I. Background and Challenges
urban public spaces tend to have tens of thousands of road surveillance cameras all over the street, around the clock monitoring and recording at the same time improving the social security also had vast amounts of video data, in the face The huge number of videos, even if a large number of police officers are deployed, adopts "human-sea tactics," but it is subject to the naked eye to identify the limits of labor intensity, and it is still unable to ensure the accuracy and timeliness of manual video search. In particular, emergency cases often occur when emergencies occur. The best time to solve the case was delayed, resulting in slower and slower responses to related intelligence investigations and case detection. Sometimes, in order to find a clue, first-line civilian police often watch the dozens of hours of videos of multiple cameras overtime and work overtime. They are mentally intense and physically overworked. They are huge overdrafts for the investigators’ mental and physical health.
As an important sub-project of Dahua* View Cloud Solution -* View Cloud Summary Solution is an overall solution based on cloud architecture and big data framework launched in the new situation. The system can analyze massive video at a high speed and quickly locate key issues. People, cars and other video information, and then profoundly change the previous policing work model - from the "view" to "search target" a revolutionary change.
Massive video structured processing logic
The Dahua View Cloud Digest Plan Dahua is to "see video" to "Search Video" - a landmark product of Video 2.0. Its basic principle is to convert various elements in the video into a parallel architecture through cloud computing. Structured description data, then according to the search conditions, these elements are screened and spliced, and the user's required condensed video is generated in real time.
The system can interface with cloud storage and combine with distributed computing technologies to fundamentally solve the problem of low efficiency of traditional video file analysis and processing. The industry-leading enrichment digest algorithm library is used to subvert the traditional way of viewing enriched summary results. In order to search for videos, quickly find key clues from massive video data and let Haidi fishing “reconstruction” is no longer a legend.
Chart system architecture
Third, the system features 1, video summary - second-level processing of massive video In the actual use of massive video second-level processing capabilities, the system can support simultaneous video channels from the cloud storage and video-based sharing platform, support for offline file upload, can easily Video files that need to be processed manually can also be easily searched for historical processing tasks.
The system supports summary task management for massive video files, and can be flexibly set for summary tasks - with an enriched summary task operation area, and can perform condensed summary operations according to policy scheduling (such as setting a queuing mechanism or a concurrent processing mechanism).
2. Video Retrieval - Global Positioning of Human and Vehicle Targets After completing the summary processing of massive video, a structured snapshot of the video can be obtained, and then the people and vehicles in the video can be structured and the snapshot can be retrieved twice.
The system supports searching based on structured data such as target type (person or car), body color, and person color (upper body or lower body color). It supports the identification of multiple colors of white, black, red, yellow, green, cyan, blue, and violet. After searching, the target with high matching will be displayed in advance.
The video summary can search for the target object according to dimensions such as the target type, color, size, speed, time, and direction. By changing the direction of the movement and the target type, the system displays the search results in real time, and the summary picture can be downloaded and saved.
Figure set the direction of motion filter target
Global Search Analysis - The cloud summary system can perform a unified target search on all video summary results.
Figure blue green vehicle global search results
3. Video Concentration Preview Video Condensation can filter the motion target in the video according to the target type, color, size, speed, concentration density, and motion direction, and can choose whether to display the time stamp.
Figure enrichment video any condition real-time playback
Change filter conditions, concentrate video presentation in real time, without the need to do video analysis again to truly follow, use and follow.
Fourth, the user value 1, massive video processing, farewell to see "video" era of cloud computing parallel computing framework, performance is several orders of magnitude higher, fault tolerance between nodes, and can be linearly extended. The smallest scale system can realize end-to-end processing of 1080P HD video for one minute. It can rapidly process the massive video involved, quickly locate the target image of human and vehicle, and form a condensed video to play, let the first-line police cherish time, cherish life, and stay away from the original video. .
2. Flexible extension of the specifications The Hadoop distributed framework can process massive amounts of view data at a very fast speed. The system scale can be flexibly expanded, and the video services can be developed on demand. It is never necessary to push back and forth again.
3. Future-oriented cloud storage design The cloud abstraction solution adopts view object cloud storage, which completely eliminates the performance bottleneck of uploading and downloading, realizes true end-to-end massive throughput read and write, and builds future-oriented business processing capabilities and system smooth expansion capabilities. ;
4. Massive video structuring and global unified retrieval of cloud abstracts The use of self-developed advanced machine vision algorithms - to achieve a structured description and enrichment of massive views, to form target image fragments + key feature descriptions, and to provide a global search portal, through many The conditions (person-vehicle classification, size, speed, color, characteristics, etc.) are used to retrieve and locate the key person's car information in seconds.
5, map detection business integration, business deep integration support and third-party video surveillance system integration and integration, users can quickly build a wealth of tactical application and intelligent platform, and the structural analysis results quickly into the policing workflow, and The business is tightly integrated to achieve centralized storage and application of valuable view information;
6, flexible deployment, capacity sharing, Internet +
Break the chimney construction model and provide unified IP and universal RESTful interfaces. The system can be deployed as an infrastructure independently, or it can provide a more flexible view summary big data service to the connected three-party platform to implement the ubiquitous view big data summary service.
Hotel Room Vending Machine
A hotel Vending Machine is a 24-hour unattended machine that does not require hiring a human to look after it, thus avoiding traditional hiring and training غير مجاز مي باشدts. In addition, the hotel intelligent vending machine is placed in the hotel room, and there is no need to go out for rent, and the cooperative operation with the hotel can realize the income.
Hotel Room Vending Machine,New Vending Machine,Hotel Vending Machine For Condom
Dongguan Kaisijin Intelligent Technology Co., Ltd , https://www.oemsmartlock.com
برچسب:
،
ادامه مطلب
بازدید:
