Hướng dẫn xuất file từ hệ thông people counting system

Can you please be more specific about your people counting application and what demos do not work? The standard 3D people counting demo keeps track of how many new targets enter the scene. Can you use this Target ID value to track the number of people entering?

Regards,

Jackson

• Hey Jackson, Thanks for your quick response! Great this is exactly the data we need! We simply need to count how many people have intered the building and give this data to some actors. Where exactly can we find the output of this data to the CAN port. We couldnt quite comprehend where this happens in the code. Regards, Sven
• Hi Sven, The Target ID will be in the UART data output from the sensor every frame. The UART data structure is detailed in the Data Format section of the user guide linked above. Please see the example python code located in the industrial toolbox at "C:\ti\mmwave_industrial_toolbox_4_5_1\labs\people_counting\visualizer". The file 'oob_parser.py' gives an example of reading this uart stream. Regards, Jackson
• Hey Jackson, sorry for the late response. We didn't get earlier into the college. Corona... We wanted to try your solution today but we have some problems with the flashing of the image. We already made another thread int he forum, because it is another problem :D but maybe you can also help us with that... Below i have posted the link to our problem:
  hopefully we will find a fast solution Regards, Sven
• Hi Sven, I understand about the timing, no problem. I have replied to the other post as well. Please reply here when you can test the target ID. Regards, Jackson
• Hey Jackson, so we was able to flash the board. We went through the User's Guide. In "4. Run the Lab" we followed the instructions in the visualizer folder. We were able to install the mmWave_People_Counting_GUI. We could open and configure the Visualizer. It also said connected to the Port Com. But when we launched the Visualizer by clicking Send Configuration, we did not get any Frames. In the developers Guide it says Python 3.6 64 Bit is requiered. So we also downloaded it installed the librarieres by using the setupEnvironment.bat. So we thought we were ready to go after all this. But still we do not get any frames in the visualizer. We also looked up if everything was downloaded and installed properly. But everthing was on place. Also the command "Python gui_main.py" did not word. (There is a picture below of the error message) Regards, Sven
• Hi Sven, The visualizer requires a specific cfg file for each binary file loaded on the sensor. Can you confirm which .bin file you loaded and which .cfg file you selected? They both need to be for the 3D people counting. The other issue can be using an older version of the toolbox, but I see you are using 4.5.1. You shouldn't need a python installation to run the .exe, it should install with all dependencies. But running the gui from python source code may help to find the issue. From your cmd window error, it seems the numpy package might need to be updated. Can you try the following command and see if it changes? 'pip install numpy -U' Regards, Jackson
• Hey Jackson, I am not sure how i can see which cfg file i loaded on the sensor. Do you mean the image i flashed on the sensor? We used the "3D_people_count_68xx_demo.bin" for the sensor. Your command worked! Now i can open it without the error! Nice! But i also get any frames there ... Regards, Sven
• Great! You must first use the 'select configuration' button before sending it to the device. Which .cfg file are you selecting when pressing this button? If you can post the cmd window output when running from python source that might help. Regards, Jackson
• I use the "ISK_6m_default.cfg" from the 3D_people_Counting folder
• Hi Sven, It seems that the device is not responding when you send the cfg file. Can you confirm that the ISK board has been put back into operational mode after flashing it? And has it been reset? And the COM port numbers are still correct? If so please try re-flashing the device as something may have gone wrong. Regards, Jackson
• Hey Jackson, Flashing does not work again. It is really frustrating... I went through the Setup Guiede now 3 times. Set all jumpers, powered first the EVM and opened then Unifalsh, pressed the NREST Button right before flashing, picked the right COM. I really don't know why it is not working... I also rebooted my PC 3 times. It always shows the error: [18.1.2021 17:08:28] [ERROR] Cortex_R4_0: Initial response from the device was not received. Please power cycle device before re-flashing. [18.1.2021 17:08:28] [ERROR] Cortex_R4_0: Not able to connect to serial port. Recheck COM port selected and/or permissions. Hope we can find a solution for this ... Regards, Sven
• Hey Jackson, So this is weird. I tried for fun to flash the device and now it worked. I did not change anything. MAGIC! :D But it did not help with the frame problem. What do u mean exactliy with operatinal mode? The functional mode? I put off the Jumper of SOP2 to get into functional mode. I also pressed thereset button on the board and the checked the coms -> still the same. Regards, Sven
• Hi Sven, I did mean functional mode. Have you confirmed that your HW and USB connection are still always working? Was this demo running before? Have you run the out-of-box demo as well? That uses a different visualizer and different binary. Can you try to load that and run from the web based visualizer? Instructions are here. Regards, Jackson
• Hey Jackson, I am quiet sure that the HW and the USB are still working, because I can always see them in the device-manager. Not sure how else i could confirm it. Yes we already tried to run the out.of-box demo. But we also had there a problem. It did not connect to the ports... So we surrendered there and hoped we will not have the same error in the other demos... Regards, Sven
• You are using the IWR6843ISK board correct? Make sure you choose the 6843 platform in the OOB visualizer. Also, one image you had showed you are using the ICBOOST board. Which revision of the IWR6843ISK board do you have? The recent ones are able to connect to the GUI by itself. Have you tried using this board standalone? This would help to isolate the problem and verify the sensor is working. Regards, Jackson
• Hey Jackson, The one picture was misstaken. I always selected the XWR68XX in the OOB visualizer. I just forgot to do it for the picture... We are getting closer. The sensor is working perfectly fine without the MMWAVEICBOOST. It gave no errors while flashing and the out-of-box demo worked perfecly! I am using the MMWAVEICBOOST with the ISK6843ISK, because we need the BUS to get the DATA into another device... So I guess the question now is what is wrong with the MMWAVEICVOOST. By the way thanks a lot for your patience and your effort!! You are helping me a lot! Regards, Sven
