Smart cameras can process imagery autonomously before initiating responses of their own accord. In industrial applications, they are used for quality control or to control plants and systems. As surveillance cameras, they analyse videos themselves and only alert personnel in dangerous situations.
Whether used for quality control in industry, building surveillance, or even for automated driving. The demand for around-the-clock, detailed video recording is growing. Its generating large data volumes that necessitate high transmission rates and a great deal of storage.
Smart cameras and image processing combined
Smart cameras are becoming more popular for all the reasons mentioned above. These smart systems combine cameras and high-performance image processing in a single unit. With the aid of these edge devices, image pre-processing or even full video analysis is performed in the camera itself. This relieves the host system. As such, additional PCs or controllers are not required. Depending on the performance of the integrated processor or FPGA, such cameras are able to identify colours and shapes. And then compare these to specified structures. They even detect human movement and behaviour. In the course of such tasks, methods from the fields of machine learning and AI are also employed.
Controlling plants and systems in real time
In industrial applications, this enables smart cameras to make decentralised decisions and react appropriately to defective parts, for example. This might involve activating an ejector device via PLC or notifying a human colleague who is monitoring the production process. Smart cameras are being deployed more frequently for typical real-time tasks, such as in pick-and-place applications. These smart cameras with powerful processors are increasingly sidelining the PC-based solutions that were previously the norm. With networked plant and robot control, embedded image-processing systems of this kind can provide comprehensive support for complex manufacturing steps.
To give one example, the company Wente/Thiedig is specialises in industrial image-processing solutions. It developed the SKG500 detection system for a client from the automotive sector. It determines the situation and position of heavy components in boxes to enable a robot arm to grip them precisely. Two smart camera systems of type VCSBC6211nano-RH manufactured by Vision Components act as independent image-processing systems.
They optically assess the contours of the uppermost blank and determine its position, lateral situation and orientation before picking. The information gathered from the 3D image is then sent via Ethernet to the plant controller by the smart camera. The fact is that the 3D detection system takes the form of an edge solution facilitates a reduced cycle time. Which in turn guarantees optimum utilisation of the subsequent processing centres.
Human, animal or vehicle?
Alongside industrial image processing, security is the second major application area for smart cameras. With integrated intelligence, modern surveillance cameras can autonomously interpret the recorded video data. In the event of potential security risks, they alert security personnel or the apartment owner in real time. Using preconfigured alarm rules. Smart-video analysis automatically detects rule violations, such as when a person enters a blocked-off area within the recorded image range.
The smart indoor camera from Netatmo can differentiate whether a car, a dog or a person has entered the area. It analyses the recorded images in real time and informs the homeowner accordingly via a smartphone app.
The indoor counterpart from Netatmo, the smart outdoor camera, can even recognise faces once they have been “trained”. However, enabling it to identify the person entering the home and pass this information on to the respective resident. In addition, the system can differentiate between a pet’s movements and other activities thanks to artificial intelligence. This helps to avoid alarms being triggered by pets. The user is only informed when something significant happens.
Racing drivers under surveillance
Surveillance cameras for professional applications are even more capable. Network cameras from Bosch feature 17 different video-analysis algorithms, for instance. This enables them to differentiate between actual safety or security incidents. Integrated video analysis enables large volumes of recorded video material to be quickly scanned for critical information. As only relevant images are transmitted, the network load and required memory are considerably reduced.
One rather atypical application scenario for these cameras can be found at the Misano World Circuit, one of the most famous MotoGP circuits in the world.
Here, the safety of the participating motorcyclists is of paramount importance to the circuit’s operators. Rule infringements on the part of the drivers must also be promptly detected. According to the rules of the Grand Prix, these are punished by points deductions. Yet maintaining a clear overview throughout 27 laps of a MotoGP race is no mean feat. With around 25 motorcyclists and speeds of more than 300 km/h, monitoring the circuit represents a real challenge.
For this reason, Bosch installed Autodome IP starlight 7000 HD cameras. Their video-analysis function was custom-configured for operations in Misano. The cameras automatically track race participants, enabling their safety and adherence to the rules to be monitored at any time. If an accident occurs, the cameras automatically inform the control centre of this. In this way, the situation on the racing circuit can be quickly analysed, while swift reactions are ensured.