Objekterkennung auf einem Förderband: Wie verwendet man Computer Vision?
Die korrekte Platzierung von Computer Vision-Kameras ist erforderlich für Objekterkennung in Förderbandanlagen. Stellen Sie sich eine Produktionsumgebung vor, in der Produkte aller Formen und Größen auf einem Förderband vorbeilaufen. Dies kann eine Obstfabrik, ein großes Logistikzentrum oder ein Hersteller von Automobilteilen sein. Dieser Artikel konzentriert sich auf die Verwendung Computer Vision Kameras für die Objekterkennung von Eiern. Ein vision-System sollte über dem Förderband platziert werden, um sicherzustellen, dass hochwertige Bilder für die Eiererkennung aufgenommen werden können. Das Ergebnis sollte sein, dass alle Eier gezählt werden und jede Kiste am Ende des Förderbands genau mit 120 Eiern gefüllt ist. Darüber hinaus sollten Eier mit abnormalen Flecken herausgefiltert werden. In nur fünf Schritten wird dieser Artikel Ihnen helfen, Computer Vision Kamera-Hardware für Ihr vision-System auszuwählen.
Table of contents
Computer vision camera selection
One of our machine vision experts began by recommending a suitable computer vision camera for the customer's object detection system. We advised using our 3MP USB3 camera ‘MER2-302-56U3C’ because of the requirements for the system. The customer planned on placing the camera above the conveyor belt at a distance of approximately 3 meters from the PC. That is why we advised them to use a camera with an USB3 interface. For applications where there is no preference for a specific interface, an USB3 interface camera is recommended for multiple reasons. Firstly, compared to a Gigabit Ethernet interface camera, an USB3 camera has a four times higher bandwidth. In general, our USB3 cameras offer a price advantage and only one USB3 cable is required to enable both data transmission and power.
All our standard industrial cameras, which are being used for applications such as object detection on the conveyor belt, are available in both monochrome and color. For this application, the customer preferred using a color camera instead of a monochrome camera, which captures images in black and white. By choosing a color computer vision camera, the setup is future-proof for their wish to one day detect colors for spots differentiation.
Furthermore, the choice for a global shutter or a rolling shutter computer vision camera should be made. Our vision expert advised using a global shutter camera. This had to do with the fact that the object detection should be done while the conveyor belt is moving. Global shutter cameras are used for applications where the camera or the object is moving while the images are captured. More information about global vs. rolling shutter cameras can be found by clicking on the link.
In addition, we supported the customer in determining the required camera resolution. The customer did not know yet what resolution was necessary for the object detection on the conveyor belt. Every time the object is captured on the conveyor belt, it should be detected as an egg and checked for spots. The customer informed us that the eggs are not always positioned on
the same spot of the conveyor belt. However, they did know that the eggs will be located in an area of maximum 300x240mm. This is what we call the required field of view (FOV). A single egg has a length between 55 and 57mm. The spots on the eggs that should be detected are 0.5x0.5mm or bigger.
To see these smallest details of only 0.5mm, a system resolution of 0.16mm/pixel is required. This is based on three pixels per smallest detail. A computer vision camera with at least 1800x1440 pixels (2.59MP) is necessary to cover the full field of view.
That is why our 3MP camera with 2048x1536 pixels offers a great solution to be mounted above the conveyor belt. By using this computer vision camera for the object detection, spots of 0.5mm can also being detected.
In conclusion, our 3MP camera offers a low-priced computer vision camera solution for object detection on a conveyor belt while benefiting from the high frame rate and easy-to-use USB3
interface.
Lens for 3MP computer vision camera
Every C-mount computer vision camera needs a suitable lens. The best suitable C-mount lens for this conveyor belt application was recommended by the machine vision expert based on a few specifications. To determine the right focal length lens to reach a certain opening angle, it is important to know the required Field of View (FOV). This is the total area (horizontal and vertical in mm) of the conveyor belt that the customer wants to capture with the computer vision camera. For this application, the required field of view is 300x240mm. The customer preferred to have a Working Distance (WD) between 600 and 800mm. This is the distance between the object to be detected and the camera located above the conveyor
belt.
The advised USB3 vision camera has a resolution of 2048x1536 pixels and a pixel size of 3.45 µm. To calculate the focal length of the C-mount lens that should be used, our online Lens Calculator tool can be used. The image shows that filling in the specifications in this tool, results in a calculated focal length of 16MM.
Lens Calculator
Our 5MP 16MM C-mount lens for computer vision cameras with a maximum sensor size of 2/3”, turned out to offer a great solution. This C-mount lens is part of our very light sensitive 5MP lens series, in which both the aperture (starting from F1.4) and focus can be adjusted and fastened by screws. A slightly bigger field of view of 320x240mm will be visible when using the 3MP camera with this 16MM lens.
Accessories for computer vision applications
All our computer vision cameras, such as the advised 3MP camera, should be connected to a PC. PCs have fast processors and powerful GPUs, allowing them to capture, process and analyze images quickly and efficiently. Additional software for the object detection needs to be used for image processing. To guarantee a stable connection between the PC and camera, we advise using one of our USB3 cables to connect the camera. For example, our 1 meter USB3 cable or 4.6 meter USB3 cable can be used. To simplify the camera mounting process, we recommend using a tripod mounting plate.
In addition, to protect the camera and lens from dust and dirt, the machine vision expert advised using our aluminium machine vision camera housing. This waterproof industrial housing protects the camera setup in the dusty production environment where the conveyor belt is located and the vision system is installed. The aluminium enclosure includes two extension rings, which are both necessary to used to extend the lens tube. This is because the lens has a length between 30mm and 45mm (33,9mm). Please check the product page of the enclosure for the installation video.
Computer vision software for conveyor belt objectdetection
As mentioned, image processing software is required to detect the objects on the conveyor belt. This 3MP camera is like our other industrial cameras GenIcam compatible, which means that is can be used with a variety of 3rd party software including MvTec Halcon, Cognex Vision Pro, OpenCV and Arm Boards software.
To start, our Software Development Kit (SDK) can be downloaded for free on our website. In this SDK, camera parameters such as the exposure time and frame rate can be set. Additionally, a first image can be made and saved. The SDK supports operating systems including Windows, Linux and Android and is compatible with regular and industrial PCs and ARM platforms. The SDK supports programming languages like C++, C#/.NET, and Python.
Our customer aims to detect the eggs and defects by writing their own software in Python. Our SDK provides example programs, including the Python sample to acquire images. One of the articles in our Knowledge Center provides more information about using a Python sample using PyCharm.
Support for computer vision applications
Can we support you in the selection of the best computer vision solution for object detection in conveyor belt systems? Or a similar vision application? Simply share the requirements of your application and we will advise you on the most suitable lens and camera for your project.
