Proper Way to Compile & Run ASI SDK Examples?

Runsva

I want to try and get some images from my ASI120MM camera, using the SDK libraries, on my Raspberry Pi. I have Raspian 11, 32-bit.

Before that, I wanted to try and compile some of the demo examples that the SDK came with. There are supposed to be some examples that would simply just let me snap an image from the camera via my USB cable, which is exactly what I want.

Firstly, in the Downloads section of the ZWO website, under the SDK section, there are different categories of the SDK that are available to download. There are the standard ones, "EFW", "EAF" and "USBST4". What are the differences between the different types of SDK, and which one should I download?

After downloading & extracting the first SDK option on my RP, there is a demo folder, alongside the doc, include and lib folders. Inside the demo folder, there is a Makefile file, that I would assume is the file that needs to be triggered in order for the demo to be built. After switching to that demo folder in my RP terminal, and entering the make command, I get an error stating -m32 is an unrecognized command. Inside the readme file that is inside the demo folder, it is stated that I need to install opencv2 first, so I tried to find how to install that library on my RP online. After searching, OpenCV appears to be a Python module, that is used for detecting camera motion. Inside the bin folder that's in the demo folder, there appear to be some compiled executables that have already been compiled. Attempting to run any of this, either outside or inside the terminal results in nothing.

There is also a document inside the doc folder of the SDK folder, however it contains no guidance on how to compile these examples. mainly just function definitions.

I'm quite a bit confused as to exactly how I'm supposed to compile this demo. How am I supposed to run this Makefile, so that the demo builds successfully?

Thanks for reading my post, any guidance is appreciated.

franco

Runsva

There are no ARM binaries prebuilt so you will need to either cross compile on a x86 Linux machine, or build the demo program on the Pi but change the compile flags.

See the readme.txt file in the demo dir.

franco

Out of curiosity I thought I would try building the demo myself.

On my Linux Mint 21.1 machine the demo would not compile, I had to make a link for the opencv2 includes which are in /usr/include/opnecv4. I also had to comment out the calls to CV_RGB which are undefined. After that the test_gui2_video ran ok.

I then tried to cross compile for arm7 but the compiler gave an error which I didn't understand.

cc1plus: warning: switch ‘-mcpu=cortex-m3’ conflicts with switch ‘-march=armv7’
cc1plus: error: ‘-mfloat-abi=hard’: selected architecture lacks an FPU

So I copied the SDK to my ASIAir Pi running Buster

I installed libopencv-dev and libncurses-dev and ran make platform=armv7, worked fine.

I ran the test_gui2_video with a remote display, thus

DISPLAY=opti:0 LD_LIBRARY_PATH=../../../lib/armv7/ ./test_gui2_video

I was using my ASI120MC-S connected to a guide scope and used all the defaults, I get video but it's black and white.

Runsva

Thanks for your reply;

Hmm, that is very interesting. If I try to compile the demo with any of the make configuration, I get an error. However, the only configuration that appears to load anything, is the armv7 config.

It's telling me that it can't find opencv2/highgui/highgui_c.h, however I installed the libopencv-dev and libncurses-dev libraries via the terminal. I also rebooted the Pi after this, and it still cannot find it.

How would I go about linking them? I thought they were already linked through the Makefile file, right?

franco

Runsva

Check if you have /usr/include/opencv4/opencv2, in which case make a symlink

ln -s /usr/include/opencv4/opencv2 /usr/include/

If you then get CV_RGB undefined, comment it out in the code.

main_SDK2_snap.cpp: //cvSet(pRgb, CV_RGB(180, 180, 180));
main_SDK2_video.cpp: //cvSet(pRgb, CV_RGB(180, 180, 180));

Runsva

franco
Thank you for your reply;

I managed to create a Symbolic Link to the opencv2 folder. Now the program compiles, after I commented out the 2 lines you mentioned, however if I try to run any of the executables on my Pi, nothing happens... Maybe compiling with the armv7 config was not the right choice after all.

