识别结果保存到对应目录下

This commit is contained in:
2025-09-09 16:30:12 +08:00
parent 0fe49bf829
commit 532a9e75e9
6 changed files with 375 additions and 325 deletions

View File

@ -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")