Imago
End of continuous monitoring
Event-Based Vision is a completely new and cost-effective approach to monitoring fast-moving processes that differs fundamentally from conventional image processing concepts.
The idea is that data is only recorded from the moment a specific event occurs.
© ImagoCounting objects, monitoring moving parts or checking vibrations is possible with image processing, but there are technological limits: The faster the processes to be monitored, the higher the demands on the vision systems - and therefore naturally also their price. As a result, many innovative ideas are abandoned if a fast and therefore expensive image processing system is required to implement them and there is therefore a risk that this will not pay off. Other ideas are often not even tackled because the image processing algorithms are considered too complex and error-prone.
A practical example: in order to avoid expensive repairs and long downtimes, it makes sense in certain cases to analyze vibrations in machine parts and detect them in good time if these components vibrate outside specified limits. One technical option for solving such tasks could be sensors that have to be adapted to the components. However, this approach is complex and may require a new design.
Theoretically, such a task can also be solved with image processing systems. However, the main problem when using this technology is the selection of the image refresh rate: if an event occurs between the images, it is not registered and the desired information is not available. Users could therefore use ever faster and therefore more expensive vision systems and still miss their target: Regardless of whether processes are observed at repetition rates of 100, 1000 or 10,000 Hz, events between two images can remain undetected.
Event as a starting signal
An alternative solution is now available for tasks of this type. The basic idea: data is only recorded from the moment a specific event occurs. This mode of operation is called event-based, describes a completely new concept for monitoring fast processes and differs fundamentally from typical camera sensors that deliver images at a certain repetition rate, which are then evaluated.
This technology is based on the 'Monet' event-based sensor, which is developed and produced by the French company Prophesee. Intelligence is already implemented in each pixel on the chip itself, allowing it to detect movements and send an event independently, without a predefined line or image cycle. Each pixel therefore decides for itself when it triggers an event and sends data to the evaluation computer. The Monet chip works with frequencies of up to 20,000 Hz and is therefore also suitable for applications in which very fast movements have to be detected.
Events instead of pictures
Imago Technologies has developed the 'VisionCam EB' camera system based on this sensor in the format of a typical smart camera. If an object moves in front of a 'VisionCam', more than 300,000 pixels are able to capture this event and trigger events. The event-based chip is directly connected to the dual-core processor with ARM architecture and Linux operating system integrated in the 'VisionCam EB'. In contrast to conventional image processing systems, this architecture no longer stores image after image in a memory and then evaluates them using suitable image processing algorithms. Instead, events are continuously stored in the working memory, which are provided with coordinates and a time stamp.
Compared to image processing, event-based vision is no longer referred to as 'images', but as two-dimensional signal processing. The art of mathematics consists of interpreting the stored events. Prophesee provides a library to simplify the mathematics for a range of applications. This can be used directly for vibration analysis applications, for example. However, the developer must develop an application-dependent program for the Linux OS, including a web GUI. Furthermore, developers are free to invent and integrate their own algorithms.
Parts that fly through the detection range of a 2D event-based sensor trigger a large number of events that can be easily analyzed statistically.
© ImagoAnother application example illustrates how event-based vision works: an NOK part is 'shot' out of a machine using compressed air and needs to be detected in a dusty environment. This task is not easy to solve for conventional image processing systems, but is no problem for the 'VisionCam EB': The ejected part flies through the detection area of the 2D event-based sensor and triggers a large number of events there, which can be statistically evaluated much more easily compared to the few detectable pixels of a high-speed camera. The system is also very robust.
Lighting and optics simplified
Event-based systems also have clear advantages in terms of the lighting and optics used. In classic image processing systems, LED lighting is often flashed to generate the necessary brightness for the image. Flash controllers are used to synchronize with the image sensor, which results in higher costs and more effort when setting up an application. In an event-based application, on the other hand, flashed lighting would trigger events, so flash controllers are not required. Simple LED spotlights that provide constantly bright lighting are completely sufficient. The question of the optimum optics for event-based systems is also answered very quickly: standard lenses with a C-mount connection meet all requirements.
The development environment itself consists of a Microsoft Visual Studio editor, which is connected to the 'VisionCam EB' via a plug-in and compiles the code there. Similar to conventional image processing systems, the 'VisionCam EB' provides the application developer with debugging information. The camera's SDK serves the required interfaces such as grabbing events, the 1 Gbit/s Ethernet, digital I/Os or a serial interface. Example programs of the simplest kind show, among other things, how events can be evaluated with statistics.
More sophisticated algorithms are available in the Metavision Library from sensor manufacturer Prophesee. Additional support is also available for the development of a web-based GUI. Due to the effort involved in developing a completely new application, the 'VisionCam EB' is more suitable for series applications than for individual projects.
The camera offers a resolution of up to 5 MPixel, which covers a wide range of applications. With the intelligent line scan camera 'VisionCam LM', Imago also has another Linux-based, freely programmable smart camera in its range, which has sufficient computing power on board for a number of applications and can be used without an external computer or camera cable. This product family is complemented by a little brother, the 'VisionSensor'. This device is based on an ARM CPU and a WVGA sensor, is in the price range of a typical sensor and enables the individual design of complete image processing applications.
Imago will soon also be presenting the 'VisionSensor' in an event-based version. The target customers for these products are image processing engineers who understand their specific application, develop optimized solutions for it, require the source code for any extensions and, in the case of line scan and area scan cameras, optionally use MVTec 's 'Halcon' library.
You might also be interested in
For 3D detection and distance measurement
Teledyne e2v has announced the Bora CMOS time-of-flight sensor, designed for 3D detection and ranging to support industrial applications including vision-guided robotics, logistics and surveillance.
read more...Quo vadis image processing?
Innovations in the fields of artificial intelligence and embedded vision as well as special processors and development environments for image processing open up a broader spectrum for the technology. An overview.
read more...Smart camera for edge AI applications
The 'Vision Cam XM2' from Imago Technologies is a compact smart camera based on the Nvidia Jetson Orin module. The embedded vision system is suitable for edge AI applications and other tasks that require multi-core ARM and GPU computing power.
read more...The e-paper special: looking for products of the year 2024
They've earned it: all 96 nominated products from 12 categories in their own e-paper. Browse through the individual areas and make your personal selection. You can also find out more about the companies and their solutions in the accompanying video...
read more...Image processing AI as a hobbyhorse
The edge AI system 'Vision Box AI' from Imago Technologies is based on Nvidia's 'Jetson AGX Orin' computer module. It is designed for vision applications and inference programs.
read more...Imago Technologies at the SPS 2023
Touch panel for vision applications
Imago launches the all-in-one panel PC 'Vision Box Touch' on the market.
read more...New condition monitoring camera
Imago Technologies is launching a new condition monitoring camera on the market: the 'Industrial DashCam 1 000'.
read more...Products of the year 2023 - your vote counts!
96 products in 12 categories are up for election for the "Products of the Year 2023"! Take part and win Saturn vouchers worth €500, among other things!
read more...AI industrial camera for beginners
Optical inspection using AI becomes easier
Artificial intelligence is penetrating more and more machine vision applications and is also becoming increasingly easy to use. Imago Technologies is presenting a corresponding AI industrial camera at SPS.
read more...