Network camera intelligentization is currently the focus of research at home and abroad

To make the network video surveillance system intelligent, its most important product form is to make the network camera intelligent. In this way, after monitoring and recognizing the video image, the front-end smart network camera can only transmit abnormal images to the back end for recording and display, thereby greatly reducing the burden on the network. But how to make the high-definition network camera (1280 × 1024) of the spare CPU resources embedded in the intelligent video analysis and identification of the software algorithm to the network camera, is currently the focus of research at home and abroad security scholars. This article describes a new single-channel high-definition IP network camera function module program, and based on this embedded intelligent features to form a smart network camera, the article also describes the composition of the program, design principles and points, and should have the appropriate intelligence Functions, etc.

It is well known that picking up the monitored electro-optical image depends on the camera, while the analog camera can only transmit unidirectional video signals, and its performance is subject to the DVR and the operator's level. However, network cameras can transmit full-duplex digital information and can be effectively integrated with other devices in the system, enabling them to achieve higher levels of performance in a distributed, scalable environment. The network camera can communicate with multiple application systems in parallel to achieve a variety of different functions, such as detecting motion in a picture or sending video streams in different formats. A camera with an IP address is more convenient for networking because it does not require additional analog-to-digital conversion and does not require the installation of a sending device at the camera end, so the size of the camera will be smaller and smaller, which is also more convenient for concealment. , is required by the security monitoring system.

Network cameras integrate the functions of traditional cameras and computers. They not only digitize and compress video signals, but also have a network interface. Its video is transmitted over an IP network and recorded on a standard PC with video management software. The network video system composed of a network camera is a true network video system, and it is also a completely digital system because no analog equipment is used. The video signal of the monitored scene is continuously transmitted from the network camera through the IP network. The camera can provide the viewer with stable image quality no matter where the viewer is located. Therefore, the network camera is most suitable for application in the network security monitoring system in Pingan City, and it should be built into an intelligent network security monitoring system, and the network cameras should be intelligent. In this way, after monitoring and recognizing the video image, the front-end smart network camera can only transmit abnormal images to the back end for recording and display, thereby greatly reducing the burden on the network. However, if all intelligent software algorithms are added to the front-end network camera, due to its own limited CPU resources, it is often necessary to increase the configuration of the front-end equipment or use a dedicated video analysis instrument to work together, which will greatly increase the volume and cost of the equipment. It is not what the user needs.

Therefore, how to make full use of the extravagant CPU resources of the network camera and embed the intelligent video analysis and recognition software algorithms into the network camera is one of the focuses of research by domestic and foreign security scholars. This article describes a new single-channel high-definition IP network camera function module program, and based on this embedded intelligent features to form a smart network camera, the article also describes the composition of the program, design principles and points, and should have the appropriate intelligence Functions, etc.

High-definition IP network camera function module solutions Compared with ordinary cameras, the network camera is mainly to increase the digital processing and digital video compression and network access functions, that is to convert the digital video signal to comply with the network transmission protocol (such as TCP / IP protocol) data stream for transmission to the Internet. Therefore, a network camera can be seen as a combination of a video camera and a computer. It captures images and transmits them directly over IP networks, enabling authorized users to view, store, and manage video data locally or remotely through a standard IP-based network infrastructure, or to monitor the camera's on-site collection of microphones. sound.

Because the network camera also has a communication (COM) interface, the network camera has control data capture and conversion capabilities. Because the COM interface can be connected to an external decoder. Therefore, the PC of any authorized user on the network can be transmitted via the COM port of the camera through the network and send instructions to the external decoder to make the camera pan and lens operate as required. At the same time, the COM interface can also receive data from the front end and transfer it to any PC on the network.

The network camera also has an alarm (ALARMI/O) interface. This alarm signal access terminal can receive the switch signal sent by various alarm sensors, and encode it, and then send it to the TCP/IP network. Through the application software, various alarm response signals can be set in response (such as Sounds, light alarms and alarm recordings, etc.).

Because each network camera has its own independent IP address, it can directly connect to the network and has built-in Web server, FTP server, FTP client, Email client, alarm management, programmability and many other features. Network cameras do not need to be connected to a PC, they can also operate independently and can be placed in any location with an IP network interface. In addition to the video information, the network camera can also realize more other functions through the same network connection and transmit other useful information such as video motion detection, audio, digitized input and output (can be used to implement alarm linkage, such as triggering Alarms or activation of live lighting, etc.), serial ports used to transmit serial data or PTZ device drivers, etc. The image cache in the webcam can also save and send video images before and after the alarm occurs.

Therefore, the IP network camera can pre-process and compress the acquired video source, and then perform remote transmission and storage over the IP network. Its important components include video preprocessing, video compression, and Video-Over IP capabilities. At present, the input data format of the IP camera is: 1 D1/25 frames/s+1 CIF/25 frames/s, which can usually be achieved by using one Cyclone II EP2C35 and one low-cost processor.

