How do I define NvInfer.h for TensorRT?

I try to start a TensorRT example with Cross Compiling on the Jetson TX2.
However, I have some problems there.

As an example I chose the TensorRT-sample sampleUffMNIST.

make all -C /home/ubuntu/CUDA_Programme/Test 3/Debug 
make: Entering directory '/home/ubuntu/CUDA_Programme/Test 3/Debug'
Building target: Test 3
Invoking: NVCC Linker
/usr/local/cuda-9.0/bin/nvcc --cudart static -L/usr/lib --relocatable-device-code=false -gencode arch=compute_62,code=compute_62 -gencode arch=compute_62,code=sm_62 -m64 -ccbin aarch64-linux-gnu-g++ -link -o  "Test 3"  ./sampleMNISTUff.o   
./sampleMNISTUff.o: In function `nvinfer1::(anonymous namespace)::createInferBuilder(nvinfer1::ILogger&)':
/usr/include/aarch64-linux-gnu/NvInfer.h:4362: undefined reference to `createInferBuilder_INTERNAL'
makefile:58: recipe for target 'Test 3' failed
./sampleMNISTUff.o: In function `main':
make: Leaving directory '/home/ubuntu/CUDA_Programme/Test 3/Debug'
/home/ubuntu/CUDA_Programme/Test 3/Debug/../sampleMNISTUff.cpp:272: undefined reference to `nvuffparser::createUffParser()'
/home/ubuntu/CUDA_Programme/Test 3/Debug/../sampleMNISTUff.cpp:288: undefined reference to `nvuffparser::shutdownProtobufLibrary()'
collect2: error: ld returned 1 exit status
make: *** [Test 3] Error 1
> Shell Completed (exit code = 2)

How do I correctly reference NvInfer.h so that the example works in crosscompile?