Trend and pain points of security video on the cloud and exploration and practice of Huawei cloud video access service
LiveVideoStack 2021-08-09 16:32:21

With the increasing proportion of video resources and content in the Internet , Accordingly, the contradiction between the utilization level of video resources and the increasing business and product requirements will become more and more prominent . This article comes from a speech shared online by Qi Yankun, product manager of Huawei cloud video access service , Combined with the practical experience of video access service, we discussed the trend of video cloud 、 Pain points and exploration and practice of Huawei cloud video access service .

writing / Qi Yankun

Arrangement / LiveVideoStack

Hello everyone , I'm Qi Yankun , From Huawei cloud media service product department , Currently, he is the product manager of Huawei cloud video access service . Today's content is divided into the following parts .


The first part is the development trend of security video , Including technology trends 、 National policies, etc ; The second part is the pain points we found in the process of practice ; The third part is to share the application practice on the cloud ; The fourth part focuses on the scene and some cases in the process of practice ; In the fifth part, there is a small video to demonstrate the operation process of cloud ; The sixth part is communication and interaction .

1. Development trend of security video


Cloud and AI Opened a new era in the security industry .

Security is closely related to everyone's life , Living in a community , There are camera related systems in the office building , These are the most basic security . From an industry perspective , Security has gone through about half a century , The industry has developed for a long time , From the beginning of the analog signal , Then it was solved “ visible ” namely “ Yes / nothing ” The problem of , It has realized the “ Thousands of miles of eyes , Comfortable ears ”. After this , In terms of the overall technical trend and business development, it still can not meet the demands of customers , for instance , There are several subsidiaries of a company distributed in different regions , The boss wants to view subsidiaries in different regions remotely , This is the Internet age . The network age is combined with the current mobile Internet , Greatly enhance the security experience , But it also brings a big problem —— bandwidth , To transmit video to the cloud, we need to solve the problem of bandwidth , This is the key technology .

Then people entered the era of intelligence , Artificial intelligence is also closely related to our life , For example, the parking lot doesn't need to be charged manually , Just recognize the license plate .

Why cloud and AI Opened a new era ? First AI Reduced demand for manpower , Increased efficiency , If the washing machine enters thousands of households , It liberates people's hands , that AI From the field of security, we can liberate people's eyes . The other is on the cloud level , No, of course you can , However, the construction of resources in the later stage, including the increasing amount of video , From SD to HD , from 2D To 3D And then to interaction , This piece is for the network 、 The demand for computing power is increasing , therefore “ On the cloud ” It's a good trend .


Just mentioned that in the field of security, we have entered the intelligent era , So what role will video play in the intelligent age ?

The intelligent world is divided into three layers : perception 、 interconnection 、 All things are intelligent . From the perception layer , Mainly data acquisition . We obtained from the data analyzed by some analysis institutions , In the digital transformation of the industry ,65% The data is video data , Enough to see that video is a very important part of the digital transformation of the whole industry . Another data is , In our intelligent world ,49% Our application is based on video , It can be seen that video is the core of all things' perception , It is the clairvoyant of the intelligent age .


The cloud has brought several benefits : share 、 to open up 、 intelligence 、 Unified management . If you don't go to the cloud , The resources mentioned earlier are a problem , Another problem is that a lot of hardware stacked together will cause data and Applications , Software and hardware cannot be decoupled . For intelligent evolution , With more and more videos , Yes GPU More and more demands , It's hard to deal with locally . If you put them all locally , A company with many subsidiaries , Its equipment is distributed in different machine rooms , It is difficult to achieve unified management of equipment . So from now on , Sharing can be well realized on the cloud 、 to open up 、 Intelligence and management .


From the perspective of national policy , Whether it's the Ministry of public security 、 Ministry of Communications 、 Postal industry 、 Dangerous goods production enterprises , Countries have clear communications , Hope that enterprises can achieve video networking and video sharing . After reading the policy , We summarize the following four levels : The first is network wide sharing , Only after sharing , The data of everyone can be unified , do AI analysis , Linkage with other equipment 、 share ; The second is global coverage , The main emphasis is the reasonable deployment of video in some areas ; The third is full-time availability , Recently, the news that some video websites have problems has frequently made headlines and hot searches , That's why it's not available all the time , If they can go to the cloud , There will be a good performance in usability ; Fourth, the whole process is controllable , The first is privacy , The most important thing is that data cannot be stolen , In addition, if you want to see the data, you need to obtain your authorization .

