Blurring the Boundaries: IP and Machine Vision Cameras Converge
- Created on Sunday, 01 June 2014
In modern production facilities, users are more frequently combining two different strands of camera technology. Classic machine vision cameras manage inspection tasks and yield management, while network cameras (also called IP cameras) handle process monitoring and bringing production to a standstill when necessary.
In many ways, the two camera breeds are actually similar. Industrial GigE cameras typically work with the same Gigabit Ethernet technology as IP cameras. Both camera types also use the same technical protocols, including TCP/IP and UDP. The similarities, in fact, make the separate technologies — both engineered from the ground up to serve different purposes — easy to combine. Customers now want to implement both camera types to fit their specific applications, and new products are being developed that combine the best features from the IP and machine vision devices.
IP vs. Classic Industrial Cameras
The core differences between a classic machine vision camera and a network (or IP) device come in the areas of image data compression, multi-streaming, and real-time compatibility.
IP cameras are specially designed to work with low bandwidths to ensure that they can fit seamlessly in existing networks without overloading them. The camera is configured via a PC on the network to send a compressed video stream, such as MPEG-4, MJPEG, or H.264. The IP camera primarily provides a strong visual impression on the operator’s monitor.
Industrial cameras, by contrast, are engineered to work with large bandwidths and optimal image quality as part of closed image processing systems. While network cameras compress the image data down to a fraction of its original data volume, in an effort to reduce the bandwidth required to transport that data, industrial cameras deliver raw images. The unprocessed images allow users to review even the most minute of details, as is necessary for quality inspections or detail measurements. If image data is missing and the error occurs precisely in that area, leading to a false positive by the inspection system, then the system has failed in its mission.
IP cameras are also capable of multi-streaming, meaning the camera sends multiple streams in different compression formats, such as H.264, MPEG-4, and MJPEG. Each end device then accesses a suitable stream. An operator, for example, can call up a high-resolution MJPEG stream offering strong detailing on his or her monitor, even while a space-saving version in the H.264 format is submitted for archiving. Machine vision cameras work with post-processed data that has already been evaluated, such as production statistics. The devices also optimize equipment performance and adapt the equipment to eliminate recurrent errors. In general, the images are only stored after inspection.
Real-time capability, another functional distinction between the two technologies, means that image capture starts immediately after a trigger signal has been sent to the camera. Image data must be acquired, transmitted, and evaluated within a set time-frame. The requirements for the maximum acceptable reaction time between trigger signal and image acquisition can vary from microseconds up to seconds. For industrial cameras, real-time images are a prerequisite. When inspecting components in a production process, for example, the components are transported on the conveyor belts at high speed. For a precise inspection, the camera must acquire the images as quickly as the components are being transported. Timing this precisely requires low latency: a small time delay between receiving the trigger signal and acquiring the image. Further, the time delay must not vary; no jitter can affect the moments of image acquisition. For an application with high image rates (e.g. 300 images per second), the required latency times can only be microseconds.
Similar requirements also exist outside of the factory, especially in traffic applications. In speed control systems, the camera activity must be synchronized with other system components, such as illumination devices. Many classic IP cameras are not real-time capable. In typical surveillance situations, such as monitoring the activity on a banking floor, the user needs an automatically captured, continuous stream of images without the need to trigger a camera. If a camera is set for a frame rate of 30 frames (images) per second, it will internally generate the signals required to initiate an image capture every 1/30th of a second. In some other situations, however, it would be desirable to be able to trigger an image capture at a specific point in time. For example, in a traffic control situation the user may want to trigger an image acquisition immediately after a car passes a sensor on a highway.