1. Object Positioner: Precise object or feature localization is a crucial function of any detection system or machine vision -guided motion system. Traditional object localization uses grayscale correlation to identify objects. While this technique is widely used, it lacks stability when image quality deteriorates. Image quality deterioration can be caused by factors such as clutter, varying brightness, and occlusion. In contrast, geometric object localization is a more recent approach that uses the object's contour to identify the object and its features. Adept's HexSight software and the AdeptVisionsAVI system both use highly advanced contour detection technology for object identification. A reliable positioner eliminates the need for fixtures to position parts, thus saving costs.
2. Optical Setup and Illumination As is well known (Copyright CONTROLENGINEERINGChina), proper optical setup and illumination are crucial for the success of vision applications. Sometimes, even with the most suitable optical setup and illumination, slight changes in the monitored object or features require corresponding adjustments to the lighting and brightness. Examples include variations in the polished surface of a wafer or the quality of laser-etched markings in OCR applications. A stable positioning tool, such as Adept's Geometric Object Positioner, can easily handle situations where changes in light focus and brightness degrade image quality. Using appropriate software allows you to eliminate the need for necessary dimming operations, regardless of changes in image quality.
3. Complete toolsets versus bundled tools: Machine vision software is primarily sold in two typical forms—complete vision toolsets or application-specific tools for task-specific applications, such as BGA inspection. The end-user's application will determine whether to use a complete vision toolset or a multitude of specific tools.
Vision tools are general applications or algorithms that perform predetermined tasks on images or parts of images. For example, a blob detection tool can identify a set of dark or bright pixels and measure the various sizes of the blob. A key point in choosing a vision system is having a complete set of vision tools. While you may not need all the tools at the beginning of a project, your requirements will change in a few years, and new applications may require different tools. Adept's HexSight software is a comprehensive machine vision library based on popular software platforms such as Visual Basic, Visual C++, and ActiveX controls. Optional HexSight symbol tools are provided along with OCR, digit matrix, and barcode decoding tools. The AdeptVisionsAVI system comes with a simple graphical interface for programming vision and motion applications.
4. Ease of Programming and Operation: A simple and intuitive graphical interface is key to ease of use and setup. A major difference between today's machine vision products lies in their graphical interfaces. An interface should be evaluated in terms of both "setup" and "operation." It should be very complex for an engineer, but very simple for an operator. For example, calibrating a vision system should be as simple as a few mouse clicks, not like an exercise in rocket science.
5. Sub-pixel Accuracy: The resolution of a vision system is the smallest feature that the system can distinguish. For example, a 1'' field of view (FOV) using a 640x480 pixel computer image will yield a resolution of 1/640, or 0.00156''. In practice, machine vision algorithms possess sub-pixel capabilities. That is, these algorithms can measure or derive units smaller than a single pixel. Adept's geometric object localization tool is accurate to 1/40th of a pixel. All other vision system tools are only accurate to 1/10th to 1/20th of a pixel. The sub-pixel accuracy of a vision tool depends on image quality and the power of the algorithm.
6. Future Upgrades Machine vision systems have diverse applications, ranging from high-end cameras to surveillance systems. When selecting a system, future upgrades should be considered. Several vendors offer general-purpose vision software that allows end-users to configure suitable lighting, optics, and vision tools. Specialized software packages such as BGA inspection and OCR are also sold pre-configured. General-purpose vision software systems are easier to upgrade. End-users should consider future system needs based on additional cameras, changes in lighting, and changes in vision tools.
For example, a system requiring multiple cameras would need to compare the price and flexibility of a system based on an image capture card versus one based on a high-end camera. Adept's HexSight software is compatible with a variety of third-party image capture cards and cameras, including FireWire cameras.
7. Image Preprocessing Detecting feature points and defects is crucial, regardless of brightness or differences in object surface or material. Image preprocessing algorithms can amplify feature points in an image, enabling vision tools to detect them more effectively. Conversely, feature points can be reduced in size so that vision tools can ignore them. For example, HexSight's morphological operations can be used to remove or fill holes in objects, separate objects at sparse points, or connect adjacent objects. Similarly, filtering operations can be used for convolution of the input image. The availability of such a wide range of preprocessing algorithms is key to the success of complex vision applications.
8. Vision-Guided Motion If your application requires a vision system to guide the robot, you must understand how the vision and motion systems are integrated. For calibration and operation, an unintegrated motion and vision system is a preliminary system; the robot or mechanism and vision system are calibrated separately. In operation, a separate vision system calculates the offset of the part's position based on its known location in the vision coordinate system, and then sends instructions to the robot arm to pick up the part at the offset from the initially programmed pick-up position.
9. System Integration If you are not very proficient in machine vision technology, you will need a system integrator for your project. Ideal vision products are widely accepted by system integrators. Adept has a program called Adept Recommended Vision Integrators (PVIs), which trains system integrators on Adept products. These system integrators have application engineering resources to handle the most challenging machine vision projects. Furthermore, these PVIs are well-suited for working with Adept's robotics and motion products. This is an ideal resource for projects requiring both motion and vision.
10. Factory Layer Connectivity Currently, various methods exist for communicating with vision systems, including common interfaces such as serial ports (RS-232), RS-485, parallel ports, Ethernet, DeviceNet, and digital I/O. Newer interfaces such as IEEE-1394 and USB are also widely used. When evaluating a vision system, the connectivity of the factory layer must be considered.
Typically, a machine vision system is a data acquisition system that interfaces with other plant-level equipment and the plant's information systems. Some vendors' software allows for remote operation of the vision system over a local area network (LAN) or the Internet. In specific industries, such as pharmaceuticals, machine vision systems are required to be application-specific, ensuring that setup is not disrupted by remote operation. When selecting a vision system, its communication interface is an important consideration and should not be overlooked. AdeptVisionsAVI controllers can connect to RS-232, RS-485, DeviceNet, Ethernet, and Firewire interfaces.
