Camera Placement

Camera placement can be characterized by either overview or detail view. The camera placement influences the resolution, frame rate and codec in use.

Overview

A camera with an overview scene is monitoring a large area such as a parking lot or a traffic camera that is viewing vehicle congestion or the number of cars parked in the lot. Because details are not important, standard definition cameras using a wide-angle lens may be sufficient. The preferred codec may be MPEG-4 with a relatively low frame rate, 1-5 frames per second. 

Overview cameras may be supplemented with a detail view camera focused on a key area of interest or by a PTZ camera to provide real-time analysis of areas of interest at a higher resolution.

Detail View

The detail view placement is targeted at observing a specific area of interest at a higher resolution than the overview. Detail view is used for Point-of-sale transactions and face or license plate recognition. The detail view may have a PTZ capability, or the camera may be close to the subject area or have a long focal length lens. Megapixel or HD cameras may be deployed to provide a sufficient number of pixels per-foot to accurately represent the subject.

The positioning of a camera for detail view is a function of the number of pixels per-foot required for the application.

Detection, Recognition, Identification

Detection, recognition, and identification are visual processes associated with the amount of detail discernable to the human eye. We detect an object when it enters the field of view. Detection means we are aware that an object (or person) now exists where previously it was not seen. Usually, this is due to movement of the object into the field of view of the surveillance camera. Detection simply means we are aware of the object, but have too little details to recognize or identify the object.

As the object moves closer, we may recognize the object from characteristics previously encountered. For example, aircraft recognition is taught to military ground troops and airmen. All aircraft have wings, engines, a fuselage, and tail assembly. They differ in size, shape, number, and position to each other. A particular model of aircraft can be recognized by recalling these characteristics from pictures, drawings or past detailed observations.

Identification is the process where sufficient details are available to uniquely discern a person or object that is previously unknown. Identification requires sufficient detail to accurately describe or recall the characteristics of the subject at a later time. For example, a mug shot (booking photograph) is taken following the arrest of a subject as a means of photographing (recording) sufficient details for later identification by a victim or witness. In video surveillance terms, sufficient detail is calibrated in pixels per foot of the area recorded by the camera.

The number of pixels per-foot to identify a subject may, at a minimum, range from 40 to over 150. If the goal, therefore, is to identify a person entering through a standard 7-foot high doorway, the camera would need to be positioned so that the pixel per-foot requirement covering the door is met. The door would then need to be covered by 1050 pixels, if the goal is to have 150 pixels per foot; 7 feet x 150 pixels per foot.

As shown, the video surveillance image is subject to uneven lighting, the subject is standing near a large window of a lab environment. There is little light from the internal space with the natural light entering from the side and rear in this scene. This image is from an analog camera that does not include a wide-dynamic range processing that would improve the image quality in this deployment. This illustrates the point that the number of pixels alone does not guarantee a high quality image.

Number of Cameras per Location

The number of cameras at any one building or facility may vary greatly depending on the coverage requirements and the nature of the business. While there are some small office deployment scenarios where only a single IP camera is needed, in most cases even a small office will require more cameras that one might initially expect.

There is a camera behind each teller station, a camera on the main entrance (both inside and outside), and two cameras in the inner office area focused on the lobby and half doorway leading into the manager office areas. Additionally, the parking lot area, side, front, and rear of the branch as well as any exterior ATM would need be covered. This small location may easily require 10 to 16 IP cameras. The Cisco Video Management and Storage System (VMSS) Network Module for the ISR router is targeted at a 16 to 32 camera deployment any may be implemented in this branch location.

Larger facilities require more cameras per location. It is not uncommon for a large retail store, home center, or warehouse retailer to need 100 to 200 IP cameras per location. Public school deployments may need 80 to 100 cameras per building.

Tip One advantage of deploying high definition cameras over standard definition is fewer cameras may be required to cover an area of interest with a similar number of pixels per foot.

Frame Rates

As image quality and frame rate increase, so does bandwidth requirements. The frame rate selected must meet the business requirements, but it does not need to be higher than what is required and should be considered carefully as frame rate influences both bandwidth and storage requirements.

