPushOutputItem, len(items))
 	for i, item := range items {
@@ -58,8 +58,12 @@ func (a PushAPI) findStreamPush(c *gin.Context, in *push.FindStreamPushInput) (*
 		if mediaID == "" {
 			mediaID = sms.DefaultMediaServerID
 		}
-		if svr, _, _ := cacheFn(mediaID); svr != nil {
-			addr := fmt.Sprintf("rtmp://%s:%d/%s/%s", web.GetHost(c.Request), svr.Ports.RTMP, item.App, item.Stream)
+		if svr, _ := cacheFn(mediaID); svr != nil {
+			port := svr.Ports.RTMP
+			if port == 0 {
+				port = 1935
+			}
+			addr := fmt.Sprintf("rtmp://%s:%d/%s/%s", web.GetHost(c.Request), port, item.App, item.Stream)
 			if !item.IsAuthDisabled {
 				addr += fmt.Sprintf("?sign=%s", hook.MD5(a.conf.Server.RTMPSecret))
 			}

pkg/gbs/server.go

@@ -49,7 +49,7 @@ func NewServer(cfg *conf.Bootstrap, store ipc.Adapter, sc sms.Core) (*Server, fu
 	iip := ip.InternalIP()
 	uri, _ := sip.ParseSipURI(fmt.Sprintf("sip:%s@%s:%d", cfg.Sip.ID, iip, cfg.Sip.Port))
 	from := sip.Address{
-		DisplayName: sip.String{Str: "gowvp"},
+		DisplayName: sip.String{Str: "gowvp/owl"},
 		URI:         &uri,
 		Params:      sip.NewParams(),
 	}
@@ -87,6 +87,20 @@ func NewServer(cfg *conf.Bootstrap, store ipc.Adapter, sc sms.Core) (*Server, fu
 	return &c, c.Close
 }
 
+// SetConfig 热更新 SIP 配置，用于配置变更时更新 from 地址而无需重启服务
+func (s *Server) SetConfig() {
+	cfg := s.gb.cfg
+	iip := ip.InternalIP()
+	uri, _ := sip.ParseSipURI(fmt.Sprintf("sip:%s@%s:%d", cfg.ID, iip, cfg.Port))
+	from := sip.Address{
+		DisplayName: sip.String{Str: "gowvp/owl"},
+		URI:         &uri,
+		Params:      sip.NewParams(),
+	}
+	s.fromAddress = from
+	s.Server.SetFrom(&from)
+}
+
 // startTickerCheck 定时检查离线，通过心跳超时判断设备是否离线
 func (s *Server) startTickerCheck() {
 	conc.Timer(context.Background(), 60*time.Second, time.Second, func() {

pkg/gbs/sip/server.go

@@ -54,6 +54,11 @@ func NewServer(form *Address) *Server {
 	return srv
 }
 
+// SetFrom 热更新 SIP 源地址配置，用于配置变更时无需重启服务
+func (s *Server) SetFrom(from *Address) {
+	*s.from = *from
+}
+
 func (s *Server) addRoute(method string, handler ...HandlerFunc) {
 	s.route.Store(strings.ToUpper(method), handler)
 }

