识别结果保存到对应目录下
This commit is contained in:
175
core/all.py
175
core/all.py
@ -1,139 +1,70 @@
|
||||
import cv2
|
||||
from core.ocr import load_model as ocrLoadModel, detect as ocrDetect
|
||||
from core.face import load_model as faceLoadModel, detect as faceDetect
|
||||
from core.yolo import load_model as yoloLoadModel, detect as yoloDetect
|
||||
from concurrent.futures import ThreadPoolExecutor, Future
|
||||
import threading
|
||||
import cv2
|
||||
import numpy as np
|
||||
|
||||
# -------------------------- 核心配置参数 --------------------------
|
||||
MAX_WORKERS = 6 # 线程池最大线程数
|
||||
DETECTION_ORDER = ["yolo", "face", "ocr"] # 检测执行顺序
|
||||
TIMEOUT = 30 # 检测超时时间(秒) 【确保此常量可被外部导入】
|
||||
|
||||
# -------------------------- 全局状态管理 --------------------------
|
||||
_executor = None # 线程池实例
|
||||
_model_loaded = False # 模型加载状态标记
|
||||
_model_lock = threading.Lock() # 模型加载线程锁
|
||||
_executor_lock = threading.Lock() # 线程池初始化锁
|
||||
_task_counter = 0 # 任务计数器
|
||||
_task_counter_lock = threading.Lock() # 任务计数锁
|
||||
# 导入保存路径函数（根据实际文件位置调整导入路径）
|
||||
from core.establish import get_image_save_path
|
||||
# 模型加载状态标记（避免重复加载）
|
||||
|
||||
|
||||
# -------------------------- 工具函数 --------------------------
|
||||
def _get_next_task_id():
|
||||
"""获取唯一任务ID、用于日志追踪"""
|
||||
global _task_counter
|
||||
with _task_counter_lock:
|
||||
_task_counter += 1
|
||||
return _task_counter
|
||||
_model_loaded = False
|
||||
|
||||
|
||||
# -------------------------- 模型加载 --------------------------
|
||||
def load_model():
|
||||
"""加载所有检测模型并初始化线程池（仅执行一次）"""
|
||||
"""加载所有检测模型（仅首次调用时执行）"""
|
||||
global _model_loaded
|
||||
if not _model_loaded:
|
||||
with _model_lock:
|
||||
if not _model_loaded:
|
||||
print("=== 开始加载检测模型 ===")
|
||||
if _model_loaded:
|
||||
print("模型已加载，无需重复执行")
|
||||
return
|
||||
|
||||
# 按顺序加载模型
|
||||
print("加载YOLO模型...")
|
||||
yoloLoadModel()
|
||||
# 依次加载OCR、人脸、YOLO模型
|
||||
ocrLoadModel()
|
||||
faceLoadModel()
|
||||
yoloLoadModel()
|
||||
|
||||
print("加载人脸检测模型...")
|
||||
faceLoadModel()
|
||||
|
||||
print("加载OCR模型...")
|
||||
ocrLoadModel()
|
||||
|
||||
_model_loaded = True
|
||||
print("=== 所有模型加载完成 ===")
|
||||
|
||||
# 初始化线程池
|
||||
_init_thread_pool()
|
||||
_model_loaded = True
|
||||
print("所有检测模型加载完成")
|
||||
|
||||
|
||||
# -------------------------- 线程池管理 --------------------------
|
||||
def _init_thread_pool():
|
||||
"""初始化线程池（仅内部调用）"""
|
||||
global _executor
|
||||
with _executor_lock:
|
||||
if _executor is None:
|
||||
_executor = ThreadPoolExecutor(
|
||||
max_workers=MAX_WORKERS,
|
||||
thread_name_prefix="DetectionThread"
|
||||
)
|
||||
print(f"=== 线程池初始化完成、最大线程数: {MAX_WORKERS} ===")
|
||||
|
||||
|
||||
def shutdown():
|
||||
"""关闭线程池、释放资源"""
|
||||
global _executor
|
||||
with _executor_lock:
|
||||
if _executor is not None:
|
||||
_executor.shutdown(wait=True)
|
||||
_executor = None
|
||||
print("=== 线程池已安全关闭 ===")
|
||||
|
||||
|
||||
# -------------------------- 检测逻辑实现 --------------------------
|
||||
def _detect_in_thread(frame: np.ndarray, task_id: int) -> tuple:
|
||||
"""在子线程中执行检测逻辑（返回4元素tuple：检测是否成功、结果数据、检测器类型、任务ID）"""
|
||||
