With the improvement of social consumption and scientific level, industrial 4.0 is the short-term goal of global industry, at least the traditional industrial production will enter into automatic production. The essence of industrial automation is to use robots instead of manual labor. Robots also need "eyes", so it is convenient to act as the eyes of robots and think. Machine vision is the application of computer vision in the field of industrial automation.
Machine vision includes two-dimensional detection and three-dimensional detection, and is in the stage of over-detection from two-dimensional detection to three-dimensional detection. The advantages and disadvantages of the two are always controversial topics of machine vision. Three-dimensional vision inspection companies tend to emphasize the advantages of three-dimensional vision detection, and two-dimensional vision inspection companies will also endorse the technology they have mastered. In fact, two-dimensional detection and three-dimensional detection have never been on the opposite side of the two detection technologies. Two-dimensional detection is a mature technology in these two technologies. It has been proved to be very effective in a wide range of automation and product quality control processes in the past 30 years, but two-dimensional detection is based on the plane dimension, compared with three-dimensional detection. In stereo dimension, 2D vision detection can not get the height information of the object to be measured.
1、limitations of two-dimensional detection
A. cannot measure shapes.
2D sensors do not support shape dependent measurements. For example, a 2D sensor cannot measure features such as object flatness, surface angle, partial volume, or distinguish objects of the same color, or distinguish between objects with contact sides.
B. is susceptible to variable lighting conditions.
Because illumination determines the position and measurement accuracy of edges, changes in illumination within the sensor's field of view sometimes lead to edge measurement errors unless specific techniques are used to compensate for such effects. Under factory conditions, lighting change is a common challenge for production workshop accident environment or field source.
C. limited contrast compensation
2D sensors rely on measuring the contrast (edge data) of objects, which means, for example, that they cannot measure black objects on a black background or distinguish some features to expose the existence and definition of edges without specific illumination.
D. is sensitive to motion of objects.
Because of errors caused by component movement, two-dimensional sensors need to be precisely fixed along the optical axis (Z axis), using scale invariant feature detection, or using large telecentric optical elements (which must match the size of FOV) to eliminate the impact.
Compared with the original 2D detection, 3D detection not only obtains the x, Y axis information and gray information, but also obtains the Z axis information of the product, which increases the detection and processing of the product height information. 3D detection is widely used in product flatness and surface height detection.
2、advantages of 3D detection
A. volume measurement (X, Y and Z direction) provides shape and position dependent parameters.
B. is suitable for detecting objects of the same color or low contrast.
C. is not affected by lighting changes and ambient light.
D. can achieve higher repeatability due to integrated optics, lighting and factory pre calibration.
E. is easy to form multi-sensor networks for scanning large workpieces.