[Machine Vision: Fundamentals of Dimensional Measurement] In traditional automated production dimensional measurement, a typical method involves using calipers or micrometers to take multiple measurements of a certain parameter on the workpiece and then averaging the results. These inspection devices or methods suffer from low measurement accuracy, slow measurement speed, and the inability to process measurement data in a timely manner, failing to meet the needs of large-scale automated production.
Machine vision-based dimensional measurement methods offer advantages such as low cost, high accuracy, and easy installation. Their non-contact, real-time, flexible, and precise characteristics effectively address the problems of traditional inspection methods. Dimensional measurement is the most widespread application of machine vision technology, particularly in automated manufacturing. Machine vision is used to measure various dimensional parameters of workpieces, such as length, circle, angle, arc, and area measurements, requiring the detection of basic geometric features in relevant areas. It not only acquires dimensional parameters of products online but also enables real-time online judgment and sorting, making its application extremely common.
Basic flowchart of workpiece inspection
The measurement of the dimensions of the object being measured usually includes multiple parameter dimensions, such as distance measurement, circle measurement, angle measurement, line arc measurement, area measurement, etc.
Machine vision dimension measurement application examples
Angle measurement
In traditional automated production dimensional measurement, a typical method involves using calipers or micrometers to take multiple measurements of a certain parameter on the workpiece and then averaging the results. These inspection devices or methods suffer from low measurement accuracy, slow measurement speed, and the inability to process measurement data in a timely manner, thus failing to meet the needs of large-scale automated production.
Machine vision-based dimensional measurement methods offer advantages such as low cost, high accuracy, and easy installation. Their non-contact, real-time, flexible, and precise characteristics effectively address the problems of traditional inspection methods. Dimensional measurement is the most widespread application of machine vision technology, particularly in automated manufacturing. Machine vision is used to measure various dimensional parameters of workpieces, such as length, circle, angle, arc, and area measurements, requiring the detection of basic geometric features in relevant areas. It not only acquires dimensional parameters of products online but also enables real-time online judgment and sorting, making its application extremely common.
Basic flowchart of workpiece inspection
The measurement of the dimensions of the object being measured usually includes multiple parameter dimensions, such as distance measurement, circle measurement, angle measurement, line arc measurement, area measurement, etc.
Machine vision dimension measurement application examples
Angle measurement
Image sensors can represent the object being inspected on a plane and calculate its position, width, angle, etc., through edge detection. An edge refers to the boundary between bright and dark areas within an image.
How to perform edge detection
(1) Projection processing
Projection processing involves performing a vertical scan relative to the inspection direction, and then calculating the average concentration of each projection line.
The average concentration waveform of the projected lines is called the projected waveform.
Calculating the average concentration along the projection direction can reduce inspection errors caused by noise within the region.
(2) Differential processing
Differential processing is performed based on the projected waveform. Areas that may be edges or have large concentration variations will have larger differential values. This can eliminate the influence caused by changes in the absolute concentration within the area. For example, the differential value for areas with no concentration variation is 0, and the value for white (255) → black (0) is -255.
(3) The maximum value of the differential is made 100% through correction.
In actual production lines, to achieve a stable edge sensitivity, appropriate adjustments are typically made to ensure the absolute value of the derivative reaches 100%. The peak value of the derivative waveform exceeding a pre-set "edge sensitivity (%)" is taken as the edge location. Based on the detection principle of peak values for intensity variations, edges can be reliably detected even on production lines where illuminance frequently changes.
(4) Subpixel processing
For the three pixels near the center of the largest portion of the differential waveform, a correction calculation is performed based on the waveform formed by these three pixels. The boundary position is calculated in units of 1/100 of a pixel (sub-pixel processing).
Representative detection applications of edge detection
(1) Utilizing various inspections at the edge position
Set edge position modes at multiple locations to measure the X or Y coordinates of the object being detected.
(2) Various checks using edge width
The “external dimension” mode of edge width is used to detect the width of the metal plate, the X/Y direction diameter of the hole, etc.
(3) Utilize various inspection methods in the circumferential area at the edge position
Using the circumference as the detection area, the angle (phase) of the cut-out part is detected.
(4) Various checks using the width of the trend edge
Using the "trend edge width" mode of the "circumference" area, scan the inner diameter of the annular workpiece and evaluate its flatness.
Trend Edge Pattern
Trend edge position (width) mode refers to detecting edge position while scanning a narrow edge window within the inspection area. Using this inspection mode, edge position (width) checks can be performed on multiple points within a window, thus ensuring the capture of minute changes in the workpiece.
Detection principle: Move the segment within a small area at small intervals to check the edge width or edge position of each point.
Methods to improve position detection accuracy: Reduce segmentation size.
Methods to shorten processing time: Reduce the segmentation shift amplitude (movement amount).
Trend direction: The direction of the segment movement.
Preprocessing filters to improve edge inspection results
The key to edge inspection lies in minimizing edge unevenness. Pre-processing filters act as a "median" or "average" filter, thus helping to maintain consistent inspection results. The characteristics and selection methods of pre-processing filters are described below.
Original image
Averaging
Medianization
Key points for using edge inspection mode on image sensors
Make effective adjustments based on an understanding of the principles of edge checking;
Understanding various derivative patterns significantly increases the likelihood of detection.
Referring to representative inspection cases is helpful for the work;
By selecting the optimal pretreatment filter through experiments, the inspection speed and inspection results can be improved.