As the resolution of current IP cameras is gradually evolving from standard definition D1 to high definition (1280×1024), and local real-time compression must be performed, only hard compression methods can be used. If multiple DSP processors are used, the cost, integration and power consumption of the system will increase, making it unacceptable to the user; if a single low-cost FPGA device is used, the performance cannot meet the design requirements, but if a single chip is used, High-performance Stratix series FPGA devices can meet the requirements. Because this device has a corresponding structured ASIC HardCopy series device, it can further reduce the cost to 1/10 and reduce power consumption by 50%. Therefore, the functional modules of the single HD IP camera using this FPGA device.

Video input preprocessing (including filtering, scaling, etc.) can be implemented using Altera's VIP suite. And 1 way 1280×1024/25 frames/s+1 way CIF/25 frames/s and H. 264 video compression will be implemented in FPGA hardware. The Nios II processor and VoIP reference design enable video-to-IP network mapping and transmission over Ethernet.

In addition, some low-cost DSP processors can be connected to assist in the analysis of some video streams. Although the logic of these tasks is relatively complex, the performance requirements are not high. Using Altera's custom on-chip bus Avalon and SOPC Builder system integration tools, the required peripherals can be automatically integrated and interconnected, which greatly improves design efficiency.

Due to the need for safe city construction, the development of video surveillance has been further promoted, which has also promoted the application of FPGA in this field.

In particular, the requirements for multi-channel, high-definition, networking, and high-speed and intelligent communication interfaces have all contributed to the further development of FPGAs in this area. Conversely, advances and updates in FPGA chip technology, IP cores, and reference designs have also contributed to the development of video surveillance.

Realizing the composition and principle of the intelligent network camera solution Through the aforementioned functional module scheme of a single high-definition IP network camera using a single high-performance Stratix series FPGA device, and the embedded intelligent software algorithm module, an intelligent network camera can be formed.

Obviously, an intelligent network camera consists of six parts: camera lens, CCD or CMOS imaging device, A/D and preprocessing section, intelligent software algorithm module, code compression module part, and network interface module. Therefore, the smart network camera solution, its composition and working principle of the block diagram shown in Figure 2.

It can be seen that the camera lens collects the optical images of the monitored scenes and focuses them onto the photosensitive surface of the CCD or CMOS imaging device, and the converted video electrical signals are preprocessed by A/D conversion, noise removal, image enhancement, and the like. Then, it is sent to the module embedded in the intelligent software algorithm. The intelligent software algorithm is the most important part of the intelligent network camera. The accuracy of the algorithm and the accuracy of abnormal signal analysis and identification are the key to the practicality of the camera. When the smart network camera starts to enter the monitoring state, the image of the monitored scene will be transmitted through the network to the monitoring center via the compression and compression. Then only the signal with the abnormality will be encoded and compressed, and sent to the network through the network interface for transmission. This can significantly save network bandwidth resources.

Due to transmission after video compression, some of the real information will be lost and some noise signals will be generated. At this time, the compressed video will be analyzed and recognized by intelligent software, and there will be more leak alarms and false alarms. However, the intelligent network camera of the present scheme performs target extraction on an uncompressed original image. Obviously, there is no such problem. Moreover, it only needs to send relevant video to the back end for monitoring and recording in case of abnormal need of pre/alarm. In general, only a few target data information needs to be transmitted through the network. These target data information The traffic is less than 1/50 of the video traffic, which reduces the network bandwidth requirements and consumption, reduces the user's investment in the network, and at the same time brings the use of intelligent video networks to users who do not have very good network conditions. The possibility of monitoring the system.

Therefore, the key to implementing intelligent network camera solutions is to select and design intelligent software algorithms and then embed them into DSPs and FPGAs to form intelligent software module parts.

Design Principles and Key Points of Intelligent Network Camera Solution The main principles and main points for implementing the intelligent network camera solution are as follows.

Modular design The modular design mentioned here is mainly to modularize the intelligent function. Personally, the eight intelligent functions that can be embedded in the smart network camera described in the previous section require that each camera, except the first three basic functions, needs to form a module, and the rest can be formed separately. This is more flexible and convenient. Different modules can be selected according to the needs of the actual monitoring site. This can reduce capital investment, and the contractor is willing to accept it. Therefore, the modular design can not only bring convenience to the manufacturer, but also suits the actual needs.

Reliability Design Any product needs to be designed for reliability, that is, the structure designed by the system is reasonable, the product is durable, and the system is stable and reliable. In this way, EMC design must be carried out to be able to resist EMI, lightning protection and all kinds of interference so as to be able to adapt to a variety of adverse climate conditions without failure.

