Machine Vision

A machine vision system consists of technology and methods based on creating a digital image of the environment, in order to perform automatic control, control the process or guide the robot. It includes software, hardware, integrated systems and methods.

As a term, it is mainly used in industrial automation but also in other environments, such as security systems and vehicle control.

The process of the machine vision system involves three steps:
1) Imaging
2) Automatic analysis
3) Providing the required information

The machine vision system replaces labour-intensive manual inspection with fast, automatic digital image processing. Thanks to rapid developments, systems have become significantly cheaper and offer a good return on investment.

There are two types of machine vision systems, 2D and 3D.

The scope of the machine vision system covers a wide range of tasks, from presence verification to real-time inspection and sorting.

Basic tasks of the machine vision system:
· Inspection
– checking the quality of the product, detecting defects and anomalies
· Measuring – identifying the dimensions of objects (length, width, height, area, volume)
· Reading – decoding and reading texts (barcodes, 2D codes, letter recognition)
· Positioning – detecting and locating an object.

2D vision
The image to be analysed is obtained from the camera image or barcode scanner. Most of the time, the image is high contrast and black and white. In such images, lighting is a key factor. 2D vision is suitable for places where objects have high contrast or where it is essential to determine their colour. The 2D system enables tasks to be performed in all four categories.

3D vision
3D vision is ideal for determining the volume, shape and spatial location of an object. In addition, 3D can detect low-contrast objects and defects when their heights vary.
3D imaging can be performed using both scanning and snapshot imaging.
· When scanning, the object moves through the scanner's field of view at a constant speed and creates a 3D profile from the object
· Imaging creates a 3D model of the object from the image.
Scanned images are generally significantly more accurate than regular images.
3D-vision techniques
There are different techniques that are used in machine vision. Time-based techniques employ a source of light to estimate information related to distance via a camera. Geometrical techniques include stereo vision, laser triangulation, light stripe projection and shape from shading.

Time of Flight

Time of flight sensors measure the time light takes to reach an object and return to the sensor. The change of phase of light provides enough information for calculating the time required, which is later converted into distance. This method gives the distance of each image point.

This technique can be used to measure distances up to 40 meters with a resolution of 5-10 mm (200x200 pixels). Up to 100 images per second can be recorded.

Given their relatively low resolution, the usage of time of flight cameras is rather restricted at the moment. They can be used, for example, for determining empty slots on pallets or level checks. These cameras can also be used for collecting traffic statistics.

Stereo vision

In a manner similar to the human visual system, a 3D image is created by observing an object from two different spatial positions. An evaluation of shared features found on both images is carried out and the XYZ coordinates of the feature are determined. The orientation of an object can also be determined when there are two or more shared features that can be extracted.

3D stereo vision solutions are cheap. All that is required is one 2D camera that is moved between two viewpoints.

Laser triangulation

The measurable object is probed by a laser beam that creates an exact relief of the object. The object is moved through the laser beam. Images of the laser beam are recorded by a camera placed at a specific angle. The height profiles are then merged into a single 3D image.

The result is a 3D model that can be rotated and positioned alongside each axis, meaning that there is no need to precisely position the objects to be surveyed. This eliminates the need for expensive mechanical components that sort and adjust products into the right position.

The main requisite of laser triangulation is that the measurable object moves relative to the camera and the laser beam. Certain objects pose a risk that the object shadows the laser beam from the camera and gathering information is not possible at that moment. One solution is to use several cameras that follow the laser beam from different angles. Different measurements are merged into a single set of data and shadowing occurs only when the laser is not observable by any of the cameras.

Light stripe projection

Light stripe projection is similar to laser triangulation. However, in contrast to triangulation, to measurable object must be static. The process of taking measurements, on the other hand, is very fast.

Light is projected in stripes onto the object and the camera uses the established projection to create a 3D image.

In comparison with the laser method, it is possible to estimate the height of individual pixels, hence ensuring greater precision.

Owing to its speed whilst collecting high volumes of data, this method is perfectly suited for industrial control tasks, such as shape deformations, completeness, position of components and determining volume.

Shape from shading

Shape from shading method means that three to four photos are captured of the shadow of the object from different spatial directions. The shadows allow assessing the shape as well as the surface texture of the object. Nonetheless, this method does not help to determine the exact height of an object, thus this method is mainly used for inspecting the surface quality of objects.

Shading is not affected by the qualities of the surface of the object, such as reflectivity, and high-resolution images can be taken.

The measurable object may be in linear motion along a straight line or rotate on spot.
Our contacts
Reg. code 11468255
VAT EE101260778

+372 5336 4352
Sinilille tee 6/1, office 9, 75312 Peetri, Estonia
By sending this message you agree to our Privacy Policy.
Made on