# OpenCV 测试应用程序
本节介绍了如何启用 OpenCV 库并运行 OpenCV 测试应用程序
## 前提条件
- 按照 [Qualcomm Linux 编译指南](https://docs.qualcomm.com/bundle/publicresource/topics/80-70020-254/introduction.html)中所述搭建基础架构。
- 刷写最新的软件版本，将其写入至开发板中。
- 设置 SSH 连接：按照[使用 SSH](https://docs.qualcomm.com/bundle/publicresource/topics/80-70020-254/how_to.html#use-ssh) 中提到的步骤在 Permissive 模式下启用 SSH
- 确认目标设备上 `/usr/lib` 存在 `libopencv_fastcv` 库。
偶尔会出现 `fastcv_linux_aarch64_2025_02_12.tgz` 下载失败的情况，导致无法生成 `libopencv_fastcv` 库。
如果缺少该库，请按以下启用 fastCV HAL 的补丁所示修改 `opencv_4.11.0.qcom.bb` 文件并重新编译版本：
**启用 fastCV HAL 的补丁**
diff --git a/recipes-support/opencv/opencv_4.11.0.qcom.bb b/recipes-support/opencv/opencv_4.11.0.qcom.bb
index 4ce2023..69c678d 100644
--- a/recipes-support/opencv/opencv_4.11.0.qcom.bb
+++ b/recipes-support/opencv/opencv_4.11.0.qcom.bb
@@ -22,6 +22,7 @@ SRCREV_boostdesc = "34e4206aef44d50e6bbcd0ab06354b52e7466d26"
SRCREV_vgg = "fccf7cd6a4b12079f73bbfb21745f9babcd4eb1d"
SRCREV_face = "8afa57abc8229d611c4937165d20e2a2d9fc5a12"
SRCREV_wechat-qrcode = "a8b69ccc738421293254aec5ddb38bd523503252"
+SRCREV_fastcv = "f4413cc2ab7233fdfc383a4cded402c072677fb0"
SRCREV_FORMAT = "opencv_contrib_ipp_boostdesc_vgg"
@@ -31,6 +32,7 @@ SRC_URI = "git://github.com/opencv/opencv.git;name=opencv;branch=4.x;protocol=ht
git://github.com/opencv/opencv_3rdparty.git;branch=contrib_xfeatures2d_vgg_20160317;destsuffix=git/vgg;name=vgg;protocol=https \
git://github.com/opencv/opencv_3rdparty.git;branch=contrib_face_alignment_20170818;destsuffix=git/face;name=face;protocol=https \
git://github.com/WeChatCV/opencv_3rdparty.git;branch=wechat_qrcode;destsuffix=git/wechat_qrcode;name=wechat-qrcode;protocol=https \
+ git://github.com/opencv/opencv_3rdparty.git;branch=fastcv/4.x_20250212;destsuffix=git/fastcv;name=fastcv;protocol=https \
file://0003-To-fix-errors-as-following.patch \
file://0001-Temporarliy-work-around-deprecated-ffmpeg-RAW-functi.patch \
file://0001-Dont-use-isystem.patch \
@@ -66,6 +68,7 @@ do_unpack_extra() {
cache data ${S}/face/*.dat
cache wechat_qrcode ${S}/wechat_qrcode/*.caffemodel
cache wechat_qrcode ${S}/wechat_qrcode/*.prototxt
+ cache fastcv ${S}/fastcv/fastcv/*.tgz
}
addtask unpack_extra after do_unpack before do_patch
@@ -91,7 +94,7 @@ EXTRA_OECMAKE:append:x86 = " -DX86=ON"
# https://github.com/opencv/opencv/issues/21597
EXTRA_OECMAKE:remove:x86 = " -DENABLE_SSE41=1 -DENABLE_SSE42=1"
-PACKAGECONFIG ??= "gapi python3 eigen jpeg png tiff v4l libv4l gstreamer samples tbb gphoto2 \
+PACKAGECONFIG ??= "gapi python3 eigen jpeg png tiff v4l libv4l gstreamer samples tbb gphoto2 fastcv \
${@bb.utils.contains_any('DISTRO_FEATURES', '${GTK3DISTROFEATURES}', 'gtk', '', d)} \
${@bb.utils.contains_any("LICENSE_FLAGS_ACCEPTED", "commercial_ffmpeg commercial", "libav", "", d)}"
@@ -126,6 +129,7 @@ PACKAGECONFIG[tests] = "-DBUILD_TESTS=ON,-DBUILD_TESTS=OFF,,"
