Machine vision depends on performance. At the same time, accuracy and reliability must not be affected. Driven by the high amount of data throughput of any machine vision application, the possibility to split the processing into several threads that run on different processors has been used for a long time. This is called multi-threading.
HALCON does more: HALCON automates the data splitting to multiple processors (Parallel HALCON), a task that the programmer has to solve with other systems. This so called Automatic Operator Parallelization (AOP) is a worldwide unique feature, developed by MVTec, and therefore fits perfectly to the launch of multi-core CPUs.
After splitting the input data, AOP distributes the data to the available CPUs to process them. Subsequently, the computed data are merged automatically to achieve the same results that are obtained on with a single processor. All this is done without any extra work by the programmer. This automation yields a performance comparable to manually programmed multi-threading, but at the same time, brings many benefits: It is much easier to use, it is not error-prone, and also programmers with little multi-threading experience can immediately use it.
Of course, if desired the programmer can deactivate the AOP to use the manually programmed multi-threading. Also in this regard, HALCON is a highly flexible software library for machine vision.
Contact: MVTec Press Contact