Constructor for GpuMat class fails with RuntimeException due to no CUDA support detected

I'm running Ubuntu Linux 24.10 with Java 21 installed. I have a copy of opencv 4.10 that I built from source with CUDA enabled in /opt/opencv and javacv 1.5.11 installed.

If I call new GpuMat() in my application I get the exception

Exception in thread "Worker Thread" java.lang.RuntimeException: OpenCV(4.10.0) /home/runner/work/javacpp-presets/javacpp-presets/opencv/cppbuild/linux-x86_64/opencv-4.10.0/modules/core/include/opencv2/core/private.cuda.hpp:106: error: (-216:No CUDA support) The library is compiled without CUDA support in function 'throw_no_cuda'

	at org.bytedeco.opencv/org.bytedeco.opencv.opencv_core.GpuMat.allocate(Native Method)
	at org.bytedeco.opencv/org.bytedeco.opencv.opencv_core.GpuMat.<init>(GpuMat.java:93)
	at jgauss/jgauss.worker.OpenImageRequest.processRequest(OpenImageRequest.java:35)
	at jgauss/jgauss.worker.BGThread.run(BGThread.java:53)

I enabled -Dorg.bytedeco.javacpp.logger.debug=true to see if anything failed to load, and the only remaining failure i have is for nppicom which appears to no longer exist with current CUDA if I have /opt/opencv/lib included first in LD_LIBRARY_PATH. If I remove /opt/opencv/lib from LD_LIBRARY_PATH then I also get an error telling me that opencv_cudev is not in java.library.path so LD_LIBRARY_PATH seems to be honored.

I compiled the following program to see if my opencv libraries supported CUDA, ran it, and saw the memory was allocated on the GPU by watching nvidia-smi

#include <opencv2/core/cuda.hpp>
#include <unistd.h>

int main(int argc, char *argv[]) {
    cv::cuda::GpuMat *mat = new cv::cuda::GpuMat(10000, 10000, CV_MAKETYPE(CV_32F, CV_32FC4));
    sleep(60);
}

So I'm not sure what's going wrong with the java code. I'm running this in a current Eclipse IDE, where there are no error message other than this when I run it, so I think my Eclipse run environment is set up correctly.

Suggestions?