视频内容分析系统Video Content Analysis System

视频内容分析系统面向各类视频业务场景,将视频里的画面、文字等杂乱的非结构化数据转化为结构化信息。基于自研先进视觉模型,系统可对视频中人物、事物、文本等任意视觉元素进行高精度解析,输出标准的结构化标签,配套可视化数据图表,实现视频内容的智能化分析。The Video Content Analysis System turns unstructured data in video, including imagery and text, into structured information for a wide range of video applications. Powered by an advanced in-house vision model, the system analyzes visual elements such as people, objects, and text with high precision and produces standardized, structured labels and charts for intelligent video analysis.

街景目标检测及行人和车辆趋势。行人 110–120 人,车辆 8–15 辆,桥梁 1 座,楼房 6 栋,地铁站入口 1 处,自动扶梯 1 部,垃圾桶 3 个。
图片加载中…Loading image…

系统能力System capabilities

多路视频画面中,红色检测框标出车辆、行人与建筑物
图片加载中…Loading image…

核心技术Core technology

画面、文字与语音经过模型处理后生成结构化标签图片加载中…Loading image…

视觉模型Vision model

由蔓藤科技自行研发的视觉模型作为核心驱动力,独立驱动起了整个系统的运行,该模型凭借其稳定的性能和灵活的应用方式,确保精准高效的对视频数据进行实时处理和分析。Manteng Technology’s in-house vision model powers the entire system. Its stable performance and flexibility support accurate, efficient video processing and analysis in real time.

三路摄像头的分析结果汇入同一分析面板图片加载中…Loading image…

联合分析Cross-camera analysis

使用了由蔓藤科技自行研发的联合分析技术,能汇聚多路视频的独立分析结果,开展跨点位联合分析,同时不干扰这些视频其它场景的分析任务。Manteng Technology’s in-house cross-camera analysis combines independent results from multiple video feeds for analysis across locations. Other analysis tasks on those feeds continue in parallel.

视频中选定的分析区域以亮框突出显示图片加载中…Loading image…

画面专注Region-focused analysis

运用了由蔓藤科技自创的画面专注技术,能根据业务场景的实际需求,重点分析视频画面中指定的一个或多个专注区域,极大的提高了准确率、增强了灵活度。Manteng Technology’s region-focused analysis lets users select one or more areas of a video frame based on their operational needs, improving accuracy and giving them greater flexibility.

高效部署Efficient deployment

灵活部署Flexible deployment

系统的部署具备高度灵活性与适应性,能够精准匹配各类业务场景与需求,无论是面向政务的专属云环境、企业内部的私有云架构、广泛共享的公有云平台,还是传统的本地化部署,均能提供最为优化的部署策略。The system adapts to a wide range of business and operational requirements. Deployment options include dedicated government clouds, enterprise private clouds, public clouds, and on-premises environments.

并行推理Parallel inference

系统原生支持多 CPU、多计算卡并行推理,通过统一资源编排,将推理任务分配给多个硬件设备并行处理,充分利用硬件算力,提升并发处理能力。经测试,NVIDIA T4 GPU 每秒可并行处理 80 路视频,Ascend 310P NPU 每秒可处理 50 路视频。The system natively supports parallel inference across multiple CPUs and accelerator cards. It coordinates these resources and distributes inference tasks across devices to make full use of available compute and increase concurrent processing capacity. In testing, an NVIDIA T4 GPU processed 80 video streams per second in parallel, and an Ascend 310P NPU processed 50 video streams per second.

多平台适配Multi-platform compatibility

系统已完成鲲鹏平台全面适配,可在国产化软硬件环境中稳定运行,能够承载高并发下的视频内容智能分析业务,保障多路视频流同步解析、特征提取与模型推理,可靠满足各类视频业务场景下的智能化分析处理需求。Fully adapted to the Kunpeng platform, the system runs reliably on Chinese hardware and software. It supports high-concurrency video analytics, including simultaneous parsing, feature extraction, and model inference across multiple video streams.

