性能优化#

This topic explains what grab strategies are and why they are important.

Industrial cameras often produce images at a constant and potentially very high rate (e.g., 60–200 or more fps). In many applications, processing pipelines, like Python, OpenCV, AI inference, etc., can't keep up with this rate.

This creates a fundamental mismatch:

Camera (Producer) → fast
Application (Consumer) → slower

Without a grab strategy, one of these situations will arise:

• Buffers grow indefinitely: not possible
• Buffers overflow randomly: undesirable

Grab strategies define how the system behaves under load.

Mental Model#

Camera → Frame Stream → Buffer Queue → RetrieveResult() → Application

If the application is slower than the camera, the queue fills up and a decision is required how to handle the situation. This is what a grab strategy is for.

Grab Strategies#

Two grab strategies are available.

LatestImageOnly#

Only the most recent frame is kept. Older frames are dropped while the application is busy. This visualization illustrates this:

Camera:    F1 → F2 → F3 → F4
App reads:        →    F4
Dropped:   F1, F2, F3

Advantages:

• You always see the most recent state.
• Minimal latency

Disadvantages:

• Frame loss

Typical Use Cases:

Use LatestImageOnly when you need current information, frame drops are acceptable, and latency must be kept to a minimum. Examples:

• Live display (GUI)
• Operator monitoring
• Real-time visualization

OneByOne#

Every frame is queued and processed in order. No frames are dropped intentionally. This visualization illustrates this:

Camera:    F1 → F2 → F3 → F4
App reads: F1 → F2 → F3 → F4

Advantages:

• No frame loss (ideal for analysis/recording)
• Deterministic processing order

Disadvantages:

• Higher latency if processing is slow
• Risk of buffer overflow with sustained overload

Typical Use Cases:

Use OneByOne when every frame matters, results must be reproducible, and processing speed is roughly the same as acquisition speed. Examples:

• Inspection systems
• Image recording
• Measurement tasks

Buffer Configuration#

Buffers act as a temporary reservoir between camera and application. You can adjust the number of buffers used as required.

camera.MaxNumBuffer.Value = 20

Increasing the number of buffers has the following advantages:

• Temporary slowdowns are tolerated.

The disadvantages are:

• Memory usage increases.
• Latency increases, i.e., the queue grows.

Key Takeaways#

There is no "best" strategy. The choice depends on whether you need the latest image always or all images. The following table may serve as guidance for choosing the right strategy.