• Hi Sven, Are you certain you are unplugging both the barrel jack power and USB cable from the sensor before trying to connect and flash with uniflash? This sounds like similar issues we see when flashpythonsubprocess gets hung up. If you are reconnecting the USB every time, try killing the flashpythonsubprocess.exe process from task manager. See the post below for more information.
• Hey Jackson, Jeah i always unplug the barrel jack power and the USB cable. But just to be sure I made a video of the whole process. Maybe you can see if I am doing anything incorrectly. Is this the right link? Because you didn't mention anything about Baudrate in your last reply. But is it still important for our error? I also tried to kill the flashpythonsubprocess.exe from the task manager. It only showed up once. After i killed it, the Console told me after 1/2 s that everything is uploaded. I doubt that it uploaded everything correctl in his short time. After this, the flashpythonsubprocess.exe never showed up again in the Task manager. Regards, Sven
• Hi Sven, The post I linked to addressed a few different issues. This is a very strange issue. I have a few more suggestions before we assume it is a HW issue with the XDS chip, as there doesn't seem to be an issue with your procedure. If you search E2E for flashpythonsubprocess.exe there are several other posts with various suggestions and fixes. Have you tried using a different PC? Have you ensured your AC adapter is rated for at least 3A? Is uniflash installed/run in administrator mode? Was this happening before you downgraded the XDS firmware the first time? It seemed like it worked once after you upgraded, can you try downgrading and upgrading again to see if resetting the onboard firmware changes anything? Hope one of these helps, Jackson
• Hey Jackson, I talked about this problem with another group in college and they said they have the exact same problem with flashing. They also did everything the way the User Guide describes it but it cannot connect to the com... Jeah my Adapter can easily 3A. But i wanted to see how much current the board needs so messured the current and it was round about 0,16A. Is this ok? I uninstalled uniflash and reinstalled it again as admin. I also tried to run it after that as admin. But both did not help. I also downgraded and upgraded it. But after the upgrade it still did not work. Regards, Sven
• so i tried now to flash the image again. It still did not work. So i just restarted the computer and tried again. and after the third restart, I was able to flash the image... Here is the console. Maybe it helps...
• The demo worked now!!! And I have no idea what i did differently :DD
• Hey Jackson Heureka!! The People Counting Visualizer works finally. Honestly I am not really sure what i changed. I simply reinstalled the MMWave_People_Counting_GUI, reinstalled Uniflash. Restarted my PC for so long until it was able to flash. And now it works! So now we can focus on the actiual question: What Data do we need, to be able to see the amount of People the Sensor counts. Is there anything we have to change in the code? Regards, Sven
• That's great! The visualizer is already setup to parse this information. Please look at the UART data format in the people counting user's guide. The TID inside the target TLV will give you the total number of tracked objects since startup. Each target gets an incremented ID number to differentiate. This should work for your counting the number of people that go through the frame. Is this what you are trying to do? The file oob_parser.py is the python code doing the actual parsing of the UART stream. You can modify this as needed if you need to change something. Regards, Jackson
• Hey Jackson, we do not necessarily want to track the total number of objects since start up. We would like to display how many people are currently in the given area. This information needs to be transmitted via CAN-bus so we can use this data for actors. Do we need to add aditional code for this? If yes, where do we have to do this and is there already existing code for this. Or does the CAN Output already deliver the DATA in the right format with the flashed image. So we only have to use the right bit. Regards, Sven
• Hi Sven, Total number of targets being tracked is filtered by the GUI but easy enough to compute, simply tally the number of target TLVs (type 7). If you are adding CAN data output, this could be pushed out that way. Please see the documentation on the CAN interface in the SDK documentation at the install path below. You will need to add this to your source code. "C:/ti/mmwave_sdk_03_05_00_04/packages/ti/drivers/can/docs/doxygen/html/index.html" Regards, Jackson
• Hey Jackson, I have two problems before i can go on with the original problem: (1) How can i start the detection of the sensor without starting the visualizer? Or is this even possible? (2) How can i see the output of the sensor? Like can i see somewhere the Can data output or the UART output? Regards, Sven
• Hey Jackson, I also flashed the xwr68xx_ccsdebug.bin image now on the device so i can work over the debugger. But when i want to test the connection to my device i always get this error: Do you know the reason for this? Regards, Sven
• Hi Sven, 1. To start the sensor without using the visualizer there a a couple options. You could write a simple new program that would only send the UART data from the .cfg file needed to start the sensor. Or you can use a 'hard coded config' setup. This would add the cfg parameters in the source code for the device. There is an example here and some documentation on the CLI bypass in the following SDK documentation. file:///C:/ti/mmwave_sdk_03_05_00_04/packages/ti/demo/xwr68xx/mmw/docs/doxygen/html/index.html 2. There will be a UART data stream coming over the COM port with or without the visualizer. If you open a teraterm or putty session and select the COM port you should see the data streaming in. This will show you the sensor is working and tracking, but you will likely need a program to parse and interpret the data to make sense of it. 3. Does the error above prevent you from debugging or flashing new code? Or does it only show up when testing the connection? Regards, Jackson
• Hi Sven, Hopefully you were able to resolve all your connection issues. I am going to close this thread. Regards, Jackson
• no it didnt. we tried it now with a new board and now everything worked fine.
• Hi Sven, Thanks for the follow up info. Weird that nothing worked on the other HW. Glad you have a working system now.