thread_name = threading.current_thread().name
|
||||
print(f"任务[{task_id}] 开始执行、线程: {thread_name}")
|
||||
|
||||
try:
|
||||
# 按照配置顺序执行检测
|
||||
for detector in DETECTION_ORDER:
|
||||
if detector == "yolo":
|
||||
success, result = yoloDetect(frame)
|
||||
elif detector == "face":
|
||||
success, result = faceDetect(frame)
|
||||
elif detector == "ocr":
|
||||
success, result = ocrDetect(frame)
|
||||
else:
|
||||
success, result = False, None
|
||||
|
||||
print(f"任务[{task_id}] {detector}检测状态: {'成功' if success else '未检测到内容'}")
|
||||
if success:
|
||||
print(f"任务[{task_id}] 完成检测、使用检测器: {detector}")
|
||||
return (success, result, detector, task_id) # 4元素tuple
|
||||
|
||||
# 所有检测器均未检测到结果
|
||||
print(f"任务[{task_id}] 所有检测器均未检测到有效内容")
|
||||
return (False, "未检测到任何有效内容", "none", task_id) # 4元素tuple
|
||||
|
||||
except Exception as e:
|
||||
print(f"任务[{task_id}] 检测过程发生错误: {str(e)}")
|
||||
return (False, f"检测错误: {str(e)}", "error", task_id) # 4元素tuple
|
||||
|
||||
|
||||
# -------------------------- 外部调用接口 --------------------------
|
||||
def detect(frame: np.ndarray) -> Future:
|
||||
def detect(frame):
|
||||
"""
|
||||
提交检测任务到线程池（返回Future对象，需调用result()获取4元素结果）
|
||||
|
||||
参数:
|
||||
frame: 待检测图像(ndarray格式、cv2.imdecode生成)
|
||||
|
||||
返回:
|
||||
Future对象、result()返回tuple: (success, data, detector_type, task_id)
|
||||
success: 布尔值，表示是否检测到有效内容
|
||||
data: 检测结果数据（成功时为具体结果，失败时为错误信息）
|
||||
detector_type: 使用的检测器类型（"yolo"/"face"/"ocr"/"none"/"error"）
|
||||
task_id: 任务唯一标识
|
||||
执行模型检测，检测到违规时按指定格式保存图片
|
||||
参数：
|
||||
frame: 待检测的图像帧（OpenCV格式，numpy.ndarray类型）
|
||||
返回：
|
||||
(检测结果布尔值, 检测详情, 检测模型类型)
|
||||
"""
|
||||
# 确保模型已加载
|
||||
if not _model_loaded:
|
||||
print("警告: 模型尚未加载、将自动加载")
|
||||
load_model()
|
||||
# 1. YOLO检测（优先级1）
|
||||
yolo_flag, yolo_result = yoloDetect(frame)
|
||||
print(f"YOLO检测结果：{yolo_result}")
|
||||
if yolo_flag:
|
||||
# 直接调用路径生成函数，无需传入原始图片名
|
||||
save_path = get_image_save_path(model_type="yolo")
|
||||
if save_path:
|
||||
cv2.imwrite(save_path, frame)
|
||||
print(f"✅ YOLO违规图片已保存：{save_path}")
|
||||
return (True, yolo_result, "yolo")
|
||||
|
||||
# 生成任务ID
|
||||
task_id = _get_next_task_id()
|
||||
# 2. 人脸检测（优先级2）
|
||||
face_flag, face_result = faceDetect(frame)
|
||||
print(f"人脸检测结果：{face_result}")
|
||||
if face_flag:
|
||||
save_path = get_image_save_path(model_type="face")
|
||||
if save_path:
|
||||
cv2.imwrite(save_path, frame)
|
||||
print(f"✅ 人脸违规图片已保存：{save_path}")
|
||||
return (True, face_result, "face")
|
||||
|
||||
# 提交任务到线程池（返回Future）
|
||||
future = _executor.submit(_detect_in_thread, frame, task_id)
|
||||
print(f"任务[{task_id}]: 已提交到线程池")
|
||||
return future
|
||||
# 3. OCR检测（优先级3）
|
||||
ocr_flag, ocr_result = ocrDetect(frame)
|
||||
print(f"OCR检测结果：{ocr_result}")
|
||||
if ocr_flag:
|
||||
save_path = get_image_save_path(model_type="ocr")
|
||||
if save_path:
|
||||
cv2.imwrite(save_path, frame)
|
||||
print(f"✅ OCR违规图片已保存：{save_path}")
|
||||
return (True, ocr_result, "ocr")
|
||||
|
||||
# 4. 无违规内容（不保存图片）
|
||||
print(f"❌ 未检测到任何违规内容，不保存图片")
|
||||
return (False, "未检测到任何内容", "none")
|
||||
111
core/establish.py
Normal file
111
core/establish.py
