3D Scanning with Kinect v2

Posts made in July, 2015

Initial Angle Detection

Initial Angle Detection

By on Jul 10, 2015 in Uncategorized | 2 comments

Working on the Angle Detection You can see here on the left is a depth-change view of the room; on the right is the new angle detection. Areas that are the same angle will be the same color; smoothing has been applied. Some glitches There was a problem where the x and y axis were influencing the angle calculation- you could see this by the line down the middle horizontally and vertically. There was a problem where only half of the possible color ranges were being used. Both of these problems have been fixed. Added Z Axis We’ve added the Z-Axis, it’s colored blue. It represents how far away from the xy axis the steepest angle is. Added Selection We’ve added the ability to select an area. You will need to select an area that seems consistent in color before specifying which surface that area belongs to. What’s lacking Now we need to test to see if correct orientation can be applied using the angles detected in the...

Read More
Optimal Smoothing for Orientation Calculation

Optimal Smoothing for Orientation Calculation

By on Jul 10, 2015 in Uncategorized | 0 comments

What is needed to calculate the camera orientation The camera orientation involves the camera’s xyz angles and the camera’s xyz position. To get the angles of the camera, we have identify the angle of 2 surfaces that are at a right angle to each other. Assuming it’s a square room, some examples of right-angle combos would be floor and back wall left side wall and ceiling backwall and left side wall right side wall and front wall Bad accuracy of angles Even after running a complex averaging algorithm of multiple frames, the surfaces will still have an effect like crumpled paper. This crumpled effect is because the averaging algorithm works on individual pixels and does not attempt to smooth; several dozens of frames would be needed to get a smooth surface by just averaging- and that’s assuming the camera is in a fixed spot and not being held by a person. If you selected a pixel on the above wall to use as an angle, you would have an error range of about 15°. Such a high error range would give you the correct orientation about 1 out of 30 tries. Smoothing By using the average position between adjacent pixels, a new position can be created which will smoothed out. Each time this smoothing is run, the new smoothed out frame will lose a bit of width and height. Using this approach of smoothing has the advantage that closer pixels will influence the position more than pixels further away. It also has the disadvantage that each additional smoothing has a lower impact compared to the smoothing that was done before. Here is the above wall after it has been smoothed 50 times. Here it is after smoothing 100 times. There is a larger difference between 0x smoothing to 50x smoothing compared to 50x smoothing to 100x smoothing. Smoothing Comparison Here each picture has run smoothing 10 more times than the prior image. Optimal Smoothing Approach So far we’ve figured 2 ways to get the correct angle. run smoothing about 50 times and have user select 1 spot to use as the angle run smoothing 10 times but have the user select more than 1 spot on the same surface...

Read More
Comparisons coming soon!

Comparisons coming soon!

By on Jul 7, 2015 in Uncategorized | 0 comments

Checking Quality We’ve completed a full 3d scan of a person and are now comparing between what we sent to be printed and what we actually got back. Multiple Printers We are actually getting this 3d model printed at multiple locations so we can check if quality is different across different locations. We will also be trying different materials so we can tell what to expect for our clients in future prints. What Next? Once we get all the full body printouts back, we’ll put up a video and show our findings. Depending on the quality of 3d prints we work on increasing the quality that the program creates or we’ll focus on simplifying the current process. It currently looks like the program is putting out higher quality than can be printed, if that’s the case then we should be out of beta soon and can start focusing on getting this into the hands of...

Read More
Note: Scan from Life is in no way affiliated with Microsoft. We are an entirely separate company that has created a product that is dependent on a Microsoft owned product.