So from the interpretation of national policies , Share the whole network 、 Global coverage 、 Full time availability 、 Controlling the whole process is important information .

2. Analysis of cloud pain points on security video


We also found some problems in the projects we participated in recently , In particular, some enterprises buy cameras from many manufacturers , Will ask us how to connect to the cloud , Because there are many protocols for each camera .

Many well-known camera manufacturers provide their products in order to better serve their products , Corresponding video access platform will be provided , This can lead to a problem , If from the customer's point of view , Multiple camera manufacturers have multiple systems , From the perspective of enterprises, this is relatively fragmented , It will hinder the sharing of data . The second point is the agreement , The current video protocols are RTMP、RTSP、Onvif、RTP、GB/T28181 etc. , Will cause trouble to enterprise customers . The third point is that the customer will report that the video is stuck 、 Blurred screen 、 The delay is large 、 There's no flow all the time , These are the problems that will be encountered in the actual project . The fourth point is the cost , It can be seen on two levels , First, if deployed locally , At this time, the demand for computing power network is relatively large ; The second is the cloud , Build on the cloud BGP The high cost .


For the above challenges , I'd like to share some experience with you .

First look at the agreement , The figure on the left can be seen from the network layer (IP) The transport layer (TCP、UDP), There will be many corresponding protocols for audio and video transmission , It can be seen that there are at least three levels from the perspective of audio and video transmission , The first point is signaling control , Include H.323, This is a more common protocol used in the traditional telecommunications industry , At the Internet level, it is used more SIP agreement , For better audio and video transmission effect , There will be RSVP、RTCP Such agreement , The other is transport layer protocol RTP, This is a relatively low-level protocol .

We can see that there are many upper layer agreements , Yes HTTP、RTMP、HLS、GB/T 28181 etc. , What do these agreements do ? First GB/T 28181 It's an agreement made by the Ministry of public security of China ( The following will focus on ),Onvif and PSIA It is also based on the security system , More agreements are used in overseas markets .

Let's talk about some typical protocols ,RTSP The delay is good , But it's complicated , And right Web Support is not very good , From these characteristics, we can see that it is a more suitable protocol for security .RTMP In the process of data stream transmission, the data is divided into many small blocks for transmission , There will be many delays , Probably <3s about , But it's important for Web The support is better , Through the plug-in, you can Web Check out , More suitable for live scenes .HLS From a cross platform perspective , It runs better , But the delay is relatively large .

To sum up, we can see that different protocols have different positioning , So in the security system or in business development , How to select an appropriate protocol to meet the demands of business development , Then you have the following GB/T 28181 agreement .


The agreement itself was made by the Chinese Ministry of public security , There are currently two versions GB/T 28181-2016 and GB/T 28181-2011, The main purpose of this protocol is to solve the problem of docking between camera systems of different manufacturers , Several important scenarios are standardized . From this point of view , In order to solve the problem of docking between different manufacturers ,GB/T 28181 The agreement is a very suitable agreement .GB/T 28181 In fact, for the just mentioned SIP、SDP、RTSP Defined , It can be considered as a protocol specification . The agreement mainly defines 12 A scenario , The device can register and log off with the server 、 Capable of real-time audio and video on demand or playback of historical video 、 Equipment control ( Camera rotation direction )、 Have the ability to locate and notify changes in equipment status .


With GB/T 28181 agreement , From the perspective of Huawei cloud , We gave priority to the implementation of this Agreement , All mainstream cameras in the current project can be based on GB/T 28181 Access , It can solve the problem of multiple manufacturers 、 Multi protocol problem . If different business scenarios have special demands , We can also support RTMP、HTTP-FLV Standard agreement .

