3D vision cameras reconstruct objects from a three-dimensional image, just like our brain does. They are increasingly used in industrial applications as they provide multiple functionalities. In this blog we explain the different types of 3D cameras that exist according to their mode of operation and the different industrial applications that we can carry out with them.

3D vision versus other types of vision

The field of computer vision has evolved a lot in recent years. Technical progress has made it possible to solve some of the existing problems in the industry with this technology. In addition to traditional 2D vision, the use of many other sensors that work with images of different types has spread: thermographic images, hyperspectral images, or 3D images.

All of them are used in applications that cannot be solved with traditional 2D vision, or that combine several techniques to do so.

3D cameras, for example, are used in applications such as: detecting possible defects such as excesses or shortages of material or metrology in cases where 3D magnitudes of the object have to be measured. It is also used in applications for robot guidance and part location. Many times, it is necessary to combine several technologies or types of images to solve the same application, so it is interesting to know the different options that exist.

There is a very wide range of possibilities and today we focus on explaining the different types of 3D cameras that exist.

Types of 3D vision cameras

  • Laser triangulation (Laser Triangulation): This type of 3D vision cameras use a laser that projects a line and a camera that acts as a receiver. The laser hits the object that modifies the projected line and the reflected light hits the receiver at different angles, depending on the distance at which it is. In this way, a point cloud is obtained that represents the object. Keep in mind that to work with this technique the object must be in motion. A linear movement must be made so that the laser line can scan the entire surface of the laser.

    These equipments are quite inexpensive compared to other 3D sensors and have a lot of precision, although, in this case, some properties of the object, such as brightness, can affect the reconstruction.

  • Stereoscopic images (Stereo Vision): the stereoscopic vision is obtained by means of two calibrated 2D cameras. In this case, two images of the same scene are captured with two cameras from two different points. The process is similar to that carried out by human eyes. With the images obtained from two points with the calibrated cameras, the 3D image is reconstructed. The reconstruction is carried out based on the epipolar geometry, with which the correlation between the points represented in the two images can be established. Pixel matching is done to reconstruct the 3D. The areas with greater texture, that is, with greater color changes, are the easiest to reconstruct since the point pairing between the two images is easier. The less textured areas are the most problematic.

    The pixels are then paired between the two sets of images to produce 3D depth. Match points between images are based on variation in texture to find distinctive edges and features, which can cause problems when surfaces in a scene are insufficiently contrasting or are too similar.

  • Structured Light: cameras use a similar principle to laser triangulation. A pattern of light is projected that provides texture to the surface of an object allowing for reconstruction.

    In this case the object must not move, but must remain static. The light patterns emitted by lasers are usually binary. They are very precise sensors and very large objects can be reconstructed, the main problem they present is that they are very sensitive to changes in light and it is not possible to use them with transparent materials.

  • Time of Flight (ToF): cameras based on the principle of time of flight emit a laser pulse and calculate the time it takes for this pulse to travel to the object and return to the image sensor The distance is calculated in as a function of the delay between the emission and the reading of the return pulse. To do this, it is necessary to know the speed at which the pulse is emitted. These cameras usually work in the near infrared wavelength.

    The reconstruction is very fast, but this technique also has problems with changes in lighting and transparent objects. The resolution usually obtained with cameras of this type is +/- 1cm. For example, these cameras are the ones used in many video games like Kinect or Xbox.

Industrial applications of 3D cameras

3D vision opens the doors to solve problems with machine vision that cannot be addressed solely using 2D vision. Next, we tell you some of the industrial applications in which 3D vision is used.

  • 3D vision for defect inspection: 3D vision is used in computer vision systems for quality processes in which surface and geometric defects are identified and located, etc. These inspections are performed, usually visually, and are expensive processes that must be performed by highly trained personnel. Automatic inspection systems can reduce inspection time and increase the reliability of results and their quality, thanks to 3D vision it is possible to see the depth and severity of the defect.

    An example is the welding processes in which possible errors lead not only to an aesthetic defect, but also affect its functionality.

  • Dimensional control and metrology: dimensional control allows the physical magnitudes of an object to be measured and they serve as a quality parameter for both the process and the product.

    The data obtained with this dimensional control, in addition to guaranteeing the quality of the product, are a source of information for the improvement of the process.

    Dimensional control systems include 3D sensors and are applicable in many sectors and applications.

  • 3D vision and robots: another of the most common applications of 3D cameras is the union with robotics. Using the 3D camera, objects can be located and positioned in order to later pick them. They are used, for example, in bin picking systems. Robotics guided by computer vision becomes a very versatile tool that can be applied in different processes.

¿Necesitas ayuda?

¡Ponte en contacto con nosotros!