Machine Vision

A Machine That Watches

December 1, 2006 By: Bill Silver Sensors

Introducing a new kind of optoelectronic sensor that detects and inspects objects, and synchronizes the action of an actuator with an object's position.


During an automated manufacturing process, the product typically moves along a production line at either constant or variable speeds. The first task of optoelectronic inspection is to discard that motion. For example, machine vision systems use triggers and shutters and photoelectric sensors use gates to freeze the object at a particular point in time so that it can be analyzed.

The Synchronizing Inspection Sensor

The synchronizing inspection sensor (SIS) is a new type of optoelectronic inspector that takes advantage of, rather than discarding, object motion to detect and inspect objects, and to synchronize the action of an actuator with their position. By using the multiple viewing perspectives afforded by object motion, the SIS provides capabilities beyond the reach of conventional optoelectronic inspection, without the need for triggers, gates, and careful arrangement of multiple detectors.

 The Ottakringer Brauerei AG in Vienna, Austria, is using the Cognex Checker 101, whose underlying technology is the synchronizing inspection sensor, to guarantee that beer crates and boxes of canned beverages contain the correct amount of product
The Ottakringer Brauerei AG in Vienna, Austria, is using the Cognex Checker 101, whose underlying technology is the synchronizing inspection sensor, to guarantee that beer crates and boxes of canned beverages contain the correct amount of product

To obtain images from multiple perspectives, multiple frames are captured and analyzed as an object moves through the field of view. To obtain multiple views, it is desirable that the object moves no more than a few pixels between successive frames, which requires a sensor-processor combination capable of very high frame rates. Our SIS uses an inexpensive DSP and a CMOS imager that can capture and analyze images at 500 frames/s, far higher than the 30–60 frames/s provided by vision systems for the past 25 years.

The keys to high frame rate operation are a fast shutter, low pixel transfer time, and fast image processing. All lead naturally to low image resolution. With fewer pixels, each pixel can be made much larger and therefore more light sensitive, allowing fast shutters. With fewer pixels, transfer and processing time is clearly reduced. Our SIS's imager is only 128 x 100 pixels, about 20 x fewer than the typical vision system and about 12,000 x more than a photoelectric sensor.

Operating Principle

An SIS captures and analyzes frames continuously, overlappping each frame capture with analysis of the previous one. Our SIS achieves 500 frames/s operation with a 300 μs shutter.

The following analysis steps are performed for each frame:

1. Attempt to locate an object in the frame.

2. If an object was located in step 1, inspect the object for presence or absence of one or more features.

3. Considering the recent history of frames, decide whether there is sufficient evidence to conclude that an object has been detected.

If we conclude in step 3 that an object has been detected, additional analysis steps are performed:

4. Considering the same recent history of frames, decide whether there is sufficient evidence to conclude that the object has passed inspection.

5. Produce output pulses indicating that an object was detected, and whether or not it passed inspection. The pulses occur at a predefined synchronization time, corresponding to a precise location of the object along the production line, e.g., at the time when the object crosses a reject actuator.

Note that if inspection is not required, steps 2 and 4 are skipped.

Locating and Detecting Objects

Step 1 answers two questions:

  • 1. How confident are we that an object appears in the current frame?
  • 2. If we are sufficiently confident, where is it?

Our level of confidence is a numerical value called an object detection weight. We are sufficiently confident if the object detection weight exceeds some threshold. Frames for which we are sufficiently confident that an object has been located are called active frames.

Being sufficiently confident does not mean that we are right, however. Artifacts of viewing direction, illumination, manufacturing process, and object appearance may make it seem that an object is present when none is, or not present when one is. Unlike a vision system looking at one image, with an SIS these unavoidable errors of judgment usually do not lead to incorrect decisions, because those decisions are based on multiple viewing perspectives.

1 2 3 4 


Add Comment




IIoT University


Deep Learning for Vision Using CNNs and Caffe: A Hands-on Tutorial – 9/22/16 – Cambridge, Mass


IDE






Sensors 2017 Call for Speakers


Sensors Midwest


Advertise


Subscribe



Twitter Feed

Find It Fix It Forum

Sensors invites you to join the Findit-Fixit Forum, where you can get answers to your sensing questions—concerning technologies, products, methods, applications, and services--and also offer help to your fellow engineers. The Forum covers all kinds of topics, from the basics to the extraordinary.

Join the discussion!


© Copyright 2016 Questex, LLC. All Rights Reserved. Sensorsmag. Privacy Policy | Terms of Use

If you are having technical difficulties or considerations, please contact the webmaster.