GB/T 28181 From the perspective of cloud on the device, it is also very simple , Just do some simple configuration , You can connect to the cloud without code development . After the video is connected to the cloud through the protocol , We also provide some capabilities on the cloud , Including equipment status management ( Online of the device 、 offline )、 Video stream management ( Video stream activation 、 Deactivation )、 Automatic invitation of abnormal video stream 、 Provide cloud storage in combination with other cloud services 、 watermark 、 Screenshot 、 Transcoding and other capabilities . most important of all AI The blessing ( I'll introduce you later ).


From a video perspective , Experience is a very important thing . If the video is played, the Caton screen is difficult to accept , From experience , First, eliminate the end side problem , Ensure smooth local video playback , After the local problem is eliminated , Focus on network reasons , Now it seems , Access from video , Different networks can be used when accessing , It is more stable to use dedicated network to access video , The effect of home broadband is poor . From a cloud perspective , Bandwidth reservation is very sufficient .

Then there is the cut-off , There are many reasons , It may be because the system itself is abnormal . Huawei cloud's solution is to use automatic invitation capability , The system automatically identifies which streams are broken , Then automatically pull up the flow , The customer experience will be greatly improved . Some disconnection means that customers need to take the initiative to disconnect based on business characteristics , The solution at this time is that the cloud system automatically recognizes , Inform customers through active subscription , The specific follow-up actions are handled by the upper application .

Finally, the end side configuration is complex , Actually GB/T 28181 The configuration is relatively simple , The camera itself is also a small system , Need to log in 、 Make some configuration , It may be complicated for construction personnel . The solution is to scan the code and register , Manage device information and complete cloud registration based on QR code scanning on the end side , This will intuitively reduce the configuration complexity . If the construction personnel are interested in GB/T 28181 More understanding , Then you can configure it directly on the device side .


Cost is of great concern to enterprises , The cost of cloud on video is roughly divided into two parts .

The first part is the enterprise bandwidth cost of enterprise customers , Need to discuss with the operator , The main method to reduce the cost on the access side is to reduce the transmission rate . The second part is the cost of cloud bandwidth , To talk about the cost of cloud bandwidth , Here we want to mention the native media network of Huawei cloud , We run a lot of business in the cloud native media network , Live broadcast 、 on demand 、 Real time audio and video , There are both big clouds and edge treatment , It is a media network composed of big cloud and edge . Based on such a media network , We can reuse network bandwidth resources well .

From the perspective of cloud video access, it is divided into several parts , Unified scheduling based on cloud , Signaling messages take priority to big cloud processing , This is the signaling scheduling process , If the access is legal , Camera or NVR Will get access points from the cloud . This access point may be the address of different edges , Scheduling based on proximity strategy or reliable signaling strategy . In this way, the original will be directly dispatched to the big cloud BGP The access camera has priority to go to the edge , The edge passes through the special line , Send the video back to Dayun and do the corresponding EI Analyze or store , This is the overall architecture . From a reliability point of view , If the edge access point fails , The video will be scheduled to other nodes that are providing reliable services . If there is an extreme situation , All the edge nodes failed , Then we'll schedule the video directly to Dayun . So from the perspective of the whole architecture , That is, the access reliability can be guaranteed , It can also ensure the lowest cost .

3. Huawei's application practice of video access to the cloud


Just now, we have analyzed three problems encountered in the process of project delivery . So what else can we do after the video is connected to the cloud ? On Huawei cloud , In particular, the services combined with video include live broadcast acceleration , If you want to let a lot of people watch the video at the same time , It can be combined with live broadcast acceleration service , Huawei's live broadcast effect has achieved a success rate >99.7%, Caton rate <2.5%, Playback delay <3s. In the cloud, Huawei's cloud storage capacity and service availability promised to the outside world have reached 99.995%, It has a good competitive advantage . There is also a junction , If you want to do some on-demand transcoding after the video is in the cloud , Or high definition low code conversion , We also have supporting video services . The other is the combination of intelligent analysis services , Next, I will focus on .


Here I list some intelligent algorithms related to daily life , It's divided into seven categories : Face recognition , The gate of the office building will use this service ; Intrusion detection , The community uses more , If an intrusion is found, it can be identified and alarmed ; Personnel statistics , Whether the personnel are gathered , Urban management scenarios can be applied to ; Safety inspection , Apply to fire alarm 、 Smoke alarm ; Vehicle detection , Detect illegal stop 、 Traffic flow statistics, etc ; Body recognition , Human body control 、 Abnormal behavior is usually used in urban management .

