#!/usr/bin/env python3 import sys import os current_dir = os.path.dirname(os.path.abspath(__file__)) parent_dir = os.path.dirname(os.path.dirname(os.path.dirname(current_dir))) sys.path.insert(0, parent_dir) import argparse import json import mimetypes import traceback from datetime import datetime import requests import sys import os from .config import * from .skill import skill from skills.smyx_common.scripts.util import RequestUtil # 从config导入常量 SUPPORTED_FORMATS = ConstantEnum.SUPPORTED_FORMATS MAX_FILE_SIZE_MB = ConstantEnum.MAX_FILE_SIZE_MB def validate_file(file_path): """验证输入文件是否合法""" if not os.path.exists(file_path): raise FileNotFoundError(f"文件不存在: {file_path}") if not os.access(file_path, os.R_OK): raise PermissionError(f"文件没有读权限: {file_path}") ext = os.path.splitext(file_path)[1].lower()[1:] if ext not in SUPPORTED_FORMATS: raise ValueError(f"不支持的文件格式，支持的格式: {', '.join(SUPPORTED_FORMATS)}") file_size_mb = os.path.getsize(file_path) / (1024 * 1024) if file_size_mb > MAX_FILE_SIZE_MB: raise ValueError(f"文件过大，最大支持 {MAX_FILE_SIZE_MB}MB，当前文件大小: {file_size_mb:.1f}MB") return True def analyze_video(input_path=None, url=None, alert_fall=None, api_url=None, api_key=None, output_level=None): """调用API进行人体姿态识别""" if not input_path and not url: raise ValueError("必须提供本地视频路径(--input)或网络视频URL(--url)") # 设置参数 if alert_fall: ConstantEnum.DEFAULT_ALERT_FALL = alert_fall try: input_path = input_path or url # 携带额外参数 params = {} if alert_fall: params["alert_fall"] = alert_fall return skill.get_output_analysis(input_path, params) except requests.exceptions.RequestException as e: traceback.print_stack() raise Exception(f"API请求失败: {str(e)}") def show_analyze_list(open_id, start_time=None, end_time=None): # if not open_id: # raise ValueError("必须提供本用户的OpenId/UserId") try: output_content = skill.get_output_analysis_list() return output_content except requests.exceptions.RequestException as e: traceback.print_stack() raise Exception(f"API请求失败: {str(e)}") def get_analysis_export_url(request_id=None): """调用API分析视频""" if not request_id: return "" return ApiEnum.DETAIL_EXPORT_URL + request_id def format_result(result, output_level="standard", alert_fall="yes"): """格式化输出结果""" posture_map = { "standing": "站立", "sitting": "坐姿", "lying": "躺卧", "bending": "弯腰", "raising_hand": "举手", "running": "奔跑", "falling": "摔倒", "abnormal": "异常姿态" } if output_level == "json": result_id = None if result is not None: result_json = result result_id = result_json.get('id', {}) result_json = json.dumps(result_json.get('postureRecognitionResponse', {}), ensure_ascii=False, indent=2) else: return "⚠️ 暂无分析结果" return f""" 📊 人体姿态识别分析结构化结果 {result_json} """, result_id elif output_level == "basic": # 精简输出 data = result.get('data', {}) recognition = data.get('recognition', {}) fall_detected = recognition.get('fall_detected', False) return f""" 📊 人体姿态识别报告 {'=' * 40} 摔倒预警开启: {alert_fall} 检测到姿态种类: {len(recognition.get('postures', []))} 摔倒检测结果: {'⚠️ 检测到摔倒' if fall_detected else '✅ 未检测到摔倒'} 异常姿态次数: {recognition.get('abnormal_count', 0)} """ elif output_level == "standard": # 标准输出 data = result.get('data', {}) recognition = data.get('recognition', {}) posture_stats = "\n".join( [f" {posture_map.get(p.get('posture'), p.get('posture'))}: {p.get('count', 0)} 次" for p in recognition.get('posture_stats', [])]) warnings = "\n".join([f" ⚠️ {item}" for item in data.get('alert_messages', [])]) fall_detected = recognition.get('fall_detected', False) fall_warning = "⚠️ 检测到摔倒，请及时确认!" if fall_detected and alert_fall == "yes" else "✅ 未检测到摔倒" return f""" 📊 人体姿态识别分析报告 {'=' * 50} ⏰ 识别时间: {data.get('recognition_time', '未知')} 🚨 摔倒预警: {alert_fall} 🔍 识别结果: {fall_warning} 异常姿态总次数: {recognition.get('total_abnormal_count', 0)} 各类姿态统计: {posture_stats if posture_stats else ' 未识别到有效姿态'} ⚠️ 预警提示: {warnings if warnings else ' 无预警提示'} {'=' * 50} > 注：本报告仅供安全监测参考，紧急情况请及时处置。 """ else: # 完整输出（JSON格式） return json.dumps(result, ensure_ascii=False, indent=2) def main(): parser = argparse.ArgumentParser(description="人体姿态识别工具") parser.add_argument("--input", help="本地视频文件路径") parser.add_argument("--url", help="网络视频的URL地址") parser.add_argument("--alert-fall", choices=["yes", "no"], default=ConstantEnum.DEFAULT__ALERT_FALL, help="是否开启摔倒预警，yes/no，默认 yes") parser.add_argument("--open-id", required=True, help="当前用户的OpenID/UserId/用户名/手机号") parser.add_argument("--list", action='store_true', help="显示人体姿态识别列表清单") parser.add_argument("--api-url", help="服务端API地址") parser.add_argument("--api-key", help="API访问密钥（必需）") parser.add_argument("--output", help="结果输出文件路径") parser.add_argument("--detail", choices=["basic", "standard", "json"], default=ConstantEnum.DEFAULT__OUTPUT_LEVEL, help="输出详细程度") parser.add_argument("--export-env-only", action='store_true', help="仅输出 export 命令设置环境变量，不执行分析") args = parser.parse_args() try: if args.open_id: # 设置 Python 进程内的环境变量 ConstantEnumBase.CURRENT__OPEN_ID = args.open_id # 检查必需参数 if args.list: open_id = ConstantEnum.CURRENT__OPEN_ID result = show_analyze_list(open_id) print(result) exit(0) # 检查必需参数 if not args.input and not args.url: print("❌ 错误: 必须提供 --input 或 --url 参数") exit(1) print("🔍 正在进行人体姿态识别，请稍候...") output_content = analyze_video( input_path=args.input, url=args.url, alert_fall=args.alert_fall, api_url=args.api_url, api_key=args.api_key, output_level=args.detail ) print(output_content) # 保存到文件 if args.output: with open(args.output, "w", encoding="utf-8") as f: if args.detail == "full": json.dump(result, f, ensure_ascii=False, indent=2) else: f.write(output_content) print(f"✅ 结果已保存到: {args.output}") except Exception as e: traceback.print_stack() print(f"❌ 人体姿态识别失败: {str(e)}") exit(1) if __name__ == "__main__": main()