The ICP algorithm (ICP = Iterative Closest Point) is able to accurately calculate the rotation in 3 spatial directions in order to transform from one 3D pose to another 3D pose of the same object. This technique can be used, for example, to compare the 3D scan of a part with a template to determine the 3D orientation of the corresponding object in 3D space. For example, a 3D image of a perfect sample part (golden template) can be compared with the 3D images of inspection parts on a production line. Dimensional deviations between the two images can thus be detected and marked in real-time, allowing quick pass/fail decisions to be made. However, the algorithm requires considerable computing time and struggles with tilt angles greater than approx. 15°. This is where Match 3D DNC delivers the perfect complement.
CVB Match 3D
Software for demanding 3D object and pose detection
In many production environments it is important that each product is checked for completeness and flaws. In 2D it is a common approach to compare a golden template image with an image of the test object. In theory this is a simple process of subtracting the test image from the original; in practical applications this is sometimes impossible because the test image contains alignment stretch and rotation errors.
In 3D the principle is the same, but the alignment of both images is more complex with additional degrees of movement and rotation possible. CVB Match 3D offers a new practical approach, providing the highest accuracy and a processing speed fast enough to keep track with modern production lines.
CVB Match 3D can be easily integrated into existing machine vision environments. As the alignment is done in software, there is no need for accurate part positioning and handling of the test sample. The mechanical effort is reduced dramatically and high inspection throughput is guaranteed.
CVB Match 3D includes two powerful algorithms that complement each other perfectly - Match 3D ICP and Match 3D DNC.
The above screen sequence shows a test object. Figure 1 shows the range map of a perfect part used as a pattern. Figure 2 shows the actual test object.
In figure 3 the two pictures are combined and aligned in the same room. The check can be carried out. Figure 4 shows highlighted areas wherever differences in elevation information have been detected.
This function allows the user to locate 3D objects in a point cloud and determine their position. A CAD file of the object to be located is used as a template. A special feature is that DNC does not require any information about the sensor used, such as spatial resolution or intrinsic calibration data. Neither the colour nor the surface texture of the object is important, hence the name DNC, which stands for "Depth - No - Color": Only depth data is analysed and used to detect objects and determine their position.
With the right choice of parameters, an object can be detected in a fraction of a second. With a suitable recording technique and a frame rate and resolution that are not too high, it is possible to search and measure the pose almost in real-time. The accuracy of the pose measurement is good enough for many practical applications. If higher precision is required, Match 3D DNC can be used for pre-positioning, followed by Match 3D ICP.
Specifications
- Software for demanding real-time 3D object and pose detection
- Highest accuracy
- High processing speed
- Algorithm works on real 3D point clouds and automatically adjusts position errors or tipping and tilts in all 3 axes (6 variants)
- No need for accurate positioning of test parts
- Allows the inspection of different parts at the same time
- Easy integration into existing machine vision environments
- Easy-to-use programming interface
- Comprehensive documentation and examples
Software for 3D surface inspection
- Allows a quick comparison of model and sample
- Best choice for complex objects
- Corrects positioning ...
Software for 3D surface inspection
- Allows a quick comparison of model and sample
- Best choice for complex objects
- Corrects positioning ...