Normal file
@ -0,0 +1,111 @@
|
||||
import os
|
||||
import datetime
|
||||
from pathlib import Path
|
||||
|
||||
|
||||
# 配置IP文件路径（统一使用绝对路径）
|
||||
IP_FILE_PATH = Path(r"D:\ccc\IP.txt")
|
||||
|
||||
|
||||
def create_directory_structure():
|
||||
"""创建项目所需的目录结构"""
|
||||
try:
|
||||
# 1. 创建根目录下的resource文件夹
|
||||
resource_dir = Path("resource")
|
||||
resource_dir.mkdir(exist_ok=True)
|
||||
print(f"确保resource目录存在: {resource_dir.absolute()}")
|
||||
|
||||
# 2. 在resource下创建dect文件夹
|
||||
dect_dir = resource_dir / "dect"
|
||||
dect_dir.mkdir(exist_ok=True)
|
||||
print(f"确保dect目录存在: {dect_dir.absolute()}")
|
||||
|
||||
# 3. 在dect下创建三个模型文件夹
|
||||
model_dirs = ["ocr", "face", "yolo"]
|
||||
for model in model_dirs:
|
||||
model_dir = dect_dir / model
|
||||
model_dir.mkdir(exist_ok=True)
|
||||
print(f"确保{model}模型目录存在: {model_dir.absolute()}")
|
||||
|
||||
# 4. 读取ip.txt文件获取IP地址
|
||||
try:
|
||||
with open(IP_FILE_PATH, "r") as f:
|
||||
ip_addresses = [line.strip() for line in f if line.strip()]
|
||||
|
||||
if not ip_addresses:
|
||||
print("警告: ip.txt文件中未找到有效的IP地址")
|
||||
return
|
||||
|
||||
print(f"从ip.txt中读取到的IP地址: {ip_addresses}")
|
||||
|
||||
# 5. 获取当前日期
|
||||
now = datetime.datetime.now()
|
||||
current_year = str(now.year)
|
||||
current_month = str(now.month)
|
||||
|
||||
# 6. 为每个IP在每个模型文件夹下创建年->月的目录结构
|
||||
for ip in ip_addresses:
|
||||
# 处理IP地址中的特殊字符（如果有）
|
||||
safe_ip = ip.replace(".", "_")
|
||||
|
||||
for model in model_dirs:
|
||||
# 构建路径: resource/dect/{model}/{ip}/{year}/{month}
|
||||
ip_dir = dect_dir / model / safe_ip
|
||||
year_dir = ip_dir / current_year
|
||||
month_dir = year_dir / current_month
|
||||
|
||||
# 创建目录（如果不存在）
|
||||
month_dir.mkdir(parents=True, exist_ok=True)
|
||||
print(f"创建/确保目录存在: {month_dir.absolute()}")
|
||||
|
||||
except FileNotFoundError:
|
||||
print(f"错误: 未找到ip.txt文件，请确保该文件存在于 {IP_FILE_PATH}")
|
||||
except Exception as e:
|
||||
print(f"处理IP和日期目录时发生错误: {str(e)}")
|
||||
|
||||
except Exception as e:
|
||||
print(f"创建目录结构时发生错误: {str(e)}")
|
||||
|
||||
|
||||
def get_image_save_path(model_type: str) -> str:
|
||||
"""
|
||||
获取图片保存的完整路径（不依赖原始图片名称）
|
||||
|
||||
参数:
|
||||
model_type: 模型类型，应为"ocr"、"face"或"yolo"
|
||||
|
||||
返回:
|
||||
完整的图片保存路径
|
||||
"""
|
||||
try:
|
||||
# 读取IP地址（假设只有一个IP或使用第一个IP）
|
||||
with open(IP_FILE_PATH, "r") as f:
|
||||
ip_addresses = [line.strip() for line in f if line.strip()]
|
||||
|
||||
if not ip_addresses:
|
||||
raise ValueError("ip.txt文件中未找到有效的IP地址")
|
||||
|
||||
ip = ip_addresses[0]
|
||||
safe_ip = ip.replace(".", "_")
|
||||
|
||||
# 获取当前日期和时间（精确到毫秒，确保文件名唯一）
|
||||
now = datetime.datetime.now()
|
||||
current_year = str(now.year)
|
||||
current_month = str(now.month)
|
||||
current_day = str(now.day)
|
||||
# 生成时间戳字符串（格式：年月日时分秒毫秒）
|
||||
timestamp = now.strftime("%Y%m%d%H%M%S%f")[:-3] # 去除最后三位，保留到毫秒
|
||||
|
||||
# 构建路径: resource/dect/{model}/{ip}/{year}/{month}/{day}
|
||||
day_dir = Path("resource") / "dect" / model_type / safe_ip / current_year / current_month / current_day
|
||||
day_dir.mkdir(parents=True, exist_ok=True)
|
||||
|
||||
# 构建图片文件名（使用时间戳确保唯一性）
|
||||