Motion pictures are captured at 24 frames per second (fps). The human eye/brain sees images captured at 24 fps as fluid motion. Televisions use 25 fps (PAL) or 30 fps (NTSC) as does analog video cameras. These full motion rates are not needed for all video surveillance applications and in most applications less than 12 to 15 fps is sufficient.

The following are some industry guidelines:

•Nevada Gaming Commission (NGC) standards for casinos—30 fps

•Cash register, teller stations—12 to 15 fps

•School or office hallways —5 fps

•Parking lots, traffic cameras, overview scenes —1 to 3 fps

•Sports Stadiums on non-event days, less than 1 fps

Movement in Relation to Camera Placement

If the camera is placed where the subject moves toward the camera or vertically, the number of frames per second can be less than if the subject moves from side to side or horizontally within the field of view. The velocity of the subject is also a consideration. A cameras observing persons jogging or riding a bicycle may require higher frame rates than a person walking.

Progressive Scanning

Analog cameras capture images using an interlaced scanning method, odd and even scan lines are done alternately. There is approximately 17 ms delay between the scanning of the odd and even lines making up the entire image. Because of this slight delay between scan passes, objects that are moving in the frame may appear blurred while stationary objects are sharp. Most IP cameras use a progressive scan that is not subject to this problem. Everything being equal, a progressive scan image has less motion blurring than an interlace scanned image.

Wide Dynamic Range Imaging

The Cisco 2500 Series Video Surveillance IP Camera offer wide dynamic range imaging. This technology increases the image quality in harsh lighting conditions, including back lighted scenes or indoor/outdoor areas such as loading docks or stadiums.

IP Transport

IP cameras and encoders communicate with the Media Server in different ways, depending on the manufacturer. Some edge devices may support only MJPEG over TCP, while others may also support MPEG-4 over UDP.

TCP

MJPEG is typically transported through TCP. TCP provides guaranteed delivery of packets by requiring acknowledgement by the receiver. Packets that are not acknowledged will be retransmitted. The retransmission of TCP can be beneficial for slightly congested network or networks with some level of inherent packet loss such as a wireless transport. Live video rendering at the receiving end may appear to stall or be choppy when packets are retransmitted, but with the use of MJPEG each image stands alone so the images that are displayed are typically of good quality.

UDP/RTP

MPEG-4/H.264 video is typically transmitted over UDP or Real-time Transport Protocol (RTP). UDP does not guarantee delivery and provides no facility for retransmission of lost packets. RTP/UDP transport is most suitable for networks with very little packet loss and bandwidth that is guaranteed through QoS mechanisms. MPEG-4 over RTP/UDP is relatively intolerant to packet loss; if there is loss in the stream, there will typically be visible artifacts and degradation of quality in the decoded images.

UDP transport does provide the option of IP multicast delivery, where a single stream may be received by multiple endpoints. In an IP multicast configuration, the internetworking devices handle replication of packets for multiple recipients. This reduces the processing load on the video encoder or IP camera and can also reduce bandwidth consumption on the network.

Some IP cameras and encoders also provide for TCP transport of MPEG-4. TCP encapsulation can be beneficial for networks with inherent packet loss. TCP may be useful especially for fixed cameras and streams that are only being recorded and not typically viewed live. TCP transport induces a little more latency in the transport due to the required packet acknowledgements, so may not be a desirable configuration for use with a PTZ controlled camera.

Required TCP/UDP Ports

IP Unicast

Applications that rely on unicast transmissions send a copy of each packet between one source address and one destination host address. Unicast is simple to implement but hard to scale if the number of sessions is large. Since the same information has to be carried multiple times, the impact on network bandwidth requirements may be significant.

The communication between the Media Server and the viewers is always through IP unicast, making the Media Server responsible for sending a single stream to each viewer.

Note The Media Server only supports IP unicast between the Media Server and the viewers.

Network Deployment Models

This chapter provides a high-level overview of different deployment models and highlights the typical requirements of campus and wide area networks. Cisco's Enterprise Systems Engineering team offers detailed network designs that have been deployed by enterprise customers to provide enhanced availability and performance. Campus Networks

An infrastructure that supports physical security applications requires several features from a traditional campus design. A hierarchical campus design approach has been widely tested, deployed, and documented. This section provides a high-level overview and highlights some of the design requirements that may apply to a video surveillance solution. A traditional campus design should provide the following:

•High availability—Avoid single points of failure and provide fast and predictable convergence times.

