VISION AWARD
by a mirror. Te per-pixel exposure of the projection is 10μs, which ensures motion blur- free, consistent data. All Photoneo’s sensors are equipped with
a band-pass filter that reduces ambient light. In addition, Photoneo’s active ambient light suppression is able to work in sync with the projector and the sensor to control light sensitivity at the sensor surface. At any given time during the acquisition, only about one per cent of the sensor surface can be exposed to direct reflections. Te camera’s control circuit can disable the rest of the sensor (99 per cent) so that it does not collect any
At any given time during the acquisition, only about one per cent of the sensor surface can be exposed to direct reflections
photoelectrons. Te technology effectively suppresses the ambient light of any source by a ratio of 1:100 and thus allows scanning under direct sunlight. Te same technology also suppresses
internal inter-reflections between pixels by the same ratio of 1:100. When two sensors operate
together in the same area, the technique rejects the second scanner’s projection in 99 per cent of the image, and the remaining one per cent of pixels can be filtered out from the result. Te PhoXi 3D camera can capture sub-
millimetre details, while its ability to scan rapidly moving objects allows it to be used in almost any situation. Potential applications for the camera include: bin picking; palletising and depalletising; quality control and metrology; autonomous delivery systems; people counting and behaviour monitoring; object sorting; face recognition and agriculture.
www.photoneo.com
A software-programmable vision system-on-chip for high-speed image acquisition and processing By Jens Döge, Christoph Hoppe, Peter Reichel, Nico Peter, Ludger Irsig, Christian Skubich, Patrick Russell and Andreas Reichel, Fraunhofer Institute for Integrated Circuits
T
he IAP VSoC2M is a novel member of the Fraunhofer vision system-on-chip (VSoC) family for high-performance
image acquisition and processing applications. It combines analogue convolution during the readout process; fast column-parallel processing for image analysis and feature extraction, including analogue storage and soſtware-defined analogue-to-digital (A/D) converters; an asynchronous readout path for sparse data compression; and an application- specific instruction set processor (ASIP) concept for soſtware-defined process control. Tis enables many new possibilities in
embedded image processing, from content- based multi-region of interest (ROI) image acquisition via optical measurement methods such as laser light sheet triangulation or optical coherence tomography (OCT), to feature- based process control. Applications such as laser light sheet
triangulation or OCT rely on a few or small ROIs, and certain optical control tasks may only require scalar output. Tus, reducing the amount of data as early as possible speeds up A/D conversion and subsequent processing steps, mitigating the digital interface bottleneck of traditional high-speed image sensors. A VSoC eliminates the need for external image processing hardware, enables latency below 100μs, and full frame rates above 10kHz. VSoC2M consists of a two-megapixel
[It] excels in applications with high frame rate and/or latency demands, and procedures requiring compression of features
charge-based pixel matrix, enabling analogue- digital column-parallel pre-processing via 1,024 processor elements. Each processor element contains 32 analogue memory cells, a dedicated AD/DA converter and basic digital processing functions. An ASIP controls the VSoC’s functional
units, including line control. A compacting readout pipeline, operating in parallel to data processing, means the VSoC can be used
40 Imaging and Machine Vision Europe • October/November 2018
for event-based, compressed sensing and conventional methods with continuous data streams. Soſtware is developed in Python with embedded Assembler. High-speed image sensors with integrated
signal processing have been widely used for measurement applications (the Sick IVP Ranger, for example) since the 2000s. Convolutional neural network-based sensors, such as the Teledyne Anafocus Eye-RIS, are not yet as prevalent in industrial imaging due to relatively low image resolution. Te VSoC concept of column-parallel,
charge-based signal processing and flexible ASIP-based control enables 1D analogue convolution operations and soſtware-defined digitisation. VSoC2M significantly improves upon its predecessor’s readout speed and flexibility with new methods for row- and column-wise data compression. VSoC2M excels in applications with high
frame rate and/or latency demands, and procedures requiring compression of features or scalars. Its flexible, soſtware-defined nature means OEM VSoC modules can be adapted to customers’ needs and opens up many new possibilities in embedded image processing, ranging from content-based multi-ROI image acquisition via optical measurement methods (such as laser light sheet triangulation and OCT) to low latency process control. O
www.eas.iis.fraunhofer.de/en/research_ topics/
industrial_image_processing.html
@imveurope
www.imveurope.com
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72