image_filename = f"resource_dect_{model_type}_{safe_ip}_{current_year}_{current_month}_{current_day}_{timestamp}.jpg"
|
||||
image_path = day_dir / image_filename
|
||||
|
||||
return str(image_path)
|
||||
|
||||
except Exception as e:
|
||||
print(f"获取图片保存路径时发生错误: {str(e)}")
|
||||
return ""
|
||||
305
core/face.py
305
core/face.py
@ -6,203 +6,217 @@ import time
|
||||
import threading
|
||||
from PIL import Image
|
||||
from insightface.app import FaceAnalysis
|
||||
# 导入获取人脸信息的服务
|
||||
# 假设service.face_service中get_all_face_name_with_eigenvalue可获取人脸数据
|
||||
from service.face_service import get_all_face_name_with_eigenvalue
|
||||
|
||||
# 用于检查GPU状态
|
||||
# GPU状态检查支持
|
||||
try:
|
||||
import pynvml
|
||||
|
||||
pynvml.nvmlInit()
|
||||
_nvml_available = True
|
||||
except ImportError:
|
||||
print("警告: pynvml库未安装、无法检测GPU状态、将默认使用0号GPU")
|
||||
print("警告: pynvml库未安装，无法检测GPU状态，默认尝试使用GPU")
|
||||
_nvml_available = False
|
||||
|
||||
# 全局变量
|
||||
# 全局人脸引擎与特征库
|
||||
_face_app = None
|
||||
_known_faces_embeddings = {} # 存储姓名到特征值的映射
|
||||
_known_faces_names = [] # 存储所有已知姓名
|
||||
_using_gpu = False # 标记是否使用GPU
|
||||
_used_gpu_id = -1 # 记录当前使用的GPU ID
|
||||
_known_faces_embeddings = {} # 姓名 -> 归一化特征值的映射
|
||||
_known_faces_names = [] # 已知人脸姓名列表
|
||||
|
||||
# GPU使用状态标记
|
||||
_using_gpu = False # 是否使用GPU
|
||||
_used_gpu_id = -1 # 使用的GPU ID（-1表示CPU）
|
||||
|
||||
# 资源管理变量
|
||||
_ref_count = 0
|
||||
_last_used_time = 0
|
||||
_lock = threading.Lock()
|
||||
_release_timeout = 8 # 5秒无使用则释放
|
||||
_is_releasing = False # 标记是否正在释放
|
||||
_ref_count = 0 # 引擎引用计数（记录当前使用次数）
|
||||
_last_used_time = 0 # 最后一次使用引擎的时间
|
||||
_lock = threading.Lock() # 线程安全锁
|
||||
_release_timeout = 8 # 闲置超时时间（秒）
|
||||
_is_releasing = False # 资源释放中标记
|
||||
_monitor_thread_running = False # 监控线程运行标记
|
||||
|
||||
# 调试用计数器
|
||||
# 调试计数器
|
||||
_debug_counter = {
|
||||
"created": 0,
|
||||
"released": 0,
|
||||
"detected": 0
|
||||
"engine_created": 0, # 引擎创建次数
|
||||
"engine_released": 0, # 引擎释放次数
|
||||
"detection_calls": 0 # 检测函数调用次数
|
||||
}
|
||||
|
||||
|
||||
def check_gpu_availability(gpu_id, threshold=0.7):
|
||||
"""检查指定GPU是否可用（内存使用率低于阈值）"""
|
||||
def check_gpu_availability(gpu_id, memory_threshold=0.7):
|
||||
"""检查指定GPU的内存使用率是否低于阈值（判定为“可用”）"""
|
||||
if not _nvml_available:
|
||||
return True # 无法检测时默认认为可用
|
||||
|
||||
return True
|
||||
try:
|
||||
handle = pynvml.nvmlDeviceGetHandleByIndex(gpu_id)
|
||||
mem_info = pynvml.nvmlDeviceGetMemoryInfo(handle)
|
||||
usage = mem_info.used / mem_info.total
|
||||
# 内存使用率低于阈值则认为可用
|
||||
return usage < threshold
|
||||
memory_usage = mem_info.used / mem_info.total
|
||||
return memory_usage < memory_threshold
|
||||
except Exception as e:
|
||||
print(f"检查GPU {gpu_id} 状态时出错: {e}")
|
||||
print(f"检查GPU {gpu_id} 状态失败: {e}")
|
||||
return False
|
||||
|
||||
|
||||
def select_best_gpu(preferred_gpus=[0, 1]):
|
||||
"""选择最佳可用GPU、严格按照首选列表顺序检查、优先使用0号GPU"""
|
||||
# 首先检查首选GPU列表
|
||||
"""按优先级选择可用GPU，优先0号；均不可用则返回-1（CPU）"""
|
||||
for gpu_id in preferred_gpus:
|
||||
try:
|
||||
# 检查GPU是否存在
|
||||
# 验证GPU是否存在
|
||||
if _nvml_available:
|
||||
pynvml.nvmlDeviceGetHandleByIndex(gpu_id)