I wanted to try to build it for the x86 config, to see if that improves the situation. While I initially got an error stating that the argument -m32 was not recognized, chaning that to -mbe32 appears to have solved that. However, I get an error stating that libASICamera2.so was not found.

Now, the readme file in the demo folder actually lists a supposed solution for that, however I have no idea how to implement it in Linux. It states that, I can either:

Add a .conf file that contains the path of the library to /etc/ld.so.conf.d/, run ldconfig
Add compile option -Wl, -rpath=<library path>

I'm not quite sure how to do either one. If I try to supply -Wl and -rpath=<library path> as arguments to the make command in my terminal, I get an unexpected newline character error.

How would I go about implementing any of those two in Linux?

franco

Runsva

Use LD_LIBRARY_PATH the way I did to specify the path to the arm asi camera lib.

If you have graphics on your Pi you don't need to set the DISPLAY.

Note that the Pi has an ARM CPU, you can't run a x86 binary or anything compiled for a different architecture.

Runsva

franco
Thank you for your reply;

That worked! The program compiles & runs, I wasn't aware that I had to run the program from the terminal, and not just double-click the executable.

At least the program is working. There is not an actual viable image, it appears like gray static. What are some good settings to use, such as the exposure time and such, that you can reccomend? I'm not familliar with these camera characteristics at all.

franco

Runsva

Is your camera the mini or the S model with the all sky lens?

Do you have a PC? You should download ASIStudio and see if you get a picture with ASICap. It will help you get an idea of what settings work best.

franco

To add the library libASICamera2.so with ldconfig, create a new file and add the path, for example

/etc/ld.so.conf.d/asi.conf

/opt/ASI_linux_mac_SDK_V1.28/lib/armv7 <-- change this to your actual path

Then type ldconfig. You only need to do this once.

To check it worked type

ldconfig -v | grep libASI

Software@ZWO

Franco, thank you very much for your kind reply. Now is the Chinese Spring Festival. We wish you a happy Spring Festival.
Later, we will write the method of compiling Demo into Manual. We will try to do better and better.

Runsva

Thanks for all your replies franco, you really saved the day;

I have another question, this time regarding some of the most nitty-gritty workings on the makefile for this project:

I wanted to try and create my own version of the test_gui2_video example. I went ahead and moved that CPP file, renamed it to TEST.cpp, alongside the include and lib folders from the ASI SDK folder, and put them all in the same folder (since I'll only be compiling this for the armv7 platform, I moved the object files in the lib/armv7 folder directly into the lib folder, and removed the other platform options). I also copied over the makefile.

Inside the makefile, I removed all the options for any platform except for armv7, as that is the only platform I'll be using.

I modified the makefile, now it looks like this:

ver = debug
platform = armv7

CC = g++
OPENCV = -lopencv_core -lopencv_highgui -lopencv_imgproc

LIBSPATH = -L/lib -I/include

ifeq ($(ver), debug)
DEFS = -D_LIN -D_DEBUG 
CFLAGS = -g  -I $(INCLIB) -L $(LDLIB) $(DEFS) $(COMMON) $(LIBSPATH)  -lpthread  -DGLIBC_20
else
DEFS = -D_LIN 
CFLAGS =  -O3 -I $(INCLIB) -L $(LDLIB) $(DEFS) $(COMMON) $(LIBSPATH)  -lpthread  -DGLIBC_20
endif

CC = arm-linux-gnueabihf-g++
AR= arm-linux-gnueabihf-ar
CFLAGS += -march=armv7 -mcpu=cortex-m3 -mthumb

all: TEST

TEST:TEST.cpp
	$(CC)  TEST.cpp -o TEST $(CFLAGS) $(OPENCV) -lASICamera2
	cp TEST bin/$(platform)/

clean:
	rm -f TEST

Running the make command for the makefile above gives me an error that the file ASICamera2.h cannot be found, even though it's path is specified at the top by LIBSPATH = -L/lib -I/include (specifically the include file).