User-friendly operation design The so-called humanized operation design is very suitable for people's simple and convenient operation and use. If you use push-button operation as much as possible, you must have clear indicators and horn prompts. When it is installed in a hidden and high way, you can also use the remote control to concentrate the daily operations on the remote control.

The practical design of the so-called practical design, that is, the system is very practical, easy to use, easy to use, less links, easy to learn. It can automatically and conveniently perform function setting and linkage with systems such as anti-theft, fire prevention, and access control. The degree of automation is high, so that the system can be pre-alarmed and alarmed in case of actual infringement. In addition, a self-check function is also required, and self-check results are regularly reported to the management center for monitoring by the management center so that the fault can be immediately maintained.

When economic design considers system economics, in addition to considering the reduction of R&D costs and production costs of the system, the use cost of the system must also be taken into account, that is, maintenance costs, spare parts fees, etc. during the use period are included.

In order to obtain a higher price/performance ratio, designers should not blindly pursue complex and advanced solutions. Under the premise of satisfying the performance index, a simple scheme is adopted as much as possible because the scheme is simple, which means that the structure is simple, the reliability is high, and the cost is low.

After comprehensive consideration of the above factors, choose to determine the appropriate program.

Intelligent Network Cameras Should Have Intelligent Functions The intelligent network cameras formed by taking advantage of network cameras are embedded network cameras with intelligent functions. In general, some basic intelligent video functions can be realized by making full use of the excess CPU resources of the front-end network video equipment. In addition, in addition to the commonly used intelligent functions, it is also possible to embed some different required intelligent functions according to different applications.

The intelligent features that can be embedded in a webcam include the following.

Intelligent functions commonly required by surveillance cameras 1. Identification and pre-/alarm functions for video occlusion and video loss detection. In a large-scale security monitoring system, the maximum number of personnel on duty in the monitoring center can take into account or view is tens. Road video image. If one of the video images in the system is blocked or lost, most of the camera on the road is obstructed or damaged by the camera or it is faulty (such as cutting off the power supply line or power supply damage, etc.), and it is difficult for the on-duty personnel to find the first time. This may bring about major security risks. However, when the system has the function of intelligent recognition and pre/alarm for video occlusion or video loss detection, the on-duty personnel can perform the viewing and processing according to sound and light alarms for the first time, so as to eliminate the occlusion or loss of the video image. Security risks.

2. There is video frequency change detection and pre-alarm function. In a general network security monitoring system, when one of the video images is changed (that is, it is not the video image range of the originally set fortification point), most of this is Criminals want to carry out crimes against the original point of defense and move the camera or video camera to get a lot of vibration and move. The on-duty staff is also difficult to find at the first time. However, when the system has an intelligent function of video recognition, detection and pre-alarming, the on-duty personnel can view and process the sound and light alarms at the first time, thereby eliminating potential security risks caused by video conversion.

3. Detection and pre-alarm function of video blur detection In the intelligent security monitoring system, when the video image of one of the roads is detected to be blurred, it may be that the camera lens is moved (ie, the focal length is lost), resulting in the original design. When the video image is blurred, the video, point display, and pre/alarm can be started automatically.

In this way, on-duty personnel are able to view and process the first time to eliminate security risks.

Of course, in the high-speed ball or integrated machine, it can also be configured to solve the fast (t ≤ 1s) autofocus system.

Intelligent functions embedded according to different applications 1. Intelligent functions for detection, identification, and pre/alarm of various types of natural disasters can be built. In the construction of a safe city, there must also be water level, dilapidated buildings, and brittle fracture sites for urban fires and floods. The intelligent functions of detecting, recognizing, pre-alarming, and alarming in places where dangerous road sections or roads are prone to rock avalanches, avalanches, and earthquakes detected by seismic stations. Usually, a camera with a cross scale is placed for monitoring. When the video image at the video fortification point moves to a certain warning scale on the cross scale, the embedded intelligent software algorithm can automatically start recording, fixed-point display and pre/alarm, so that the on-duty personnel can alarm according to sound and light in the first time. Check and handle to prevent hazards caused by various disasters.

2. There can be video motion detection recognition and pre/alarm function In the intelligent security monitoring system, there must be at least video motion detection pre/alarm function, especially in banks, jewelry stores, shops, warehouses, accounting rooms, Armory rooms, security rooms, museums, oil depots and all important monitoring sites need to be set up where safety requirements are relatively high. Of course, this function must remove small animals such as rats, cats, and dogs, as well as interference caused by changes in light and other environments, especially in complex weather conditions such as rain, snow, fog, and high winds. Measure and identify the movement of a single object or multiple objects, including their direction of movement, motion characteristics, etc. In addition, the detection area of ​​the video movement can also be turned on or off individually to meet the requirements of the entrances, the halls, the parking lots, and the like during a specific time period.