•Scalability—Support the addition of new services without major infrastructure changes.

•Simplicity—Ease of management with predictable failover and traffic paths.

A highly available network is a network that provides connectivity at all times.

As applications have become more critical, the network has become significantly more important to businesses. A network design should provide a level of redundancy where no points of failure exist in critical hardware components. This design can be achieved by deploying redundant hardware (processors, line cards, and links) and by allowing hardware to be swapped without interrupting the operation of devices.

The enterprise campus network is a typical campus network. It provides connectivity to several environments such as IDFs, secondary buildings, data centers, and wide area sites. An Intermediate Distribution Frame (IDF) is the cable infrastructure used for interconnecting end user devices to the Main Distribution Frame (MDF) or other buildings and is typically located at a building wiring closet.

Quality-of-service (QoS) is critical in a converged environment where voice, video, and data traverse the same network infrastructure. Video surveillance traffic is sensitive to packet loss, delay, and delay variation (jitter) in the network. Cisco switches and routers provide the QoS features that are required to protect critical network applications from these effects.

Hierarchical Design

The goal of a campus design is to provide highly available and modular connectivity by separating buildings, floors, and servers into smaller groups. This multilayer approach combines Layer 2 switching (based on MAC addresses) and Layer 3 switching or routing (based on IP address) capabilities to achieve a robust, highly available campus network. This design helps reduce failure domains by providing appropriate redundancy and reducing possible loops or broadcast storms.

With its modular approach, the hierarchical design has proven to be the most effective in a campus environment. The following are the primary layers of a hierarchical campus design:

•Core layer—Provides high-speed transport between distribution-layer devices and core resources. The network's backbone.

•Distribution layer—Implements policies and provides connectivity to wiring closets. This layer provides first-hop redundancy such as Hot Standby Router Protocol (HSRP) and Gateway Load Balancing Protocol (GLBP).

•Access layer—User and workgroup access to the network. Security and QoS can be defined at this layer and propagated to the higher layers.

In smaller environments, it is typical to collapse the distribution and core layers into a single layer.

Wide Area Networks

A wide-area network (WAN) is used to connect different local-area networks (LANs) and typically covers a broad geographic area. WAN services are leased from service providers who provide different speeds and connectivity options.

Deploying a video surveillance solution through a WAN environment presents challenges that are not typically seen in a LAN. In a LAN environment it is common to see 1 Gbps and 10 Gbps of bandwidth, while in a WAN environment, most connections are less than 10 Mbps; many remote connections operate on a single T1 (1.544 Mbps) or less.

These inherent bandwidth constraints require careful evaluation of the placement of cameras and Media Servers and how many viewers can be supported at remote sites simultaneously. By using child proxies, bandwidth requirements can be reduced to transport video streams across WAN connections.

The placement of recording devices also becomes important. The video may be streamed to a central site using lower frame rates or resolution, but another attractive alternative is to deploy Media Servers at the remote sites and stream the traffic using the LAN connectivity within the remote site.

A point-to-point or leased line is a link from a primary site to a remote site using a connection through a carrier network. The link is considered private and is used exclusively by the customer. The circuit usually is priced based on the distance and bandwidth requirements of the connected sites.

Technologies such as Multilink PPP allow several links to be bundled to appear as a single link to upper routing protocols. In this configuration, several links can aggregate their bandwidth and be managed with only one network address. Because video surveillance traffic requirements tend to be larger than other IP voice and data applications, this feature is attractive for video surveillance applications.

Hub-and-spoke, also known as star topology, relies on a central site router that acts as the connection for other remote sites. Frame Relay uses hub-and-spoke topology predominantly due to its cost benefits, but other technologies, such as MPLS, have mostly displaced Frame Relay.

Example 1: Network Bandwidth Usage

Two OM Viewers are displaying video streams from Camera 1 and Camera 2 while one OM Viewer is displaying three video streams: two streams from Camera 1 and one stream from Camera 2. The network bandwidth required to display video streams for Camera 2 in Site A are relatively small for a LAN environment, but the traffic from Camera 1 can be significant for WAN environments since four different 1Mbps streams have to traverse the WAN locations.