Optimizing loop time

The first step to optimizing your project's processing time is to generate a trace file from the Matrox Profiler; the file includes statistics that will provide a very useful overview of the project's execution including, for example, the duration of individual steps. This might be the quickest way to cut out large chunks of unproductive processing time. For more information on generating a trace file, see the Profiling and troubleshooting logs section later in this chapter.

With this overview in mind, there are a number of optimizations you can still perform in your flowchart and operator view at design-time to reduce the amount of processing required when the project is deployed:

At design-time the project's execution time is determined by your host CPU, but the typical desktop computer is significantly faster than a supported Matrox smart camera's processor. Therefore, timing results at runtime are very different from the timing results at design-time, and it might be beneficial to adjust the Timeout of a particular step accordingly, using the information in the Matrox Profiler's generated trace file. The default Timeout for most steps is 2 sec.
Whenever possible, specify the number of occurrences you expect to find when using the PatternMatching step, the StringReader step, or the CodeReader step. The default settings will try to find all occurrences, even though you might only be interested in finding a specific few.
Whenever an Analysis and Processing step needs to work with an image, specifying a search region will force it to only work inside that region and ignore the rest of the image. This can eliminate unnecessary work, and minimize any unwanted noise that might affect the usefulness of the step's results. For more information, see the Search regions section in Chapter 2: Building a project.
If you are using an ImageProcessing step and the processing destination(s) will fill the entire output image, you do not need to initialize the destination buffer. Set the Image property to None and ClearImage to False to save the processing time required to perform a copy or clear.

When grabbing

There are certain optimizations you can do when grabbing images:

In the Platform Configuration dialog you can specify a reduced image size, or change the portion of the image that is scanned during each grab, decreasing the grab and/or processing time. For more information, see the Changing the camera's inputs subsection of the Platform settings for your camera section in Chapter 28: Acquisition.
If you enable the use of image queues in the Platform Configuration dialog, then grabbed images can be kept in the queue will be processed as soon as the previously grabbed image has been processed. Note that, by default, enabling the triggered grab mode will also enable image queues for the camera. For more information, see the Grab timing section in Chapter 28: Acquisition.

Analysis and Processing steps

There are several possible optimizations that can be done in the Analysis and Processing steps:

When using a Measurement step, specifying several of the advanced constraints can further optimize edge and stripe determination. For more information, see the Measurement step advanced constraints section in Chapter 7: Measurement step.
When using a BlobAnalysis step, ensure that only the necessary operations for your task are performed. It might not be necessary, for example, to fill in all the holes. The time to fill in a hole is proportional to the number of pixels across its smallest dimension, so high resolution images are susceptible to long processing times. For more information, see the Configure settings subsection of the Procedure for using the BlobAnalysis step section in Chapter 3: BlobAnalysis step.
When using a Metrology step, the extraction of measured features when the image contains visibly rough texture or clutter can significantly increase processing time, which is compounded if the image is high resolution. Depending on the application, other steps might be better suited to the task. For more information, see the (Physically) measured features section in Chapter 18: Metrology step.
When using a ModelFinder step, a search done in a geometrically complex image should be done in Controlled mode, instead of the General mode. Geometrically complex images contain many edgels that, depending on the shape of the model, can lead the ModelFinder step to find many unhelpful occurrences. In addition to Controlled mode, consider using one of the shape finder steps.
When using a StringReader step, you can use the MIL StringReader Interactive Utility to specify a maximum acceptance threshold, reducing the search time. For more information, see the Add a string model subsection of the Creating and testing a context using the MIL String Reader Interactive Utility section in Chapter 19: StringReader step.
When using a CodeReader step, if possible, specify the number and size of cells in the code. For more information, see the Fine-tuning your read subsection of the Procedure for using the CodeReader step section in Chapter 4: CodeReader step.

Operator view

There are some optimizations that can be done for the operator view:

Decrease the overall load of the operator view by reducing the size of the operator view display. A smaller display can be compensated for with a more detailed settings page, or with several simpler pages.
Reduce the number of annotations whenever possible.
Specify the Maximum update rate to limit the number of times a publishing group will update each second. For more information, see the When to change the defaults subsection of the Controlling when data is sent to the operator view section in Chapter 50: Customizing the operator view.
Be selective about which inspection results are saved or displayed in the filmstrip by specifying inspection labels, and only publish them when actual changes occur. For more information, see the Controlling what gets saved and displayed subsection of the Filmstrip section in Chapter 50: Customizing the operator view.

Other

There are more general optimizations that can be done, outside the above steps:

The system automatically frees up memory and improves performance using a garbage collector, which periodically cleans up data objects that are no longer in use. The details of when the cleanup occurs can be set in the Performance tab of the Runtime properties dialog, accessible from the Project Runtime Properties menu item.
Unused data also need to be cleaned during the project's execution. You can manually decide when cleanup should occur:
- At the end of each loop. Executes the cleanup each time it reaches the end of the selected loop. This option tends to give a more consistent loop time. Note that, even with this option selected, the system will perform the cleanup at any point should it become necessary.
- Automatic. Allows the system decide when the cleanup will be made. This option tends to add a longer cleanup time during some iterations and no cleanup time during others.