BMS显著性模型Linux Mint环境配置出错，寻求解决方案

I’ve worked through similar porting issues with Windows-based MATLAB mex projects to Linux, so let’s walk through the most likely fixes for your BMS (from the Exploiting Surroundedness for Saliency Detection: A Boolean Map Approach paper) setup:

1. Fix OpenCV Path & Version Mismatches #

First, confirm your OpenCV installation is solid:

• Open a terminal and run pkg-config --modversion opencv to verify you have a 2.4.x version installed. If you get a 3.x+ version, you’ll need to install OpenCV 2.4.x (the API changes between 2.x and 3.x will break the BMS code).
• Open /mex/compile.m and update the path definitions to use Linux-style forward slashes (/) instead of Windows backslashes (\). For example:

  % Replace Windows-style paths like:
  % INCLUDEPATH = 'C:\opencv\include';
  % With Linux paths (adjust to your actual install location):
  INCLUDEPATH = '/usr/local/include/opencv';
  LIBPATH = '/usr/local/lib';
  

  Also, ensure the linker flags reference Linux library names (.so instead of .lib). For example, use -lopencv_core instead of opencv_core.lib.

2. Configure MATLAB’s Mex Compiler #

Linux requires a compatible compiler for mex files:

• Open MATLAB and run mex -setup to select a supported GCC compiler. Make sure the GCC version matches your MATLAB release’s supported versions (check MATLAB’s docs if you’re unsure—older MATLAB versions won’t work with the latest GCC).
• Scan compile.m for Windows-specific compiler flags (like /EHsc or /MT) and replace them with Linux equivalents. A common one is adding -fPIC (position-independent code) to the mex flags, which is mandatory for Linux mex files.

3. Fix File Permissions #

Linux enforces stricter permissions than Windows:

• Run chmod -R u+rwx /path/to/your/BMS/mex/ to give your user full read/write/execute access to the mex directory and its files. This prevents "permission denied" errors during compilation.
• If you installed OpenCV from source, double-check that the library files in /usr/local/lib are readable by your user (run ls -l /usr/local/lib/libopencv*.so to confirm permissions).

4. Debug the Compile Command #

To get clear error details, modify compile.m to print the full mex command before running it:

% Add this line right before the mex execution line
disp(mex_command);

Run the script again, copy the printed mex command, and execute it directly in a terminal. This will show you exactly where the compilation is failing (e.g., missing include files, wrong library paths) instead of vague MATLAB error messages.

5. Check for Missing Dependencies #

Even if OpenCV is installed, you might be missing system libraries it depends on:

• Run ldd /usr/local/lib/libopencv_core.so in the terminal. Look for any lines marked not found—install those libraries with sudo apt install <library-name>.

内容的提问来源于stack exchange，提问作者sana