import os import cv2 import torch import numpy as np from redis import Redis from ultralytics import YOLO import json from kafka import KafkaConsumer import threading import redis import torch from config import * torch.cuda.set_device(1) # 配置 MODEL_PATH = FACE_MODEL_PATH KAFKA_BROKER = KAFKA_BROKER KAFKA_TOPIC = WORKER_CONFIGS["face"]["kafka_topic"] KAFKA_GROUP_ID = f"face_{KAFKA_GROUP_ID_PREFIX}" REDIS_HOST = REDIS_HOST REDIS_PORT = REDIS_PORT REDIS_PASSWORD = REDIS_PASSWORD REDIS_DB = WORKER_CONFIGS["face"]["redis_db"] # Worker使用的Redis DB MAIN_REDIS_DB = MAIN_REDIS_DB # 主Redis DB UPLOAD_DIR = UPLOAD_DIR RESULT_DIR = RESULT_DIR # 初始化 Kafka consumer = KafkaConsumer( KAFKA_TOPIC, bootstrap_servers=[KAFKA_BROKER], group_id=KAFKA_GROUP_ID, auto_offset_reset='earliest', enable_auto_commit=True, value_deserializer=lambda x: json.loads(x.decode('utf-8')) ) # 初始化 Redis redis_client = Redis( host=REDIS_HOST, port=REDIS_PORT, password=REDIS_PASSWORD, db=REDIS_DB ) main_redis_client = Redis( host=REDIS_HOST, port=REDIS_PORT, password=REDIS_PASSWORD, db=MAIN_REDIS_DB ) class faceDetector: def __init__(self, model_path): self.model = YOLO(model_path).to('cuda:1') def detect(self, frame): results = self.model(frame, device='cuda:1') return results def format_results(self, results, original_shape): formatted_results = [] for r in results: boxes = r.boxes keypoints = r.keypoints for i in range(len(boxes)): box = boxes[i] kpts = keypoints[i] # 调整边界框坐标以适应原始图像大小 orig_h, orig_w = original_shape[:2] model_h, model_w = r.orig_shape scale_x, scale_y = orig_w / model_w, orig_h / model_h bbox = box.xyxy[0].cpu().numpy() bbox_scaled = [ bbox[0] * scale_x, bbox[1] * scale_y, bbox[2] * scale_x, bbox[3] * scale_y ] # 调整关键点坐标以适应原始图像大小 kpts_scaled = kpts.xy[0].cpu().numpy() * np.array([scale_x, scale_y]) formatted_results.append({ "bbox": bbox_scaled, "confidence": box.conf.item(), "keypoints": kpts_scaled.tolist() }) return formatted_results def draw_results(self, frame, results): annotated_frame = frame.copy() for r in results: bbox = r["bbox"] keypoints = r["keypoints"] # 绘制边界框 cv2.rectangle(annotated_frame, (int(bbox[0]), int(bbox[1])), (int(bbox[2]), int(bbox[3])), (0, 255, 0), 2) # 绘制关键点 for kp in keypoints: cv2.circle(annotated_frame, (int(kp[0]), int(kp[1])), 5, (255, 0, 0), -1) return annotated_frame detector = faceDetector(MODEL_PATH) def process_image(image_path): try: original_img = cv2.imread(image_path) original_shape = original_img.shape img = cv2.cvtColor(original_img, cv2.COLOR_BGR2RGB) img = cv2.resize(img, (640, 640)) img = img.transpose((2, 0, 1)) img = np.ascontiguousarray(img) img = torch.from_numpy(img).float() img /= 255.0 img = img.unsqueeze(0) results = detector.detect(img) json_results = detector.format_results(results, original_shape) annotated_img = detector.draw_results(original_img, json_results) return json_results, annotated_img except Exception as e: print(f"处理图像时出错: {str(e)}") return None, None def process_video(video_path): try: cap = cv2.VideoCapture(video_path) frame_count = 0 json_results = [] fps = int(cap.get(cv2.CAP_PROP_FPS)) width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH)) height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT)) original_shape = (height, width) out = cv2.VideoWriter(video_path.replace(UPLOAD_DIR, RESULT_DIR), cv2.VideoWriter_fourcc(*'mp4v'), fps, (width, height)) while cap.isOpened(): ret, frame = cap.read() if not ret: break if frame_count % fps == 0: preprocessed_frame = preprocess_frame(frame) results = detector.detect(preprocessed_frame) frame_json_results = detector.format_results(results, original_shape) json_results.append({"frame": frame_count, "detections": frame_json_results}) annotated_frame = detector.draw_results(frame, frame_json_results) out.write(annotated_frame) frame_count += 1 cap.release() out.release() return json_results except Exception as e: print(f"处理视频时出错: {str(e)}") return None def preprocess_frame(frame): frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB) frame_resized = cv2.resize(frame_rgb, (640, 640)) frame_transposed = frame_resized.transpose((2, 0, 1)) frame_contiguous = np.ascontiguousarray(frame_transposed) frame_tensor = torch.from_numpy(frame_contiguous).float() frame_normalized = frame_tensor / 255.0 frame_batched = frame_normalized.unsqueeze(0) return frame_batched def process_task(): print("开始处理任务,等待Kafka消息...") for message in consumer: print(f"收到Kafka消息: topic={message.topic}, partition={message.partition}, offset={message.offset}") task = message.value task_id = task['task_id'] filename = task['filename'] file_type = task['file_type'] print(f"解析任务信息: ID={task_id}, 文件名={filename}, 类型={file_type}") file_path = os.path.join(UPLOAD_DIR, filename) # 检查键类型并更新状态 task_key = f"task:{task_id}" try: key_type = main_redis_client.type(task_key) if key_type != b'hash': main_redis_client.delete(task_key) main_redis_client.hset(task_key, "status", "processing") print(f"任务 {task_id} 状态更新为 'processing'") except redis.exceptions.ResponseError as e: print(f"更新任务 {task_id} 状态时出错: {str(e)}") continue # 跳过这个任务，继续处理下一个 try: if file_type == "image": print(f"开始处理图像: {filename}") json_results, annotated_img = process_image(file_path) if json_results and annotated_img is not None: result_filename = f"face_{filename}" result_path = os.path.join(RESULT_DIR, result_filename) cv2.imwrite(result_path, annotated_img) redis_client.hmset(f"face_result:{task_id}", { "result": json.dumps(json_results), "result_file": result_filename }) main_redis_client.hmset(f"task:{task_id}", { "status": "completed", "result_type": "face", "result_key": f"face_result:{task_id}" }) print(f"图像 {filename} 处理完成,结果已保存") else: print(f"图像 {filename} 处理失败") main_redis_client.hset(f"task:{task_id}", "status", "failed") else: # video print(f"开始处理视频: {filename}") json_results = process_video(file_path) if json_results: result_filename = f"face_{filename}" redis_client.hmset(f"face_result:{task_id}", { "result": json.dumps(json_results), "result_file": result_filename }) main_redis_client.hmset(f"task:{task_id}", { "status": "completed", "result_type": "face", "result_key": f"face_result:{task_id}" }) print(f"视频 {filename} 处理完成,结果已保存") else: print(f"视频 {filename} 处理失败") main_redis_client.hset(f"task:{task_id}", "status", "failed") except Exception as e: print(f"处理任务 {task_id} 时出错: {str(e)}") main_redis_client.hmset(f"task:{task_id}", { "status": "failed", "error": str(e) }) print(f"任务 {task_id} 处理完毕,等待下一个Kafka消息...") def listen_redis_changes(): pubsub = redis_client.pubsub() pubsub.psubscribe('__keyspace@3__:face_result:*') # 监听所有face_result键的变化 for message in pubsub.listen(): if message['type'] == 'pmessage': key = message['channel'].decode('utf-8').split(':')[-1] operation = message['data'].decode('utf-8') if operation == 'hmset': value = redis_client.hgetall(f"face_result:{key}") if value: result = {k.decode(): v.decode() for k, v in value.items()} print(f"Result update for task {key}: {result}") if __name__ == "__main__": print("face处理程序启动...") # 启动处理任务的线程 task_thread = threading.Thread(target=process_task, daemon=True) task_thread.start() print("任务处理线程已启动") # 启动Redis监听线程 redis_thread = threading.Thread(target=listen_redis_changes, daemon=True) redis_thread.start() print("Redis监听线程已启动") print("主程序进入等待状态...") # 保持主线程运行 task_thread.join() redis_thread.join()