584,841 active members*
4,608 visitors online*
Register for free
Login
IndustryArena Forum > Hobby Projects > RC Robotics and Autonomous Robots > Eclipze's SMD Pick'n'Place Build....
Page 37 of 37 27353637
Results 721 to 733 of 733
  1. #721
    Join Date
    Jan 2007
    Posts
    148
    Nisma

    Hello , thanks for the reply to my post ... yes i prefer c or c++ , i have used python .
    using opencv is new to me , i'm looking to be able to pick using cannyedge and i was looking at saving as a blob and converting to real x y
    although the more i think on this , i feel i should realy just concentrate on picking the pad's and look at formating the output to find pad 1 and enumerate the part pads , but rotating the part after vision camers has taken a shot , then move to match the placement , either by another vision shot , or i feel reading the layout database is a better solution , is their any code to do the majority of this already done i wonder .

    i'm also looking at the moment at camera calibration , and camera lens distortion similar to the chessboard examples , so that i can take an undistorted view for better placement , over all i think i have the idea right , although in the back of my brain , i may be makeing this over complicated to what is actualy needed ..

    i'm looking at the code snippet you posted , to grasp your idea and get up to speed on this ,

    i would welcome any code and guidance , and welcome the oppertunity to bring this to a working solution , so spending some time reading your posts

    i am also looking at openpnp although i must admit Java is not my strong point , i much prefer c !

  2. #722
    Join Date
    Sep 2006
    Posts
    70
    First of all, Illumination is the key and preferibly you need two types of illumination, diffused and direkt lightning with at least tree power levels.
    Illumination don´t need to be perfect, a lot can be corrected with processing and it´s further possible to not correct at all some thing but then doing things different.
    In the last 10 years, there have emerged annother possibility. Use sift/surf/... or better the non patented algorithm to search for the component, the computing power and
    maybe gpu performance have allowed that and it´s maybe faster then other method with less problems with illumination.

    Yes, i have code for doing the vision processing using. For porting it to new machine, the camera and lightning must be evaluated how many noise is present, what type
    of noise and how well the illumination works. Further basically how the automatic exposition time circuit and image dsp processor modify the image.
    I do morphological operations, blur/erode/dilate/threshold/mask/.... and then blob processing, called contour on OpenCV, but third part blob libs works too.

    I can generally evaluate it really fast and this is the reason i ask for images befor giving out code.
    I use canny only as backup, usually i use laplace type, not because it works better, just because it works too and it´s times faster, and little bit different.
    Annother thing is, that different parts need different morphological code. The simplest ist his:
    I send you the code as pm, this is very old code that i wrote when testing cvBlobsLib. Never used CvBlobsLib on production.
    The test image is soic8, and one pixel is equivalent to 0.127mm. This for info. I explain closed loop centering.
    First Threshold is applied with value 210. Then Dilate + Erode both with value 4 (4*0.127*2=1,016mm) is used. This results to the displayed picture.
    Then blob detection is made. I had some filtering, but this don´t apply to this case. The floating point coord are added togedder and divided by the number of blobs.
    This is the center coordinate. The floating point difference between the center of the image and the center of all blob coordinates is the moving difference in x/y direction.
    cvRound(diff_pixel * mm_x_pixel) / mm_x_microstep ) gives the number of microsteps for offsetting the next/current x or y movement.
    For the angle, use atan2 for getting the angle. Note only that Y axis starts from top as usual with imaging algorithm and because that Y must have negativ sign in order to
    account for that. Now at example if you have sot23 or in the case you don´t have one single blob for pins but one blob every pin, then you must divide the blob coordinate
    to right/left or top/bottom and make one right and left blob center, that you can use for angle detection. I have this type of code, not on CvBlobsLib, i have used Contours from
    OpenCV. The simplest is to remove the single pin, the single blob in +x/+y direction after the center. Then it´s possible to read the angle of the sot23 chip.
    Using this you adjust angle on closed loop. Just trial and error. The calculate the center offset and add it to the next move. That´s it.
    Open loop involves more math. I don´t have made it, because i rotate the image on image center and simulate closed loop from memory image.
    Usually i have enougth time for that during move. Using this simple Algorithm, most parts can be placed. It´s possible to have errors on this algorithm. For excluding this
    infrequent errors automatically, complexer code is used. For the placement location, the pnp file has to be used .

  3. #723
    Join Date
    Jan 2007
    Posts
    148
    Nisma,
    Thank you for your comments and process , i look forward to code , i now see the need to implement lighting better than i have at the moment , so i will tackle that first .
    i was looking at using high intensity white leds mounted around the bottom camera , i had initialy thought using Infra red LED'S both top and bottom would be better due to camera wavelength etc i will introduce pwm to be able to control the brightness , i had also thought of making a matrix of white and infra red led's . at this stage i can define a pattern or size as needed , as i do not have a specific stage area to constrain the build .

    for the purpose of diffusion i will go with some white acrylic or frosted glass , and see what results i obtain , as to the best result , that will then give us some definitive pictures to works from .