|
||||
|
||||
# 检查GPU是否可用
|
||||
# 验证GPU内存是否充足
|
||||
if check_gpu_availability(gpu_id):
|
||||
print(f"GPU {gpu_id} 可用、将使用该GPU")
|
||||
print(f"GPU {gpu_id} 可用，将使用该GPU")
|
||||
return gpu_id
|
||||
else:
|
||||
if gpu_id == 0:
|
||||
print(f"GPU 0 内存使用率过高（繁忙）、尝试切换到其他GPU")
|
||||
print("GPU 0 内存使用率过高，尝试其他GPU")
|
||||
except Exception as e:
|
||||
print(f"GPU {gpu_id} 不存在或无法访问: {e}")
|
||||
continue
|
||||
|
||||
# 如果所有首选GPU都不可用、返回-1表示使用CPU
|
||||
print("所有指定的GPU都不可用、将使用CPU进行计算")
|
||||
print(f"GPU {gpu_id} 不可用或访问失败: {e}")
|
||||
print("所有指定GPU均不可用，将使用CPU计算")
|
||||
return -1
|
||||
|
||||
|
||||
def _release_engine():
|
||||
"""释放人脸识别引擎资源"""
|
||||
def _release_engine_resources():
|
||||
"""释放人脸引擎的所有资源（模型、特征库、GPU缓存等）"""
|
||||
global _face_app, _is_releasing, _known_faces_embeddings, _known_faces_names
|
||||
if not _face_app or _is_releasing:
|
||||
return
|
||||
|
||||
try:
|
||||
_is_releasing = True
|
||||
# 释放InsightFace资源
|
||||
if hasattr(_face_app, 'model'):
|
||||
# 清除模型资源
|
||||
_face_app.model = None
|
||||
_face_app = None
|
||||
print("开始释放人脸引擎资源...")
|
||||
|
||||
# 清空人脸数据
|
||||
# 释放InsightFace模型资源
|
||||
if hasattr(_face_app, "model"):
|
||||
_face_app.model = None # 显式置空模型引用
|
||||
_face_app = None # 释放引擎实例
|
||||
|
||||
# 清空人脸特征库
|
||||
_known_faces_embeddings.clear()
|
||||
_known_faces_names.clear()
|
||||
|
||||
_debug_counter["released"] += 1
|
||||
print(f"Face recognition engine released. Stats: {_debug_counter}")
|
||||
_debug_counter["engine_released"] += 1
|
||||
print(f"人脸引擎已释放，调试统计: {_debug_counter}")
|
||||
|
||||
# 清理GPU缓存
|
||||
# 强制垃圾回收
|
||||
gc.collect()
|
||||
|
||||
# 清理各深度学习框架的GPU缓存
|
||||
# Torch 缓存清理
|
||||
try:
|
||||
import torch
|
||||
if torch.cuda.is_available():
|
||||
torch.cuda.empty_cache()
|
||||
torch.cuda.ipc_collect()
|
||||
print("Torch GPU缓存已清理")
|
||||
except ImportError:
|
||||
pass
|
||||
|
||||
# TensorFlow 缓存清理
|
||||
try:
|
||||
import tensorflow as tf
|
||||
tf.keras.backend.clear_session()
|
||||
print("TensorFlow会话已清理")
|
||||
except ImportError:
|
||||
pass
|
||||
|
||||
# MXNet 缓存清理（InsightFace底层常用MXNet）
|
||||
try:
|
||||
import mxnet as mx
|
||||
mx.nd.waitall() # 等待所有计算完成并释放资源
|
||||
print("MXNet资源已等待释放")
|
||||
except ImportError:
|
||||
pass
|
||||
|
||||
except Exception as e:
|
||||
print(f"释放资源过程中出错: {e}")
|
||||
finally:
|
||||
_is_releasing = False
|
||||
|
||||
|
||||
def _monitor_thread():
|
||||
"""监控线程、检查并释放超时未使用的资源"""
|
||||
global _ref_count, _last_used_time, _face_app
|
||||
while True:
|
||||
time.sleep(5) # 每5秒检查一次
|
||||
def _resource_monitor_thread():
|
||||
"""后台监控线程：检测引擎闲置超时，触发资源释放"""
|
||||
global _ref_count, _last_used_time, _face_app, _monitor_thread_running
|
||||
_monitor_thread_running = True
|
||||
while _monitor_thread_running:
|
||||
time.sleep(2) # 缩短检查间隔，加快闲置检测响应
|
||||
with _lock:
|
||||
# 只有当引擎存在、没有引用且超时、才释放
|
||||
# 当“引擎存在 + 无引用 + 未在释放中”时，检查闲置时间
|
||||
if _face_app and _ref_count == 0 and not _is_releasing:
|
||||
elapsed = time.time() - _last_used_time
|
||||
if elapsed > _release_timeout:
|
||||
print(f"Idle timeout ({elapsed:.1f}s > {_release_timeout}s), releasing face engine")
