OpenCV CascadeClassifier多线程中的C ++接口 [英] OpenCV CascadeClassifier C++ Interface in Multiple Threads

问题描述

我想在多个线程中使用CascadeClassifier对象运行OpenCV C ++接口。

I would like to run an OpenCV C++ Interface using the CascadeClassifier Objects in multiple threads.

我的程序工作原理是我的主线程加载some_file.xml转换为CascadeClassifier对象。生成三个或更多线程，并且它们通过级联对象。程序很快崩溃。我做了几个测试，并得出结论，CascadeClassifier对象在执行detectmultiscale函数时不是线程安全的。

The way my program works is my main thread loads "some_file.xml" into a CascadeClassifier object. Three or more threads are spawned and they are passed the cascade object. The program soon crashes thereafter. I have done several tests and concluded that the CascadeClassifier object is not thread-safe when doing a "detectmultiscale" function.

我想避免读取同一个文件关闭硬盘驱动器每次生成一个新的线程。

I would like to avoid having to read the same file off of the hard drive every time a new thread is spawned. How can this be avoided?

推荐答案

如果你使用HaP级联的LBP级联存储新格式，那么你可以避免从每个新线程的文件系统读取级联：

If you are working with LBP cascade of with Haar cascade stored in new format then you can avoid reading cascade from file system for each new thread:

将级联加载到内存中：

cv::FileStorage fs(path_to_cascade_file, cv::FileStorage::READ);
if (!fs.isOpened())
    HandleError();

接下来将fs对象传递给每个新线程并创建CascadeClassifier对象：

Next pass fs object to the each new thread and create CascadeClassifier object:

cv::CascadeClassifier cc;
if (!cc.read(fs.getFirstTopLevelNode())
    HandleError2();

这篇关于OpenCV CascadeClassifier多线程中的C ++接口的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！