PACKAGECONFIG[text] = "-DBUILD_opencv_text=ON,-DBUILD_opencv_text=OFF,tesseract,"
PACKAGECONFIG[tiff] = "-DWITH_TIFF=ON,-DWITH_TIFF=OFF,tiff,"
PACKAGECONFIG[v4l] = "-DWITH_V4L=ON,-DWITH_V4L=OFF,v4l-utils,"
+PACKAGECONFIG[fastcv] = "-DWITH_FASTCV=ON ,-DWITH_FASTCV=OFF,,"
inherit pkgconfig cmake setuptools3-base python3native
@@ -224,9 +228,6 @@ SUMMARY = "Opencv : The Open Computer Vision Library, Qualcomm Fork"
SRC_URI += "file://0001-PENDING-Removed-cluster-euclidean-in-fastcv-ext.patch;patchdir=${S}/contrib/ \
file://0001-FROMLIST-Using-fastcv-static-lib-for-compilation.patch"
-EXTRA_OECMAKE += "-DOPENCV_ALLOW_DOWNLOADS=ON "
-EXTRA_OECMAKE += " -DWITH_FASTCV=ON "
-
do_install:append() {
rm -f ${D}/usr/lib/libfastcv.a
}
Copy to clipboard
刷写编译版本后，确认 `libopencv_fastcv` 库存在于 `/usr/lib` 中。
## 启用 OpenCV 库和测试包
1. 要启用 **tests** 包，请将 `tests` 包括在 `/layers/meta-qcom-hwe/recipes-support/opencv/opencv_4.11.0.qcom.bb` 配方文件的 `PACKAGECONFIG` 中，如下所示。
PACKAGECONFIG ??= "gapi python3 eigen jpeg png tiff v4l libv4l samples tbb gphoto2 tests \
${@bb.utils.contains("DISTRO_FEATURES", "x11", "gtk", "", d)} \
Copy to clipboard
默认情况下，只有编译版本中显示为测试 bin 的库会被清理。
2. 要保留测试 bin，请在 `/layers/meta-qcom-hwe/recipes-support/opencv/opencv_4.11.0.qcom.bb` 配方文件中包括以下代码：
RM_WORK_EXCLUDE += "opencv"
Copy to clipboard
3. 当版本已经被编译过了，必须在重新编译之前清理 OpenCV。要清理 OpenCV，运行以下命令：
bitbake -fc cleanall opencv
Copy to clipboard
编译期间不会修改代码，因此直接编译不会生成相应的二进制文件。
4. 要编译 OpenCV，运行以下命令：
bitbake opencv
Copy to clipboard
库的路径为 `tmp-glibc\sysroots-components\armv8-2a\opencv\usr\lib`。
Bin 的路径为 `tmp-glibc\work\armv8-2a-qcom-linux\opencv\4.11.0.qcom\build\bin`。
## 运行现有测试应用程序
要调用 OpenCV API，原生 OpenCV 测试示例可以参考示例程序。使用以下步骤调用 Snapdragon 设备上的应用程序。完整地编译镜像以确保所有库都是应用程序镜像的一部分。
1. 按照 [Qualcomm Linux 编译指南](https://docs.qualcomm.com/bundle/publicresource/topics/80-70020-254/)的描述在设备上刷写镜像。
OpenCV 库位于设备上的 `/usr/lib`。
2. 关于测试数据，`git clone` 在 [https://github.com/opencv/opencv_extra/tree/4.11.0](https://github.com/opencv/opencv_extra/tree/4.11.0) 的项目。
3. 为了避免在尝试将测试数据和 bin 推送到设备时出现只读错误，使用设备的 IP 地址登录 SSH 终端，并运行以下命令重新挂载设备。
mount -o remount,rw /usr
Copy to clipboard
4. 使用 `scp` 命令将测试数据推送到首选主机。
例如：`scp -r [file] root@[IP-ADDR]:/tmp`
5. 使用 `scp` 命令将所需的测试 bin 推送到 `/usr/bin`。
例如：`scp -r [test] root@[IP-ADDR]:/usr/bin/`
6. 登录 SSH 终端并运行以下命令。
chmod 777 /usr/bin/
cd /usr/bin
export OPENCV_LOG_LEVEL=DEBUG
export OPENCV_VIDEOIO_DEBUG=1
export OPENCV_TEST_DATA_PATH=/tmp
./ --gtest_output=json:/tmp/results.json --gtest_also_run_disabled_tests
exit
Copy to clipboard
结果存储在 `/tmp/results.json`。要更改结果文件名，请修改 `--gtest_output=json:/tmp/{results.json}` 参数。
Last Published: Aug 12, 2025
[Previous Topic
概述](https://docs.qualcomm.com/bundle/publicresource/80-70020-21SC/topics/overview.md) [Next Topic
实现新示例应用](https://docs.qualcomm.com/bundle/publicresource/80-70020-21SC/topics/sample-applications.md)