Ubuntu 16.04 Docker中OpenCV链接CMake失效编译报错问题

I’ve run into this exact frustrating issue with older OpenCV versions before—even when CMake confirms it’s detected the library, function signature mismatches or incomplete linking can throw these confusing "no matching function" errors. Let’s break down the most likely fixes, tailored to your setup:

1. Verify Your Function Call Matches OpenCV 3.4.2’s Exact Signature #

OpenCV 3.4.2’s cv::imdecode has a strict, non-overloaded signature that’s easy to misalign with newer version habits:

cv::Mat imdecode(cv::InputArray buf, int flags);

Double-check two critical details:

• The first argument must be a cv::InputArray compatible type (e.g., std::vector<uchar> for raw image bytes, not a raw char* or std::string). If you’re passing a string of bytes, convert it first:

  std::string raw_image_data = "...";
  std::vector<uchar> image_buffer(raw_image_data.begin(), raw_image_data.end());
  cv::Mat img = cv::imdecode(image_buffer, cv::IMREAD_COLOR);
  

• The second argument must be an integer flag (stick to the cv::IMREAD_* constants like IMREAD_GRAYSCALE—don’t use enums or string values from newer OpenCV versions).

2. Fix CMake Linking to Include Required OpenCV Modules #

The most common culprit here is missing the imgcodecs module (which houses imdecode). Even if find_package(OpenCV) succeeds, it might not link all needed components by default.

Update your root CMakeLists.txt to explicitly declare required modules:

cmake_minimum_required(VERSION 3.10)
project(YourProjectName)

# Explicitly request the modules your code uses
find_package(OpenCV 3.4.2 REQUIRED COMPONENTS core imgcodecs) # Add others like highgui if needed

# Link OpenCV to your main target
add_executable(main src/main.cpp)
target_link_libraries(main PRIVATE ${OpenCV_LIBS})

# Pass OpenCV config to tests directory
add_subdirectory(tests)

In your tests/CMakeLists.txt, ensure test targets also link against OpenCV properly:

add_executable(image_decoding_test test_imdecode.cpp)
target_link_libraries(image_decoding_test PRIVATE ${OpenCV_LIBS})

3. Confirm the Docker Image Has a Complete OpenCV Build #

Some minimal OpenCV Docker images skip non-core modules to save space. To check if imgcodecs is present in jjanzic/docker-python3-opencv:contrib-opencv-3.4.2:

1. Spin up an interactive shell in the container:

   docker run -it --rm jjanzic/docker-python3-opencv:contrib-opencv-3.4.2 /bin/bash
   

2. List compiled OpenCV modules:

   opencv_version --verbose
   

Look for imgcodecs in the "Modules" section. If it’s missing, you’ll need to either:

• Switch to a full-featured OpenCV 3.4.2 Docker image, or
• Build OpenCV from scratch in your Dockerfile with the imgcodecs module enabled (add -DBUILD_opencv_imgcodecs=ON to your cmake build flags).

4. Resolve Local Ubuntu 18.04 OpenCV Version Conflicts #

Ubuntu 18.04’s default repos include OpenCV 3.2.0, which often overrides manually installed 3.4.2 builds. To force CMake to use your desired version:

1. Add this line to your CMakeLists.txt before find_package(OpenCV):

   set(OpenCV_DIR /path/to/your/opencv-3.4.2/build) # Replace with your actual build directory path
   

2. Wipe your existing build directory and re-run CMake to pick up the correct version:

   rm -rf build && mkdir build && cd build && cmake ..
   

内容的提问来源于stack exchange，提问作者Gertjan Brouwer