|
||||
_release_engine()
|
||||
idle_time = time.time() - _last_used_time
|
||||
if idle_time > _release_timeout:
|
||||
print(f"引擎闲置超时（{idle_time:.1f}s > {_release_timeout}s），释放资源")
|
||||
_release_engine_resources()
|
||||
|
||||
|
||||
def load_model(prefer_gpu=True, preferred_gpus=[0, 1]):
|
||||
"""加载人脸识别模型及已知人脸特征库、默认优先使用0号GPU"""
|
||||
"""加载人脸识别引擎及已知人脸特征库（默认优先用0号GPU）"""
|
||||
global _face_app, _known_faces_embeddings, _known_faces_names, _using_gpu, _used_gpu_id
|
||||
|
||||
# 确保监控线程只启动一次
|
||||
if not any(t.name == "FaceMonitor" for t in threading.enumerate()):
|
||||
threading.Thread(target=_monitor_thread, daemon=True, name="FaceMonitor").start()
|
||||
print("Face monitor thread started")
|
||||
# 启动后台监控线程（确保仅启动一次）
|
||||
if not _monitor_thread_running:
|
||||
threading.Thread(
|
||||
target=_resource_monitor_thread,
|
||||
daemon=True,
|
||||
name="FaceEngineMonitor"
|
||||
).start()
|
||||
print("人脸引擎监控线程已启动")
|
||||
|
||||
# 如果正在释放中、等待释放完成
|
||||
# 若正在释放资源，等待释放完成
|
||||
while _is_releasing:
|
||||
time.sleep(0.1)
|
||||
|
||||
# 如果已经初始化、直接返回
|
||||
# 若引擎已初始化，直接返回
|
||||
if _face_app:
|
||||
return True
|
||||
|
||||
# 初始化InsightFace模型
|
||||
# 初始化InsightFace引擎
|
||||
try:
|
||||
# 初始化InsightFace
|
||||
print("正在初始化InsightFace人脸识别引擎...")
|
||||
_face_app = FaceAnalysis(name='buffalo_l', root='~/.insightface')
|
||||
_face_app = FaceAnalysis(name="buffalo_l", root=os.path.expanduser("~/.insightface"))
|
||||
|
||||
# 选择合适的GPU、默认优先使用0号
|
||||
# 选择GPU（优先用0号）
|
||||
ctx_id = 0
|
||||
if prefer_gpu:
|
||||
ctx_id = select_best_gpu(preferred_gpus)
|
||||
_using_gpu = ctx_id != -1
|
||||
_used_gpu_id = ctx_id if _using_gpu else -1
|
||||
|
||||
if _using_gpu:
|
||||
print(f"成功初始化、使用GPU {ctx_id} 进行计算")
|
||||
else:
|
||||
print("成功初始化、使用CPU进行计算")
|
||||
if _using_gpu:
|
||||
print(f"引擎初始化成功，将使用GPU {ctx_id} 计算")
|
||||
else:
|
||||
print("引擎初始化成功，将使用CPU计算")
|
||||
|
||||
# 准备模型
|
||||
# 准备模型（加载到指定设备）
|
||||
_face_app.prepare(ctx_id=ctx_id, det_size=(640, 640))
|
||||
print("InsightFace人脸识别引擎初始化成功。")
|
||||
_debug_counter["created"] += 1
|
||||
print(f"Face engine initialized. Stats: {_debug_counter}")
|
||||
print("InsightFace引擎初始化完成")
|
||||
_debug_counter["engine_created"] += 1
|
||||
print(f"引擎调试统计: {_debug_counter}")
|
||||
|
||||
except Exception as e:
|
||||
print(f"初始化失败: {e}")
|
||||
print(f"引擎初始化失败: {e}")
|
||||
return False
|
||||
|
||||
# 从服务获取所有人脸姓名和特征值
|
||||
# 从服务加载已知人脸的姓名和特征值
|
||||
try:
|
||||
face_data = get_all_face_name_with_eigenvalue()
|
||||
|
||||
# 处理获取到的人脸数据
|
||||
for person_name, eigenvalue_data in face_data.items():
|
||||
# 处理特征值数据 - 兼容数组和字符串两种格式
|
||||
# 兼容“numpy数组”和“字符串”格式的特征值
|
||||
if isinstance(eigenvalue_data, np.ndarray):
|
||||
# 如果已经是numpy数组、直接使用
|
||||
eigenvalue = eigenvalue_data.astype(np.float32)
|
||||
elif isinstance(eigenvalue_data, str):
|
||||
# 清理字符串: 移除方括号、换行符和多余空格
|
||||
cleaned = eigenvalue_data.replace('[', '').replace(']', '').replace('\n', '').strip()
|
||||
# 按空格或逗号分割（处理可能的不同分隔符）
|
||||
values = [v for v in cleaned.split() if v]
|
||||
# 转换为数组
|
||||
# 清理字符串中的括号、换行等干扰符
|
||||