  4. #724
    Join Date
    Sep 2006
    Posts
    70
    Ir works, if you have samsung type nozzle and the camera directly over the nozzle, as this special nozzles are ir trasmissive and you have realtime image during travel of the centering from the chip.
    For downlooking camera, ir don´t work because paper and plastic tapes behave different. Further it´s possible that some component are coatet, and this coating reflects the ir like if there are pins.
    Don´t use PWM for intensity, use lm317 as example, fixed resistor and two variable resistors that set the intensity driven by uln2003 as example. Use i2c expander for selecting the type of light
    as may webcams can drive external i2c bus from sw. This allow you to integrate lightning control into the camera, but if you are using i2c type drivers for C and Z axis because it have some advantages specially about expandability, then drive it with the same bus too. You only need tree level of light, direct, indirect and a mix of both.
    Annother aspect is, that if the camera don´t have enought light, it slow down the fps rate. If you want make it cheap, then it´s possible to have the combined light of both at full intensity and
    the single direct/indirect light with less fps.

  5. #725
    Join Date
    Jan 2007
    Posts
    148
    thanks for the comments ,
    i'm looking at full real time imaging during travel , so i'll take your advice thanks , i always tend to make the job more difficult than it has to be ...

    i'm looking at using a square illumination box with 16 led's per side around the bottom camera . and a ring of 16 led's for the top camera
    i dont need this to be a cheap build , so whatever is needed will be built in . i'd like to go with a cheap usb snake type camera on the z if only for size , we shall see how it performs

  6. #726
    Join Date
    Jun 2013
    Posts
    1
    Hi all, nice to see others working on there own PnP machines
    I have been working on my PnP for some time now, but i do not have many hours free so it will take some time to complete the system.

    But i have made some pre research for my laser alignment unit,
    Simple test setup with simple laser and line sensor

    Attachment 207904

    Then i get something like this in response
    Attachment 207906

    and when i rotate the component i can measure the laser light pos on the sensor and i get
    Attachment 207908

    so from that is is easy to get the component rotation and pos.
    But i have to make some more test before i can be sure how fast i could do the measurement.

  7. #727
    Join Date
    Jul 2011
    Posts
    39
    hello
    I decided to make a pick and place machine, in the wall story of the most bigger problem is software for me , I can not understand the programing so I can not to write the software , do you have to give the program free , or we can recommend a free program to download .
    Regards

  8. #728
    Join Date
    Jul 2011
    Posts
    39
    Please , can you recommend free software for DIY pick and place machine , LPT communications 4 axis ( X , Y , Z and rotation head ) max speed is 1200 pcs for one hour .

  9. #729
    Join Date
    Oct 2006
    Posts
    46

  10. #730
    Join Date
    Jul 2011
    Posts
    39
    Quote Originally Posted by atlab View Post
    I tried to find a program that is all but coda, written in Java, and anything you do not understand what needs to these codes to work.
    Do you have any program that is easier to use, and can be limited version?
    Should any program that may try to use a machine that right, if everything is ok I'll find someone who will write me a program to machine ?.

  11. #731
    Join Date
    Oct 2006
    Posts
    46
    1. All you need is to ask Google "how to run jar" and read instructions.
    2. I don't know another open souce project.
    3. If you want personally yours program, try to find programmer near your location: you and the programmer will need to do a lot of collaboration.

    Architecture of the PnP machine, control electronics and control program very much related to each other, so:
    - if you take the ready-to-use program, you must fully replicate the architecture of the machine and its control electronics;
    - If you already have a mechanical part of the machine, you will need to construct the control electronics and software for it;
    - if you use an open source project then you can adapt it to your needs (or somebody for you);
    - etc.

  12. #732
    Join Date
    Jul 2011
    Posts
    39
    I tried to run the program according to the instructions that I found on several pages on the internet but I could not, I do not know if the problem is in Windows or something else, do you have experience with running open pnp ?

    I can make a machine per program, just first have to open to see how it works and what all of the options used .

  13. #733
    Join Date
    Oct 2006
    Posts
    46
    3. If you want personally yours program, try to find programmer near your location: you and the programmer will need to do a lot of collaboration.

Page 37 of 37 27353637

Similar Threads

  1. Newbie - To build or not to build Router/Plasma Table
    By dfranks in forum Waterjet General Topics
    Replies: 10
    Last Post: 04-08-2011, 05:16 AM
  2. NEW BUILD: PVC as a build material
    By Smiler in forum DIY CNC Router Table Machines
    Replies: 12
    Last Post: 11-13-2009, 11:57 PM
  3. New Large Table Build in Houston, TX (Build Log)
    By anitel in forum Plasma, EDM / Other similar machine Project Log
    Replies: 12
    Last Post: 12-30-2008, 09:45 AM
  4. Replies: 4
    Last Post: 08-16-2005, 11:46 PM

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •