• State Resolved
  • Locked Locked
  • Replies 4 replies
  • Subscribers 101 subscribers
  • Views 1875 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

OpenCV and DSP Selection

Frank Maldonado
Prodigy 190 points
Other Parts Discussed in Thread: OMAP-L138, TMS320C6748, OMAPL138

We want to convert a MS Visual Studio C++ program with OpenCV (Version 2.4.6) to either TMS320C6748 or OMAP-L138's embedded code. The required OpenCV functions are: findContours; GaussianBlur; cvtColor; Canny; dilate; erode; contourArea; approxPolyDP; matchTemplate; cvCaptureFromCAM; imread; and VideoCapture. Some OpenCV complementary functions are: resize; getRotationMatrix2D; warpAffine; and RotatedRect.

At the moment I have the following questions:

  • Considering development time (i.e. converting OpenCV to embedded), which DSP processor would be the best choice the TMS320C6748 or the OMAP-L138?
  • In the case of the OMAP-L138 and considering the OpenCV requirement, does the ARM core introduce a bottleneck in the program execution? We are looking for real time video analysis and low power consumption.
  • Is it strictly necessary to work with Linux when using the OMAP-L138 (some E2E's forums talk about using Linux in their cases)? If true, what are the main advantages of doing this?
  • We are using a USB Web Cam with a resolution of 480x640. Access to the camera is through OpenCV. Could you recommend us a camera for working with the system prototype? Is it required to write a driver?

Thanks, 

- Frank M. 

  • 0
    TI__Mastermind 35645 points

    Hi Frank,

    Thanks for your post.

    OMAPL138 is a low power device that combines the power of fixed point ARM9 with a fixed and floating point C674x DSP that is ideally suited for several real time applications and has a wide variety of peripheral integration. A good Starting point to get information on the Software offering on the device is the OMAPL138 Software Design guide which also has links to available training and demos that exist on the device

    http://processors.wiki.ti.com/index.php/OMAP-L138_Software_Design_Guide

    If you are looking for a low cost development platform the OMAPL138 based LCDK should be a good starting point:

    http://processors.wiki.ti.com/index.php/L138/C6748_Development_Kit_(LCDK)

    there is a face detection demo on that platform that demonstrates the use of OpenCV on the C674x DSP. In addition to OpenCV, we also support a large number of DSP optimized libraries for image and vision processing:

    http://processors.wiki.ti.com/index.php?title=Software_libraries#Image_and_Video_Processing_Libraries

    We have facedetect example code that uses some opencv API, other than that we don't have a procedure or examples of opencv.

    http://processors.wiki.ti.com/index.php/BIOS_C6SDK_2.0_User_Guide#Face_Detect_Demo

    Install C6SDK package to know about facedetect example code.

    c6sdk_02_00_00_00/demos/facedetect/

    The cxcore library of OpenCV is integrated for identifying the face. The demo integrates prebuild OpenCV functions for facedetect. We don't have sufficient information on opencv 3rd parties and it is limited to our scope.

    OpenCV libraries are located at c6sdk_02_00_00_00\demos\facedetect\bin. The following functions are used to detect the face from Open CV algorithms

    Apply OpenCV haardetect algorithm

    dsp_sequence = cvHaarDetectObjects(image, cascade, storage, 3, 2, 0, cvSize(30,30));

    Thanks & regards,

    Sivaraj K

    ----------------------------------------------------------------------------------------------------------------

    Please click the "Verify Answer" button on this post if it answers your question.

    ----------------------------------------------------------------------------------------------------------------

  • 0 in reply to Sivaraj Kuppuraj
    Prodigy 190 points

    Hi Sivaraj Kuppuraj, 

       I think we will get the OMAPL138 based LCDK and to try to benchmark some of the algorithms. At the same time we will evaluate the OMAPL138's OpenCV features.

       I wonder, could you recommend us a cam that can be used with the LCDK? we are using a USB Web Cam with a resolution of 480x640, but it can change it according to your suggestions. We would like to avoid writing drivers and then a plug-and-play solution would be the ideal choice. 

       Is there needed any additional hardware tool (such as a debug probe) for working with the OMAPL138 LCDK?

       Thanks & regards,

       - Frank M.

  • 0 in reply to Frank Maldonado
    TI__Mastermind 35645 points
    Hi Frank,

    Sounds good. Leopard imaging sensor camera will hold good to use with LCDK and JTAG XDS510 USB emulator, you could use this debug tool to work with OMAPL138 code. KIndly check the below wiki:

    processors.wiki.ti.com/.../AM1808_EVM_board_using_CCS%3F

    Thanks & regards,
    Sivaraj K
  • 0 in reply to Sivaraj Kuppuraj
    Prodigy 190 points

    Hi Sivaraj,

       Looking at the "XDS510 USB emulator" I found several models such as:

    • XDS510 USB JTAG Emulator 701900
    • XDS510 USB Galvanic JTAG Emulator 701910
    • XDS510 USB PLUS JTAG Emulator 701907

    the three from Spectrum Digital Inc. and being their cost $ 989.00, $1,995, and $1,199.00 respectively.

       I have the following three questions:

    1. Why are they so expensive? Do they have some kind of advanced features?
    2. Could we use the low cost "USB2.0 XDS510 TI DSP JTAG Emulator Programmer" ($95.00) with the OMAPL138?
    3. Is it needed an external emulator for working with the TMDXLCDK6748 development kit? We are trying to avoid using OpenCV and then we could target the TMS320C6748 instead of the OMAPL138. 

       Thanks & regards, 

       - Frank M.