cleaned = eigenvalue_data.replace("[", "").replace("]", "").replace("\n", "").strip()
|
||||
# 分割并转换为浮点数数组
|
||||
values = [v for v in cleaned.split() if v] # 兼容空格/逗号分隔
|
||||
eigenvalue = np.array(list(map(float, values)), dtype=np.float32)
|
||||
else:
|
||||
# 不支持的类型
|
||||
print(f"Unsupported eigenvalue type for {person_name}")
|
||||
print(f"不支持的特征值类型（{type(eigenvalue_data)}），跳过 {person_name}")
|
||||
continue
|
||||
|
||||
# 归一化处理
|
||||
# 特征值归一化（保证后续相似度计算的一致性）
|
||||
norm = np.linalg.norm(eigenvalue)
|
||||
if norm != 0:
|
||||
eigenvalue = eigenvalue / norm
|
||||
@ -210,100 +224,103 @@ def load_model(prefer_gpu=True, preferred_gpus=[0, 1]):
|
||||
_known_faces_embeddings[person_name] = eigenvalue
|
||||
_known_faces_names.append(person_name)
|
||||
|
||||
print(f"成功加载 {len(_known_faces_names)} 个人脸的特征库")
|
||||
|
||||
except Exception as e:
|
||||
print(f"Error loading face data from service: {e}")
|
||||
print(f"加载人脸特征库失败: {e}")
|
||||
|
||||
return True if _face_app else False
|
||||
return _face_app is not None
|
||||
|
||||
|
||||
def detect(frame, threshold=0.4):
|
||||
"""检测并识别人脸、返回结果元组(是否匹配到已知人脸, 结果字符串)"""
|
||||
global _face_app, _known_faces_embeddings, _known_faces_names, _using_gpu, _used_gpu_id
|
||||
global _ref_count, _last_used_time
|
||||
def detect(frame, similarity_threshold=0.4):
|
||||
"""
|
||||
检测并识别人脸
|
||||
返回：(是否匹配到已知人脸, 结果描述字符串)
|
||||
"""
|
||||
global _face_app, _known_faces_embeddings, _known_faces_names, _ref_count, _last_used_time
|
||||
|
||||
# 验证前置条件
|
||||
# 校验输入帧有效性
|
||||
if frame is None or frame.size == 0:
|
||||
return (False, "无效帧数据")
|
||||
return (False, "无效的输入帧数据")
|
||||
|
||||
# 增加引用计数并获取引擎实例
|
||||
# 加锁并更新引用计数、最后使用时间
|
||||
engine = None
|
||||
with _lock:
|
||||
_ref_count += 1
|
||||
_last_used_time = time.time()
|
||||
_debug_counter["detected"] += 1
|
||||
_debug_counter["detection_calls"] += 1
|
||||
|
||||
# 初始化引擎（如果未初始化且不在释放中）
|
||||
# 若引擎未初始化且未在释放中，尝试初始化
|
||||
if not _face_app and not _is_releasing:
|
||||
if not load_model(prefer_gpu=True):
|
||||
_ref_count -= 1 # 恢复引用计数
|
||||
return (False, "引擎初始化失败")
|
||||
# 初始化失败，恢复引用计数
|
||||
with _lock:
|
||||
_ref_count = max(0, _ref_count - 1)
|
||||
return (False, "人脸引擎初始化失败")
|
||||
|
||||
# 获取当前引擎引用
|
||||
engine = _face_app
|
||||
engine = _face_app # 获取引擎引用
|
||||
|
||||
# 检查引擎是否可用
|
||||
if not engine or not _known_faces_names:
|
||||
# 校验引擎可用性
|
||||
if not engine or len(_known_faces_names) == 0:
|
||||
with _lock:
|
||||
_ref_count = max(0, _ref_count - 1)
|
||||
return (False, "人脸识别引擎不可用或未初始化")
|
||||
return (False, "人脸引擎不可用或特征库为空")
|
||||
|
||||
try:
|
||||
# 如果使用GPU、确保输入帧在处理前是连续的数组
|
||||
if _using_gpu and not frame.flags.contiguous:
|
||||
# GPU计算时，确保帧数据是连续内存（避免CUDA错误）
|
||||
if _using_gpu and engine is not None and not frame.flags.contiguous:
|
||||
frame = np.ascontiguousarray(frame)
|
||||
|
||||
faces = _face_app.get(frame)
|
||||
# 执行人脸检测与特征提取
|
||||
faces = engine.get(frame)
|
||||
except Exception as e:
|
||||
print(f"Face detect error: {e}")
|
||||
# 检测到错误时尝试重新选择GPU并重新初始化
|
||||
print("尝试重新选择GPU并重新初始化...")
|
||||
print(f"人脸检测过程出错: {e}")
|
||||
# 出错时尝试重新初始化引擎（可能是GPU状态变化导致）
|
||||
print("尝试重新初始化人脸引擎...")