You can use Huawei cloud services in combination with different business scenarios .


For the above , Let me make a summary , Based on the three pain points we encountered in the early stage : Multi manufacturer and multi protocol 、 Experience 、 cost .

The first is cloud capability blessing , Cloud 、AI Combined with storage and codec . The second is full real-time , Multi edge media network can ensure business continuity , If there is any abnormality , It can switch automatically , Provide cut-off automatic invitation . The third is wide coverage , Compatible with multiple manufacturers , This is in some projects , Including Haikang Dahua 、 Huawei and other mainstream cameras and NVR The equipment has been docked . Fourth, low cost , High cost performance , Combined with media service network, we can achieve low cost . The fifth is the whole scene , From the support scenario , We are full scene access , Including security 、 traffic 、 Park 、 Business, super, etc. can support .

Now let's discuss the security scenario .

4. Scene and case introduction


From the perspective of video access services, the main scenarios include smart stores 、 Home Security 、 Wisdom park 、 Intelligent environmental protection , Transportation networking .

Smart stores are generally remote shop patrols , Passenger flow analysis ; Home Security , At present, it seems that it will become a trend , Install cameras at home to observe the situation at home ; Wisdom park , It's widely used , Park Visitor Management , License plate recognition, etc ; Intelligent environmental protection , It is also an increasingly important scene , The follow-up meeting will focus on the cases of solid waste monitoring ; Traffic networking is also an important scenario .

Next, let's introduce some cases that have been delivered .


Bright kitchen and stove are closely related to everyone's life . Now parents are very concerned about students' meals at school , This is a big appeal , Secondly, takeout is becoming more and more popular , You want to see the food processing process , If all these are arranged, manpower supervision will form a large manpower cost , Can't reach such a large scale , So the video is saved to the cloud through AI Analysis and big data analysis , It can automatically identify abnormal violation scenarios and present them to relevant supervision and management departments .

In the bright kitchen and stove scene, the main intelligent recognition scenes are 9 individual : Not wearing a mask 、 Without a chef's hat 、 Play with the smarthphone 、 Smoking monitoring 、 Naked upper body 、 Pests 、 The camera is blocked 、 The trash can is not covered 、 Without gloves . Video and AI And big data can greatly improve the regulatory efficiency .


Smart venues are related to children , Now there are more and more training venues , Parents will record videos during the escort , There is a large amount of recorded video in this process , Look at the trouble , By collecting video from venue Cameras , Then enter the child portrait with permission , Search for relevant highlights based on portraits , Make relevant videos through editing and push them to parents , Parents received more wonderful pictures , Easy to find, easy to store .


Once the solid materials produced by chemical enterprises are stored 、 transport , Improper handling can lead to major problems , The most intuitive Yancheng explosion was caused by improper disposal , The local video was also lost . From this point of view , Video cloud can eliminate the problem of video loss . This business scenario not only has video applications , There are others. IoT Device joint sensing , Complete the whole life cycle management of solid waste , In this way, enterprises can be well supervised , Help enterprises standardize production .

The above are several typical cases .

5. Operation demonstration of video access to the cloud


Next, let's learn how to quickly go to the cloud through a small video .

( Video fast cloud operation demonstration )

You can see that the cloud on the video is very simple , No need to write any code , It only takes a few steps to configure .

6. summary

I would like to make a brief summary of what I shared today .

First, security has entered a new era , Entered the era of intelligence .

The second is the benefits of cloud on video : share 、 Development 、 intelligence 、 Unified management .

The third is the ability of Huawei cloud video access service itself, including supporting national standards 、RTMP、HTTP-FLV Other protocols , Go to the cloud easily .

Fourth, it can be combined with cloud services after going to the cloud , Especially algorithm 、 Storage 、 Live service 、 On demand service, etc .

The above is today's sharing , thank you !


1、GB/T 28181 The plan , Is it your main access scheme ?

GB/T 28181 It was formulated by the Ministry of public security of China , This protocol is widely used in domestic video management . More access protocols used overseas are Onvif、PSIA Such agreement . therefore GB/T 28181 Focus on domestic market , It is to solve different manufacturers , Management costs caused by different agreements .

