Accurate Real-Time 3D Color cameras with Full HD

Inspired by the Kinect

Although different 3D cameras and scanners have existed for some time, present solutions have been limited by several unwanted compromises. If you wanted high speed, you would get very low resolution and accuracy (e.g. Time-of-Flight cameras and existing stereo vision cameras, which despite being fast typically have resolution in the millimeter to centimeter range). If you wanted high resolution and accuracy, you would typically get a camera that was slow and expensive (e.g. the high accuracy scanners).

The Zivid 3D camera provides full HD 3D color snapshots at 100ms per image (incl. acquisition and 3D processing time), while at the same time delivering depth resolution at 100µm. (Bild: Zivid 3D camera image courtesy of Zivid Labs AS/Yumi robot image courtesy of ABB Robotics)

Additionally, most 3D cameras and high accuracy scanners only had depth data with no color information. This changed with the first generation of Microsoft Kinect, a real-time 3D RGBD camera which, despite being made for Xbox and gaming with no industrial packaging, was adopted by robot researchers in masses around the world. Although creating significant attention in robotic vision research, the Kinect did not provide the data quality and robustness needed for broad usage in industrial automation. It was just not accurate enough to be used in many industrial applications, such as e.g. inline inspection of 3D shape in production lines, or in robot guidance or machine vision applications requiring sub-millimeter accuracy. Inspired by the Microsoft Kinect and the likes, Zivid Labs has introduced a new kind of 3D color camera without the common compromises. Several innovations set the Zivid camera apart and define the scope of applicability. Providing high quality 3D snapshots in full HD and color at rapid rates (100ms per image, including both acquisition and 3D processing time), while at the same time delivering true to reality depth resolution at 100µm. That is like the width of a human hair and about 50 to 100 times more accurate than other high-speed 3D color cameras in the market. The availability of both 3D and full color (RGBD) data makes it much easier to locate and characterize features in an automation operation. Whether it’s a PCB (with components in different colors), a multitude of consumer products in a ware-house rack, a multi-color plastic material or a food product (e.g., for inspection or processing of fruit, fish or meat, where color makes it easier to distinguish e.g. between fat, blood and connective tissue). Data quality lies at the core of any successful automation operation, and is extremely important to make systems robust, reliable and flexible towards rapidly changing production or logistics operations with a broad variety of products and parts. In machine vision objects such as black and absorbing (e.g. carbon or dark rubber), partly shiny or reflective objects (e.g. semi polished metal or plastic wrapped items) or partly translucent materials (e.g. fish or meat) has always been a challenge. To address this, the 3D camera comes with several flexible High Dynamic Range (HDR) modes in addition to a smart and adaptive projection technology and noise filtering, which has been shown to make a big difference.

Comparison of the Microsoft Kinect 2 (Time-of-Flight technology) and the Zivid 3D camera. The Zivid camera is about 100 times more accurate (50 to 100µm compared to 5 to 10mm. (Bild: Image courtesy of Zivid Labs)

Application Areas and Availability

Some possible application areas include robust detection and pick&place of randomly organized objects, challenging depalletizing tasks, order picking and preparation in a logistic warehouse, demanding kitting and assembly operations, automated positioning in jig-less setups or high accuracy guiding and control. Additionally, by taking precise 3D images of components and revealing small defects or anomalies (either in shape or color), it allows for fast inspection and quality control. By being so fast, it can be integrated into the production line and inspect 100% of the production. It’s also suited for lightning fast 3D digitization of objects, e.g. for 3D printing, 3D visualization and a broad set of other industrial automation applications. The real-time 3D color camera is now available for purchase to an affordable price and with low lead time. The feedback from those who have the camera running in industrial setups, has been very positive.

Accurate Real-Time 3D Color cameras with Full HD
The Zivid 3D camera provides full HD 3D color snapshots at 100ms per image (incl. acquisition and 3D processing time), while at the same time delivering depth resolution at 100µm. (Bild: Zivid 3D camera image courtesy of Zivid Labs AS/Yumi robot image courtesy of ABB Robotics)


Das könnte Sie auch interessieren

Yaskawa hat den Erweiterungsbau seines Robotikzentrums am Standort Allershausen bei München eingeweiht. Im Rahmen einer Eröffnungsfeier mit Kunden und Partnern des Unternehmens sagte Yaskawa-Europa-President Manfred Stern: „Wir sehen uns immer stärker als europäischer Player der Robotik.“

Anzeige

Einen Gegenstand entgegennehmen oder ihn einer anderen Person reichen: Für Menschen gehört das zu den natürlichsten Handlungen. Doch für Roboter ist es eine schwierige Angelegenheit.

Anzeige

Ocado Technology, die Forschungsabteilung des großen Online-Supermarktes Ocado, hat ein neues Robotersystem entwickelt, das eine große Anzahl der 50.000 verschiedenen Artikel auf der Website des Supermarktes greifen kann. Es verwendet ein proprietäres Computer-Vision-System, das von der Forschergruppe konzipiert wurde, um Greifpunkte für einen bestimmten Gegenstand zu berechnen, ohne dass ein 3D-Modell des zu entnehmenden Objekts benötigt wird.

Anzeige

Vor kurzem fand das Richtfest für das neue automatische Kleinteilelager von Eisele Pneumatics in Waiblingen bei Stuttgart statt. Das 1939 gegründete Familienunternehmen entwickelt, produziert und vertreibt als Anbieter von Verbindungslösungen mit rund 160 Mitarbeitern Anschlüsse aus Ganzmetall, darunter Steckanschlüsse, Schnellschlusskupplungen und Schlauchverschraubungen. Das neue Kleinteilelager soll die schnelle Verteilung der Waren sicherstellen.

In der achten Episode der YouTube-Interviewreihe Robots in Depth spricht Ian Bernstein, Gründer verschiedener Robotikfirmen wie Sphero, im Interview mit Robotikexperte Per Sjöborg über seine Erfahrungen beim Aufbau eines weltweiten Vertriebsnetzes sowie über die Herausforderungen der Software- und Hardware-Entwicklung.

Das von einem Forscherteam des Robotics Innovation Center am DFKI und der Universität Bremen verfasste wissenschaftliche Paper ‚Intrinsic interactive reinforcement learning – Using error-related potentials for real world human-robot interaction‘ beschreibt ein Verfahren des Maschinellen Lernens, bei dem ein Roboter aus dem eigenen Fehlverhalten in der gestengesteuerten Interaktion mit dem Menschen lernt.

Anzeige
Anzeige
Anzeige
Anzeige
Anzeige
Anzeige