: Easy 3D Calibration of laser triangulation systems Fredrik Nilsson Product Manager, SICK, BU Vision
Using 3D for Machine Vision solutions : 3D imaging is becoming more important and well accepted for Machine Vision solutions : Gives information about Shape, Size, Volume, and Position in space, which is vital in e.g.: - Robot guidance and bin picking (3D position and pose) - Food portioning (volume) - Log scanning (shape and volume) - Electronic component inspection (size and position) : 3D also adds value in what could be 2D solution, e.g.: - Code verification on tires (difficult contrast) - Width measurements (independent of height position) Fredrik Nilsson, Product Management 2
Calibrated 3D for real measurements : Not all applications requires exact measurements, e.g.: - Reading the relief code on tires - Part identification (from 3D features) - Assembly verification - Detect broken parts : To know how large, how much, the size of faults, or where, then calibrated data in e.g. millimeters is essential - Food portioning (same volume in each slice) - Dough volume (volume within limits) - Extrusion control (size within limits) - Random bin-picking (where and how) - Gap control (distance) - Robot guidance (where) - Surface coplanarity Fredrik Nilsson, Product Management 3
Optimized solutions for reliable results : 3D devices that comes with inbuilt laser and fixed geometry can be factory calibrated, i.e. instantly provide 3D data in e.g. millimeters + Easy to integrate - Inflexible if optimization of data quality is required (fixed field-of-view, resolution, occlusion, etc.) : 3D Cameras with free choice of lens, laser, base line, laser/camera angles, gives full flexibility to optimize data quality for each specific task: + Free choice of field-of-view + Tune height resolution vs. depth-of-field (height range) + Reduce occlusions and secondary reflexes + Save machine space by separating laser and camera - Requires final calibration procedure to get data in millimeters : Flexibility allows for optimization of data quality to build reliable 3D solutions Fredrik Nilsson, Product Management 4
What is 3D calibration? : Calibration is to find the correspondence between sensor pixels and the world coordinates in the laser plane. : Sensor, lens, laser, and angles between them, all affects this relationship. : The calibration need to compensate for the lens distortion, the perspective view, and the laser source, to get accurate results in millimeters Changing the lens, laser-, or camera-angle changes the scene and the system then needs to be re-calibrated Fredrik Nilsson, Product Management 5
Challenges of 3D calibration : Perceived as a challenging task and is indeed a non-trivial task - Examples of integrators who have spent months of valuable time and effort : Methods exists but with limitations in practice. Examples: - Checker board requires external light source, do not compensate for laser properties, less suitable for large FOV - Accurate positioning in height requires time to perform or expensive equipment (and space) - Multi-facetted objects requires stable and controlled movement, challenges for large FOV : Integrators often finds themselves in situations where they have: - Limited time - Limited budget - Limited machine space Prefer to re-use previous work to lower risk and save time (i.e. method must be scalabe and suitable for various scenarios) : Most methods are not appropriate for integrators when needing in-machine calibration Fredrik Nilsson, Product Management 6
SICK Introduce: Easy 3D Calibration : A complete concept to help integrators pass the challenge of 3D Calibration : Allow integrators to use the full flexibility of Ranger to optimize data quality and reach reliable solutions and still get calibrated 3D data (as of factory calibrated devices) : Vastly reduced effort and development time to get calibrated 3D data into Ranger applications : Consist of several components: - A new method that do not suffer from the known limitations of existing methods - GUI that guides the operator through the complete calibration - Calibration targets in different sizes as accessories and as CAD - GUI to visualize calibrated 3D data with measurements - API to incorporate calibrated data into Vision Applications Fredrik Nilsson, Product Management 7
The new 3D Calibration method : Calibration method based on showing a known calibration target at several random positions in the laser plane : Two phases of the calibration - Showing a flat surface (for lens distortion) - Showing a saw-tooth shaped target (perspective and scale) : SICK s method has several advantages: - No need to know the positions of the target no need for expensive equipment for accurate positioning - Does not rely on stable translation movement vibrations in conveyor systems will not influence calibration result - Does not require much space to perform suitable for in-machine calibration - Scalable method suitable from ~150mm up to 1,5m (different targets required) - Uncomplicated calibration target easy and cheap to manufacture - Does not require additional light sources - Relatively fast calibration completed in minutes Fredrik Nilsson, Product Management 8
The Coordinator (GUI Guidance tool) : Assists the user through the complete calibration procedure step by step : Indicates were more of the calibration target needs to be shown : Immediate feedback about calibration errors : Iterative until results are satisfactory : Uploads calibration result to the Ranger (flash memory) Fredrik Nilsson, Product Management 9
The Calibration target : The calibration target has a un-complex shape - easy to adapt and duplicate : Different sides with different purposes: - Saw tooth to get scale and perspective (known distances between feature points) - Flat side for compensation or lens distortion : Available as accessories of different sizes : Select target size to fit the field of view : CAD drawings for own manufacturing Fredrik Nilsson, Product Management 10
Complete 3D Calibration in practice : Movie that shows the calibration procedure Fredrik Nilsson, Product Management 11
Result of Calibration : Complete in-machine calibration accomplished in a few minutes : Calibration errors less than 0.05% of field-of-view : Compensated for lens distortion and perspective view : Calibration data stored in camera flash (ready at power up) : Ranger ready to provide 3D data in millimeters rather than pixels via API : Data in global coordinate system in laser plane [X,Z] (Y from encoder) : icon API for data as point clouds or in rectified format = Quality optimized AND calibrated 3D data directly into the Vision Application Fredrik Nilsson, Product Management 12
Benefits of calibration concept : Fast - Calibration completed in minutes : Easy hand held calibration - No need for positioning equipment or controlled conveyor movement : Uses 3D laser only - No need for additional light sources : Flexible - Suitable for inspection widths from 150 to 1500mm : Accurate - Calibration errors as low as 0.05% of FOV is possible : User friendly - Guide through all steps with immediate accuracy feedback : Minimize errors - Calibration is using the same measurement principles and setup as the final machine vision inspection : Inexpensive - No additional equipment than the un-complex target. Software free of charge to all Ranger customers Fredrik Nilsson, Product Management 13
Summary and conclusions The Easy 3D Calibration concept will : Significantly reduce the integrators efforts, allowing them to save valuable time and cost : Allow for more robust solutions with the flexibility of free camera-lens-laser arrangements, still getting calibrated data : Allow for easier retro-fit and save of machine space by separating laser and camera : Provide calibrated 3D data directly into integrators application in the most appropriate format for the task (point clouds for 3D-matching, rectified images for analysis) : Make 3D more available by lowering the step from 2D solutions Fredrik Nilsson, Product Management 14
Welcome to learn more in our booth: Hall 4.0, Booth 4B71 Or read more at: www.sick.com/rangercalibration Thank you for your attention! Fredrik Nilsson, Product Management 15