2、 For some stock Cameras , I don't think the compatibility is very good ?

This is a real problem , We now find that the camera itself is for GB/T 28181 Support for , Some are still supporting very old 2011 Version of .GB/T 28181 There are two major versions , The first is 2011 edition , This is usually supported by very old cameras , The new camera is supported 2016 Version of ,2016 The version of the protocol will also be more perfect .

3、 In the future, we will add a pair of SRT,QUIC Wait for the support of the agreement ?

The video access service of Huawei cloud will continue to develop , We will combine business characteristics and customer needs , Any technological development is combined with these two points , If there are technical requirements in the future , Huawei cloud will consider this issue .

4、 Just now, the access protocol supports GB/T 28181, For example, some small equipment manufacturers or previous old models do not support how to deal with ?

If you really don't support GB/T 28181, Then you can find one NVR, Because now let's look at NVR Can support , Access first through small devices NVR, Re access GB/T 28181, because NVR Support GB/T 28181, It's equivalent to making a conversion . If there is no support GB/T 28181 Of NVR, You can do a conversion at the local gateway , The cost will be a little higher .

5、 Today's projects are all public cloud or private cloud delivery ?

Today's cases are all based on public cloud delivery . The cloud I mentioned here today includes public cloud and private cloud , Going to the cloud does not necessarily require going to the public cloud , Private cloud can also .

6、H.265 Can the stream be accessed ?

Sure .

7、 Monitor whether the stream can be encrypted , Encrypt after streaming media is received , Player decryption ?

GB/T 28181 The agreement needs to be improved . National secret algorithm can be used for encryption and decryption .

8、 If the camera only supports UDP Push flow , What do I do ?

GB/T 28181 The protocol itself supports two protocols , One is UDP, One is TCP.

9、 How about the real-time video after going to the cloud ?

From the current project , If the local bandwidth is good enough , Click on the cloud to open the video 1-2s It's no problem to play in , The real-time performance is good . From the perspective of experience, it can meet the demands of most business scenarios .

10、 For Dahua 、 Remote configuration of parameters on Haikang camera , It directly calls the manufacturer's SDK Do you ?

GB/T 28181 The agreement itself does not need to use SDK, Directly log in to the camera for configuration .

11、GB/T 28181 Is there a security problem on the protocol ? It is said that uplink is easy to be hijacked ?

If you want better security , Security reinforcement can be carried out at the link layer , Ensure data security from the physical channel .

12、 Video access service supports mobile phone real-time streaming ? What are the performance requirements for the equipment terminal ?

Huawei's video access service supports three protocol national standards 、RTMP、HTTP-FLV, Can be based on RTMP Push flow , It can be realized . General mobile phones can meet the demands , Now most live broadcasts are streaming through the mobile terminal .

13、 The public cloud is similar to mouse detection. Can you elaborate on these ? The detection cost of each frame is relatively high , Frame extraction will miss detection .

That's a good question , My team focuses more on video access services , The mouse test is carried out by another team , I can take this question back and EI Team discussion .

14、H.265 After the video stream is delivered to the cloud ,H5 Direct use H.265 Play or transcode into H.264?

In theory, both can .

**15、** Just shown 40+ Are all these algorithms made by Huawei itself ? It was developed in cooperation with ecology ?

It is mainly made by Huawei itself , Huawei cloud itself EI There are many combinations of algorithms and partners , Partner algorithms will also be introduced .

16、 Can you realize voice intercom ?

GB/T 28181 It supports voice .

17、 I think some companies now use RTC The plan , This is more real-time , Better network compatibility , Does the teacher recommend this way ? The service requires that the video is turned on in seconds .

Some scenes, such as smart doorbell recommendation RTC programme , because RTC The delay of the scheme is relatively low . If the delay requirement is not high , There is no need to use RTC programme , Finally, the scheme is selected based on the business scenario .

18、 How many streams can be used for existing project concurrency ?

Huawei cloud single tenant can go to 10 10000 level concurrency .



Please scan the picture for details __ QR code __ Or click on __ Read the original __ Learn more about the conference .

This article USES the Article synchronization assistant Sync