3D to 2D Template Matching

Project Background

Measurement Background

To ensure battery quality, safety, and optimize performance, manufacturers need to strictly control every link in the production process, including battery dimensional accuracy. The 3D inspection of battery height and width aims to achieve precise measurement of battery physical dimensions through a combination of 2D and 3D inspection tools.

Camera Selection

Hepersen LCF

Measurement Items

Battery length, width, and height

Solution

AI-Vision first converts the point cloud image into a depth map, creates a template and performs template matching on the depth map to confirm the sample position, then uses 3D tools to measure the sample's width and height.

Design Concept

Implementation Effect Display

• Project Result Display
  • Width Measurement Result
  • Height Measurement Result

Project Process

1. Initialization

1. Use the Lua Script tool to initialize global variables.
2. Use the Load Point Cloud tool to load the point cloud image to be processed.

-- Initialize global variables for loop judgment
SetIntVariable("count",1)

2. Obtain Sample Position and Angle Information

1. Use the Point Cloud to Depth Map tool to convert the 3D point cloud image into a depth map for 2D template matching.
2. Use the Load Image tool to load the saved depth map.
3. Use the 2D Template Creation tool to create a 2D template.
4. Use the 2D Template Matching tool to perform template matching on the depth map based on the created template.
5. Use the Lua Script tool to obtain all sample position and angle information. Output data to global variables.

• Bound variables: Template matching results; resolution, origin position (output variables of Point Cloud to Depth Map)
• Edit the script to obtain all sample matching angles and insert them into an array; obtain position information from the template matching results, and convert 2D matching XY positions to 3D XY coordinates according to 2D matching position * resolution + origin position
• Output data to global variables

-- Obtain matching angles
angles = {}

for i = 1,5,1 do
   table.insert(angles,Result[i].Angle)
end

-- Obtain actual point cloud image position: 2D matching position * resolution + origin position
positionX = {}
positionY = {}

for i = 1,5,1 do
    table.insert(positionX,Result[i].Center.X*resolutionx + offsetx)
    table.insert(positionY,Result[i].Center.Y*resolutiony + offsety)
end

for i = 1,5,1 do
   DrawPoint2D(0, "blue", positionX[i], positionY[i])
end

SetFloatArrayVariable("angles",angles)
SetFloatArrayVariable("positionX",positionX)
SetFloatArrayVariable("positionY",positionY)
SetIntVariable("counts",5)

3. Cyclically Measure Width and Height

Adjust the orientation of the point cloud based on the obtained template matching angle and position information to make the left and right edges of the measured block parallel to the Y-axis, and the upper and lower edges parallel to the X-axis. Then push the box to measure the width and height of the block.

1. Use the Marker tool to mark the start of the loop.

2. Obtain the current template matching position and angle through a script: The script obtains the current position and angle based on the count value and outputs them to global variables.

3. Use the 3D Position Adjustment tool to adjust and locate to the XY position of the current sample to be measured.

   

4. Use the 3D Transformation tool to rotate the point cloud according to the template matching angle of the sample to be measured.

   

5. Copy the point cloud to IM4 for width and height measurement.

6. Use four Square Probe tools to find the four edges of the sample.

7. Use four 3D Geometric Intersection tools, bind the two edges found in the previous step as input geometries, and find the intersection points of the four edges respectively.

8. Use four 3D Distance tools, bind two of the four intersection points as input points respectively, and measure the length and width of the sample.

9. Use two 3D Region tools to set the datum region.

10. Use the 3D Plane tool, bind the region set in the previous step as the area, and fit the datum plane.

11. Use the 3D Height tool to measure the distance from the upper layer of the sample to the bottom surface and the distance from the upper layer of the sample to the lower layer of the sample respectively.

12. Use the Lua Script tool to cyclically increment the global variable value by 1.

• Bind the current loop value count and the total number of samples counts
• Edit the script to judge whether the loop ends

if count < counts then
    Jump("ReStart")
    SetIntVariable("count",count+1)
else
    Jump("End")
end

13. Mark the end of the loop: Use the Marker tool to create a loop stop marker.

Experience Summary

Core Steps:

1. Convert the 3D point cloud image into a depth map and perform 2D template matching to obtain the relative template angle of the sample and the XY position in the 2D image.
2. Convert the angle and position information obtained from 2D to 3D information.
3. Perform measurements on the 3D point cloud image based on the obtained information.