When detecting and recognizing the intrusion of a moving human body that is not recognized, the video, the fixed point display, and the pre/alarm are automatically started, so that the on-duty personnel can perform the viewing and processing according to the acousto-optic alarm at the first time. In addition, PTZ control commands can be automatically sent according to their movements so that cameras with pan/tilt heads can automatically track objects. If the human body exceeds the surveillance range of the camera, it can also automatically notify other cameras in the area where the human body is located to continue tracking until the criminals are caught.

3. The smart function of counting the number of entrances and exits can embed the entrance and exit number counting function. This function can automatically analyze and count the number of people or objects crossing important national departments, important entrances and exits or designated areas through the analysis of the images monitored by the cameras.

The application of this intelligent function can also be widely used in the service, retail, and other industries to help managers analyze business conditions, or improve service quality, such as calculating the number of customers patronizing their stores on a particular day.

4. Intelligent functions that can be embedded in the crowd such as attention detection and congestion detection and pre-alarming. This intelligent function can identify the overall movement characteristics of the crowd, including speed, direction, etc., in order to avoid the formation of congestion, or to discover in time that there may be Racks and other abnormal conditions. Typical application scenarios include places where people gather at supermarkets, train stations, and entertainment venues.

In addition, you can count the time people stay in front of an object. Based on this, it can also be used to evaluate the attractiveness of new products or new promotional strategies. It can also be used to calculate the time taken to provide services to customers.

5. Intelligent function for detecting, detecting, pre-alarming, and other dangerous articles that can be embedded in non-detained explosives or flammable and dangerous articles. This intelligent function is mainly used for suspiciously retained objects (such as flammable, chemical, etc.) in important facilities or public places. Explosive and other dangerous objects) Exceeding the edge for a predetermined period of time has not taken away and other abnormal behavior, processing, identification, and locking tracking and pre-alarm, so as to eliminate as soon as possible to protect the safety of people's lives and property.

6, can have automatic light switch, adjust the light intensity and identify whether the intrusion of smart intelligence in party and government agencies, schools, businesses and large home environment, but also use intelligent video analysis technology to eliminate the huge energy waste and security risks, such as Automatically turn off unnecessary lights or turn on lights for work or adjust the light intensity as needed. Because the intelligent security video surveillance system can calculate at any time in a certain environment where people walk or work, automatically turn on or enhance the light of the corresponding location, automatically turn off or weaken the light in other locations. In addition, it can also be used to automatically analyze whether the people in these environments are consistent with what is already in the office or home database. If any outsiders are found to be invading, they can be promptly pre-alarmed.

7. Integrated safety precautions for other safety precaution devices capable of linking corresponding defense zones. An intelligent security monitoring system, in addition to the necessary video surveillance systems, should at least be able to link the security protection devices of corresponding defense zones, such as entrance and exit control systems, and fire and burglar alarm systems. Wait. That is, it needs to be linked with access control detectors, fire prevention and anti-theft alarms, and safety alarm detectors such as gas and radiation leaks to achieve integrated safety precautions. For example, when an access control detector, or an alarm signal such as fire prevention, anti-theft, and gas and radiation leakage, is received, the network camera at the corresponding location can automatically switch to the corresponding preset position via the presettable pan and tilt to automatically record the video. , And trigger alarm bells and lights for sound and light alarms, so as to achieve integrated security features.

In addition, it can also be embedded with human characteristics, biometrics, and intelligent functions required for electronic police and intelligent transportation.

The above-mentioned intelligent functions can be respectively arranged in the cameras in different positions at the front end of the security monitoring system according to the needs of different monitoring locations of the users. Of course, some of the more complicated ones can also be embedded in the DVS and the DVR.

It can be seen from the above that the intelligent network camera formed by taking advantage of the high-definition network camera is a network camera embedded with intelligent functions. Obviously, this intelligent front-end monitoring device can effectively save bandwidth resources. It is a better component of the network intelligent video surveillance system. In this way, on the back end of an intelligent video surveillance system with an intelligent network camera, it is only necessary to centrally manage the target data information sent from the front-end camera without processing and recognizing the video signal, so the back-end system does not need to be expensive. The device can also perform efficient intelligent video analysis. Because the back end only has the associated picture display of the pre/alarm events sent from the front end, the monitoring center staff can easily complete the monitoring of the entire system. The function of this linked video also makes video search and playback easy and rapid, which in turn saves valuable time. Therefore, the solution of the above smart network camera is a very practical and effective solution.

It is worth pointing out that although smart network cameras have better implementations, they cannot implement all intelligent analysis and recognition functions through front-end network cameras. Because its own CPU resources are limited, to embed more complex intelligent software algorithms, often need to improve the front-end device configuration, or use a dedicated video analysis instrument to work together in order to achieve the front-end video analysis and recognition functions, so need to be properly configured .