Feature enhancement: PnP algorithm for pose Matrix calculation and other functions improves

[kalwalt]

From this issue #21 i'm trying to implement the PnP algorithm in our WebARKit project.

[kalwalt]

I have this issue when the pinball image is tracked:

Uncaught RuntimeError: divide by zero
    at WebARKit::getPoseMatrix() (00d7f0c6:0x2be8cb)
    at emscripten::internal::MethodInvoker<emscripten::val (WebARKit::*)(), emscripten::val, WebARKit*>::invoke(emscripten::val (WebARKit::* const&)(), WebARKit*) (00d7f0c6:0x171b2a)
    at WebARKit.WebARKit$getPoseMatrix [as getPoseMatrix] (eval at EI (WebARKit.js:2:4774185), <anonymous>:8:10)
    at A.value (WebARKit.js:2:4803539)
    at A.value (WebARKit.js:2:4802767)
    at processFrame (worker.js:54:8)
    at self.onmessage (worker.js:16:7)
$WebARKit::getPoseMatrix() @ 00d7f0c6:0x2be8cb
$emscripten::internal::MethodInvoker<emscripten::val (WebARKit::*)(), emscripten::val, WebARKit*>::invoke(emscripten::val (WebARKit::* const&)(), WebARKit*) @ 00d7f0c6:0x171b2a
WebARKit$getPoseMatrix @ VM32:8
value @ WebARKit.js:2
value @ WebARKit.js:2
processFrame @ worker.js:54
self.onmessage @ worker.js:16

I dont't get why this happens but i think it's related on how the poseMatrix is passed out.
To be more specific: i think that this issue not come from the Typescript or javascript side, but it's in WASM code.

[kalwalt]

I was trying to make WebARKitGL file as only header, but i think it's better to continue with the convention as with other classes files(header and source in .cpp).

[kalwalt]

I have implemented the getCameraProjectionMatrix method so we have, in theory, all the "tools" to display a 3d mesh with our own 3d Engine now.

[kalwalt]

The action script fails with a seg fault probably because ubuntu need libjpeg or libopenjpeg installed.

[kalwalt]

The action script fails with a seg fault probably because ubuntu need libjpeg or libopenjpeg installed.

That is not right, otherwise it should fails also before.
The fact is: the image is 3 channels (RGB) but we expect 4 channels (RGBA), so the seg fault could be caused by this. The strange fact is this not happens on my local machine, and i'm testing with WSL and Ubuntu 22.04.
I will create a RGB ColorSpace and make required changes to WebARKitTracker class.

EDIT: There is already ColorSpace::RGB, no needs to add...

[kalwalt]

now the github action si passing regularly, i can add also support to RGB also for the processFrameData, even if i don't know if this will be ever used.

[kalwalt]

It would preferable, and this is appliable only to test for linux or Windows or other OS excluded Emscripten, to have an overloaded initTracker function with refData as a cv::Mat, maybe it could be useful also processFrameData, but we can start with the initTracker for now.
Moreover we have considered only RGB, RGBA or GRAYSCALE data, but data coming from imread i think are BGR. so we should consider also that options.

[kalwalt]

test failing because i have not updated the tests with the last changes, i will fix it.

[kalwalt]

test failing because i have not updated the tests with the last changes, i will fix it.

fixed with 63946b4

[kalwalt]

I think that the PnP code could be correct, i tested in the webarkit-testing repository in a browser with WASM. Maybe i could make a simple test in an desktop environment, testing on WASM i have some difficulties passing the data in the correct way, maybe with linux or Windows is simpler?

[kalwalt]

PnP code is implemented, merging it.