pkg/zlm/config.go

@@ -310,42 +310,43 @@ type SetServerConfigRequest struct {
 	RtcStartBitrate                    *string `json:"rtc.start_bitrate,omitempty"`
 	RtcTCPPort                         *string `json:"rtc.tcpPort,omitempty"`
 	RtcTimeoutSec                      *string `json:"rtc.timeoutSec,omitempty"`
-	RtmpDirectProxy                    *string `json:"rtmp.directProxy,omitempty"`
-	RtmpEnhanced                       *string `json:"rtmp.enhanced,omitempty"`
-	RtmpHandshakeSecond                *string `json:"rtmp.handshakeSecond,omitempty"`
-	RtmpKeepAliveSecond                *string `json:"rtmp.keepAliveSecond,omitempty"`
-	RtmpPort                           *string `json:"rtmp.port,omitempty"`
-	RtmpSslport                        *string `json:"rtmp.sslport,omitempty"`
-	RtpAudioMtuSize                    *string `json:"rtp.audioMtuSize,omitempty"`
-	RtpH264StapA                       *string `json:"rtp.h264_stap_a,omitempty"`
-	RtpLowLatency                      *string `json:"rtp.lowLatency,omitempty"`
-	RtpRtpMaxSize                      *string `json:"rtp.rtpMaxSize,omitempty"`
-	RtpVideoMtuSize                    *string `json:"rtp.videoMtuSize,omitempty"`
-	RtpProxyDumpDir                    *string `json:"rtp_proxy.dumpDir,omitempty"`
-	RtpProxyGopCache                   *string `json:"rtp_proxy.gop_cache,omitempty"`
-	RtpProxyH264Pt                     *string `json:"rtp_proxy.h264_pt,omitempty"`
-	RtpProxyH265Pt                     *string `json:"rtp_proxy.h265_pt,omitempty"`
-	RtpProxyOpusPt                     *string `json:"rtp_proxy.opus_pt,omitempty"`
-	RtpProxyPort                       *string `json:"rtp_proxy.port,omitempty"`
-	RtpProxyPortRange                  *string `json:"rtp_proxy.port_range,omitempty"`
-	RtpProxyPsPt                       *string `json:"rtp_proxy.ps_pt,omitempty"`
-	RtpProxyRtpG711DurMs               *string `json:"rtp_proxy.rtp_g711_dur_ms,omitempty"`
-	RtpProxyTimeoutSec                 *string `json:"rtp_proxy.timeoutSec,omitempty"`
-	RtpProxyUDPRecvSocketBuffer        *string `json:"rtp_proxy.udp_recv_socket_buffer,omitempty"`
-	RtspAuthBasic                      *string `json:"rtsp.authBasic,omitempty"`
-	RtspDirectProxy                    *string `json:"rtsp.directProxy,omitempty"`
-	RtspHandshakeSecond                *string `json:"rtsp.handshakeSecond,omitempty"`
-	RtspKeepAliveSecond                *string `json:"rtsp.keepAliveSecond,omitempty"`
-	RtspLowLatency                     *string `json:"rtsp.lowLatency,omitempty"`
-	RtspPort                           *string `json:"rtsp.port,omitempty"`
-	RtspRtpTransportType               *string `json:"rtsp.rtpTransportType,omitempty"`
-	RtspSslport                        *string `json:"rtsp.sslport,omitempty"`
-	ShellMaxReqSize                    *string `json:"shell.maxReqSize,omitempty"`
-	ShellPort                          *string `json:"shell.port,omitempty"`
-	SrtLatencyMul                      *string `json:"srt.latencyMul,omitempty"`
-	SrtPktBufSize                      *string `json:"srt.pktBufSize,omitempty"`
-	SrtPort                            *string `json:"srt.port,omitempty"`
-	SrtTimeoutSec                      *string `json:"srt.timeoutSec,omitempty"`
+	// RTCEnableTurn                      *string `json:"rtc.enableTurn,omitempty"`
+	RtmpDirectProxy             *string `json:"rtmp.directProxy,omitempty"`
+	RtmpEnhanced                *string `json:"rtmp.enhanced,omitempty"`
+	RtmpHandshakeSecond         *string `json:"rtmp.handshakeSecond,omitempty"`
+	RtmpKeepAliveSecond         *string `json:"rtmp.keepAliveSecond,omitempty"`
+	RtmpPort                    *string `json:"rtmp.port,omitempty"`
+	RtmpSslport                 *string `json:"rtmp.sslport,omitempty"`
+	RtpAudioMtuSize             *string `json:"rtp.audioMtuSize,omitempty"`
+	RtpH264StapA                *string `json:"rtp.h264_stap_a,omitempty"`
+	RtpLowLatency               *string `json:"rtp.lowLatency,omitempty"`
+	RtpRtpMaxSize               *string `json:"rtp.rtpMaxSize,omitempty"`
+	RtpVideoMtuSize             *string `json:"rtp.videoMtuSize,omitempty"`
+	RtpProxyDumpDir             *string `json:"rtp_proxy.dumpDir,omitempty"`
+	RtpProxyGopCache            *string `json:"rtp_proxy.gop_cache,omitempty"`
+	RtpProxyH264Pt              *string `json:"rtp_proxy.h264_pt,omitempty"`
+	RtpProxyH265Pt              *string `json:"rtp_proxy.h265_pt,omitempty"`
+	RtpProxyOpusPt              *string `json:"rtp_proxy.opus_pt,omitempty"`
+	RtpProxyPort                *string `json:"rtp_proxy.port,omitempty"`
+	RtpProxyPortRange           *string `json:"rtp_proxy.port_range,omitempty"`
+	RtpProxyPsPt                *string `json:"rtp_proxy.ps_pt,omitempty"`
+	RtpProxyRtpG711DurMs        *string `json:"rtp_proxy.rtp_g711_dur_ms,omitempty"`
+	RtpProxyTimeoutSec          *string `json:"rtp_proxy.timeoutSec,omitempty"`
+	RtpProxyUDPRecvSocketBuffer *string `json:"rtp_proxy.udp_recv_socket_buffer,omitempty"`
+	RtspAuthBasic               *string `json:"rtsp.authBasic,omitempty"`
+	RtspDirectProxy             *string `json:"rtsp.directProxy,omitempty"`
+	RtspHandshakeSecond         *string `json:"rtsp.handshakeSecond,omitempty"`
+	RtspKeepAliveSecond         *string `json:"rtsp.keepAliveSecond,omitempty"`
+	RtspLowLatency              *string `json:"rtsp.lowLatency,omitempty"`
+	RtspPort                    *string `json:"rtsp.port,omitempty"`
+	RtspRtpTransportType        *string `json:"rtsp.rtpTransportType,omitempty"`
+	RtspSslport                 *string `json:"rtsp.sslport,omitempty"`
+	ShellMaxReqSize             *string `json:"shell.maxReqSize,omitempty"`
+	ShellPort                   *string `json:"shell.port,omitempty"`
+	SrtLatencyMul               *string `json:"srt.latencyMul,omitempty"`
+	SrtPktBufSize               *string `json:"srt.pktBufSize,omitempty"`
+	SrtPort                     *string `json:"srt.port,omitempty"`
+	SrtTimeoutSec               *string `json:"srt.timeoutSec,omitempty"`
 }
 
 func (e *Engine) GetServerConfig() (*GetServerConfigResponse, error) {