Is there something I'm doing wrong here? I'm not sure why it would not be able to find it, it's directory is explictly listed. Is there perhaps a syntax mistake in the makefile? I have no experience with makefiles at all.

Also, I was wondering, where is the -lASICamera2 reference towards the bottom of the makefile defined? It's not defined anywhere previously in the makefile, or any of the code. How does the makefile know where to look for that, as it wasn't explictly listed in the original makefile for the demo either?

franco

Runsva

-I/include, is an absolute path, it points to /include.

Include paths are normally in CFLAGS. If your includes are in a directory below the directory with the Makefile, you shoud use -Iinclude. Same goes for libs.

Runsva

franco
Thanks for your reply;

That makes so much sense, I replaced the libspath with LIBSPATH = -lib -include. That appears to have solved that issue., what was the purpose of the -I and -L flags in front of each one of the paths of the original makefile?

However, there's still an issue with the Linker flags. That's why I asked the second part of my question in my previous reply;
It's currently telling me that it can't find -lpthread, or any other Linker flags. How does the Linker know what -lpthread points to? In the makefile of the original demo, there were no issues like that, so that makes me wonder what exactly the issue here is, as I didn't change anything else with the project. Are these Linker flags defined in a file that I may have not moved over?

franco

Runsva

Include files are used at compile time and are defined in cflags, you can compile to objects then link them with ld. In your case you compile and link in one go, so all the flags are passed to the compiler.

-L adds a path for the linker to search for libraries, which are normally called lib'name'.so When specifying a library you drop the lib and .so extension and just use -l'name'

Runsva

franco
Thanks for your reply;

That makes much more sense, and I was finally able to get my program to compile after adding the -I and -L prefixes to my paths. Thank you!

I now edited my code substantially, so it runs better, and tuned it to my needs.

There is one more thing I would like to ask, and it's regarding the opencv2 library I'm using to process the image data.

Currently, the image data is stored in a pointer variable of type IplImage*. There should be a function named cvSaveImage in the opencv2 library that would allow one to save the contents of the image data pointer to a file, however whenever I try to call that function, it prompts me with an error because it can't be found...

After looking into the opencv2 library, I found out that I was actually including the opencv2/highgui/highgui_c.h file to access all the opencv functions. While that file contains all the functions that I was previously called, it's curiously missing the cvSaveImage function...

Would you know where the cvSaveImage function is declared & defined in the opencv2 library for the Raspberry Pi?

franco

Runsva sorry but I've never used the software myself.

w7ay

Runsva Would you know where the cvSaveImage function is declared & defined in the opencv2 library for the Raspberry Pi?

Did you include the header file imgcodecs_c.h ?

https://github.com/cinder/Cinder-OpenCV3/blob/master/include/opencv2/imgcodecs/imgcodecs_c.h

See the documentation for cvSaveImage here:

https://docs.opencv.org/3.0.0/da/d0a/group__imgcodecs__c.html

It tells you to use the above header file.

Also, if you are using the v3 or v4 API for openCV, you might as well just use imWrite instead of cvSaveImage:

https://docs.opencv.org/3.4/d4/da8/group__imgcodecs.html

Chen

franco

w7ay Runsva that file on my machine just contains the line

#error "This header with legacy C API declarations has been removed from OpenCV. Legacy constants are available from legacy/constants_c.h file."

/usr/include/opencv4/opencv2/imgcodecs/imgcodecs_c.h

I get the feeling the C API has been dropped from the OpenCV project.

w7ay

franco #error "This header with legacy C API declarations has been removed from OpenCV. Legacy constants are available from legacy/constants_c.h file."

If you can't find the older (ver 2 API) anymore, just use ver 3, and the imWrite call.

Chen