场景案例Application scenarios

图片加载中…Loading image…

智慧城市安全监管Smart city safety monitoring

系统可接入城市现有道路、路口、地铁公交站等各类公共区域的多路监控摄像头,汇聚采集的实时视频流数据,开展自动化解析识别与数据分析,动态感知公共区域内人员、车辆、设施、环境等要素的分布状态、流量起伏、秩序变动、险情灾害、异常事件等变化趋势,并将分析结果、统计报表、告警信息实时推送至城市管理人员,支撑管理人员快速掌握全域态势、提前处置隐患、高效开展精细化城市治理。The system integrates live feeds from existing cameras on city roads, at intersections, at subway and bus stops, and in other public areas. It automatically analyzes the feeds to track the distribution of people, vehicles, and facilities, changes in pedestrian and vehicle volumes and environmental conditions, and signs of disruptions, hazards, disasters, or unusual events. Results, reports, and alerts reach city staff in real time, helping them assess conditions across the city and address risks early.

图片加载中…Loading image…

医疗影像辅助筛查Assisted medical image screening

系统支持导入CT 影像、病灶影像、病理切片影像等多类型医学影像作为输入数据源,自动对影像内的目标区域开展智能解析,自动定位标注可疑病变位置,为疾病初步预筛查工作提供数据支撑,医务专业人员可对照原始影像、系统标注后的筛查影像以及配套结构化报告逐项核查、二次研判,提升工作效率,为诊断工作提供可靠参考依据。The system accepts CT scans, images of lesions, digitized pathology slides, and other medical images. It automatically analyzes target regions and locates and marks suspected lesions to support preliminary screening. Medical professionals can compare the original images, annotated images, and structured reports for item-by-item verification and secondary review, improving workflow efficiency and providing a reference for diagnosis.

图片加载中…Loading image…

电动汽车充电桩安全管控EV charging station safety monitoring

系统可对接现有的电动车充电车位、出入口、通道和场所的多路监控摄像头,从实时视频流数据中识别目标对象,动态感知充电车位空闲状态、充电桩使用情况、电动车长时间滞留停留情况、消防、通行通道被占用等各类情况,同时自动汇总、统计相关监测数据,为运营管理人员对优化现场巡检计划、人员调度、秩序管控以及隐患处置提供数据支撑与决策依据。The system connects to existing cameras covering EV charging bays, entrances, exits, access lanes, and surrounding areas. It detects objects in live video to monitor bay availability, charger use, vehicles staying for extended periods, and blocked fire or access lanes. It automatically compiles monitoring statistics to help operators plan inspections, deploy staff, manage the site, and address hazards.

图片加载中…Loading image…

智慧社区安全巡查Smart community safety patrols

系统可对接现有的社区出入口、步行通道和公共活动区的视频监控摄像头,从实时视频流数据中感知、分析和捕捉社区公共区域内的人员流动、聚集、异常行为等各类动态信息,帮助社区管理人员随时随地掌握各公共区域的现场实况,及时发现风险隐患、秩序问题,快速响应居民诉求,助力开展人员疏导、治安防范、便民管理等工作,进一步提升社区治理效率与响应速度,为社区居民提供更加安全、优质、精细化的社区服务。The system connects to existing cameras at community entrances, along walkways, and in shared activity areas. It analyzes live video to detect pedestrian movement, gatherings, unusual behavior, and other changes in public spaces. Community staff can check conditions remotely, identify safety hazards and disruptions early, and respond quickly to residents’ requests. The information supports pedestrian flow management, security patrols, and day-to-day resident services, helping staff respond faster and provide safer, more attentive service.

关于我们About us

公司成立于 2016 年 3 月 9 日,拥有技术扎实、经验成熟的研发团队与深厚的行业积累,致力于为客户提供一体化系统集成解决方案、应用软件开发及全流程 IT 服务支撑。

我司自研的视频内容分析系统具备良好的开放性与可扩展性,支持基于客户实际业务需求深度定制、二次开发,快速适配差异化业务场景。

Founded on March 9, 2016, the company draws on an experienced R&D team and extensive industry knowledge to provide integrated systems solutions, application development, and end-to-end IT support.

Its in-house Video Content Analysis System is open and extensible, supporting extensive customization and further development to meet specific customer needs and adapt quickly to different business scenarios.

图片加载中…Loading image…
返回顶部Back to top