Computer Vision-Based 3-D Reconstruction for Object Replication


Ryan Cullinane
Cady Motyka
Elie Rosen

Prof. Kristin Dana

• Design software utilizing the Microsoft Kinect to generate a 3-D mesh of a chosen object and convert the data into a printable 3-D model.


• Utilize the Kinect’s infrared laser pattern projecter and IR Depth Sensor to gather information about depth from the scene.

• Convert depth data into a 3-D mesh where the vertices of the mesh are put together to create an object file that allows for a 3-D model to be viewed on a computer.

3-D Modeling

• Calibration of the printer is done manually by adjusting the endstop controllers for the motors and the bed height. The bed height calibration involves adjusting the height of the bed so that the extruder nozzle is approximately 1mm away from the glass pane at all times.

3-D Printing

• Printing is executed layer by layer, where the object is built vertically from the bottom up. Supports generated from the model conversion are printed thinner so they can be

• Normal vectors for the triangles are then computed to determine which direction the triangle is facing. These triangles then form a 3-D Model Group.

• Use the 3-D model, generate G-Code which can be sent to a 3-D printer to create a physical replica of the object.