Machine vision is primarily used in manufacturing for several key areas, including visual guidance, dimension measurement, product inspection, and object recognition. In these areas, one of the most fundamental algorithms is product recognition and localization—for example, in visual-guided robots: the robot must first identify the product to be picked up from an image and determine its precise coordinates before it can guide itself to the correct position. The same principle applies to dimension measurement and product inspection—before performing any measurements or inspections, it’s essential to first confirm whether a product is present and to pinpoint its exact location, so that subsequent analytical tools can be properly applied.

Therefore, product identification and positioning is a fundamental issue.

1.Components of a visual positioning system

The robot positioning system based on machine vision comprises a camera system and a control system.

The camera system includes a computer (equipped with an image acquisition card) and a camera, which primarily capture visual images and apply machine vision algorithms. The control system comprises a control box and a computer, enabling precise control over the end-effector’s position. The workspace is imaged using a CCD camera, and the computer processes these images to identify tracking features, perform data computation and recognition, and then uses inverse kinematics to determine the positional error at each robot joint. Based on this error information, the high-precision end-effector module is controlled, allowing for scientific adjustment of the robot’s position and orientation.

2.Key Factors of Visual Positioning Systems

In the field of industrial production, especially in the application of industrial robots,Visual Recognition and Localization SystemThis is particularly important. In actual production, we need to pay attention not only to whether objects can be grasped accurately but also to the speed of the operation—a challenge that has long plagued the industry. Typically, when we encounter industrial robots performing gripping tasks, their speeds tend to be relatively slow. Yet once we try to increase the speed, the precision of the grasp often suffers. This is precisely the crux of the challenge faced by visual recognition and positioning systems. Next, let’s join Xiao Ju as we explore this issue further.

First is the data volume,In a relatively complex production environment, the system needs to accurately locate and identify the products that require recognition.

Next is speed,How can we boost the speed of some standard production lines to the millisecond level? While previous algorithms could still function effectively under normal conditions, as algorithms continue to evolve, deep-learning algorithms often require more powerful GPUs to achieve optimal performance.

Then comes the heart of the issue: positioning accuracy.In deep learning systems, the images we see are all subject to some degree of scaling. Therefore, the entire system needs to achieve pixel-level alignment with the original image.

The remaining challenge is recognition accuracy. In many cases, we have very limited learning data. Under these circumstances, how can we further improve the accuracy of recognition?

3.Visual localization faces challenges.

If you want to design a feasible algorithm for product recognition and localization, you’ll need to overcome several challenges:

1、Quickly specify the product

Industrial products vary widely in their characteristics. Therefore, for each specific application, it’s necessary to quickly identify the target product from just a few images—or even a single image. For example, if a production line needs to locate the position of rivets, taking a single photograph and performing the corresponding training allows the system to subsequently search for and pinpoint those rivets in subsequent images.

2、Quick product search

For a 2-megapixel image, it is typically required to identify and locate the product’s position within tens of milliseconds.

3High-precision positioning

Industrial production has stringent requirements for precision and tolerances; therefore, product positioning must be as accurate as possible. Currently, there is a widespread demand that identification and positioning algorithms achieve positioning accuracy at the pixel level—or even subpixel levels.

4、Can adapt to the effects of product absence, occlusion, and dirtiness.

If a product is partially occluded, resulting in a missing portion of the product in the image, it should still be possible to detect and locate the object. Conversely, if the surface of the product becomes dirty, causing changes in its surface features, it should still be possible to detect and locate the object.

5、Can adapt to the effects of uneven light brightness.

If the product’s brightness changes—for example, becoming half bright and half dark—it should still be identifiable and locatable.

6、Products that can be recognized by rotation

The product can typically rotate within a 360-degree range.

7、Can recognize multiple products

An image may contain multiple products, which need to be identified and located separately.

8、Can accurately identify nearly symmetrical objects.

Objects that are nearly symmetrical can easily be misidentified as being oriented in the wrong direction, so corresponding design adjustments are necessary.

9、Can handle objectsPolarity reversal

For example, the product being learned has white background with black text, but in reality, the product image might have a black background with white text—so it needs to be able to recognize that.