|
||||
with _lock:
|
||||
_ref_count = max(0, _ref_count - 1)
|
||||
load_model(prefer_gpu=True) # 重新初始化时保持默认GPU优先级
|
||||
load_model(prefer_gpu=True)
|
||||
return (False, f"检测错误: {str(e)}")
|
||||
|
||||
result_parts = []
|
||||
has_matched = False # 标记是否有匹配到的已知人脸
|
||||
has_matched_known_face = False # 是否有任意人脸匹配到已知库
|
||||
|
||||
for face in faces:
|
||||
# 特征归一化
|
||||
embedding = face.embedding.astype(np.float32)
|
||||
norm = np.linalg.norm(embedding)
|
||||
# 归一化当前检测到的人脸特征
|
||||
face_embedding = face.embedding.astype(np.float32)
|
||||
norm = np.linalg.norm(face_embedding)
|
||||
if norm == 0:
|
||||
continue
|
||||
embedding = embedding / norm
|
||||
face_embedding = face_embedding / norm
|
||||
|
||||
# 对比已知人脸
|
||||
max_sim, best_name = -1.0, "Unknown"
|
||||
# 与已知人脸特征逐一比对
|
||||
max_similarity, best_match_name = -1.0, "Unknown"
|
||||
for name in _known_faces_names:
|
||||
known_emb = _known_faces_embeddings[name]
|
||||
sim = np.dot(embedding, known_emb)
|
||||
if sim > max_sim:
|
||||
max_sim = sim
|
||||
best_name = name
|
||||
similarity = np.dot(face_embedding, known_emb) # 余弦相似度
|
||||
if similarity > max_similarity:
|
||||
max_similarity = similarity
|
||||
best_match_name = name
|
||||
|
||||
# 判断匹配结果
|
||||
is_match = max_sim >= threshold
|
||||
if is_match:
|
||||
has_matched = True # 只要有一个匹配成功、就标记为True
|
||||
# 判断是否匹配成功
|
||||
is_matched = max_similarity >= similarity_threshold
|
||||
if is_matched:
|
||||
has_matched_known_face = True
|
||||
|
||||
bbox = face.bbox
|
||||
# 记录该人脸的检测结果
|
||||
bbox = face.bbox # 人脸边界框
|
||||
result_parts.append(
|
||||
f"{'匹配' if is_match else '不匹配'}: {best_name} (相似度: {max_sim:.2f}, 边界框: {bbox})"
|
||||
f"{'匹配' if is_matched else '未匹配'}: {best_match_name} "
|
||||
f"(相似度: {max_similarity:.2f}, 边界框: {bbox.astype(int).tolist()})"
|
||||
)
|
||||
|
||||
# 构建结果字符串
|
||||
if not result_parts:
|
||||
result_str = "未检测到人脸"
|
||||
else:
|
||||
result_str = "; ".join(result_parts)
|
||||
# 构建最终结果字符串
|
||||
result_str = "未检测到人脸" if not result_parts else "; ".join(result_parts)
|
||||
|
||||
# 减少引用计数、确保线程安全
|
||||
# 释放引用计数（线程安全）
|
||||
with _lock:
|
||||
_ref_count = max(0, _ref_count - 1)
|
||||
# 持续使用时更新最后使用时间
|
||||
if _ref_count > 0:
|
||||
_last_used_time = time.time()
|
||||
# 若仍有引用，更新最后使用时间；若引用为0，也立即标记（加快闲置检测）
|
||||
_last_used_time = time.time()
|
||||
|
||||
# 第一个返回值为: 是否匹配到已知人脸
|
||||
return (has_matched, result_str)
|
||||
return (has_matched_known_face, result_str)
|
||||
14
core/ocr.py
14
core/ocr.py
@ -167,7 +167,19 @@ def detect(frame):
|
||||
items_to_process = [line]
|
||||
|
||||
for item in items_to_process:
|
||||
# 跳过纯数字列表（可能是坐标信息）
|
||||
# 精确识别并忽略图片坐标位置信息 [[x1,y1], [x2,y2], [x3,y3], [x4,y4]]
|
||||
if isinstance(item, list) and len(item) == 4: # 四边形有4个顶点
|
||||
is_coordinate = True
|
||||
for point in item:
|
||||
# 每个顶点应该是包含2个数字的列表
|
||||
if not (isinstance(point, list) and len(point) == 2 and
|
||||
all(isinstance(coord, (int, float)) for coord in point)):
|
||||
is_coordinate = False
|
||||
break
|
||||
if is_coordinate:
|
||||
continue # 是坐标信息，直接忽略
|
||||
|
||||
# 跳过纯数字列表（其他可能的坐标形式）
|
||||
if isinstance(item, list) and all(isinstance(x, (int, float)) for x in item):
|
||||
continue
|
||||
|
||||
|
||||
Reference in New Issue
Block a user