Why Image Acquisition Boards Are Essential in Machine Vision Systems

In new-energy battery inspection, smartphone display defect detection, automotive component measurement, and many other industrial scenarios, machine vision systems play a central role in quality control. While image acquisition boards may not “capture scenes” like cameras or “make decisions” like algorithms, they act as the system’s invisible workhorses—converting camera signals into computer-processable data and serving as the essential data bridge.

Many assume that “a camera can directly send images to a PC,” but that’s not the case. Without an acquisition board, analog camera signals appear as unreadable noise, high-speed digital streams become congested, and multi-camera systems easily fall out of sync. This article explains why acquisition boards are indispensable by examining their definition, working principles, key parameters, and application requirements.

1. Reframing the Role: The Acquisition Board as a “Translator + Controller”

When people hear “acquisition board,” they often confuse it with a “memory card.” In machine vision, however, an image acquisition board is the bidirectional communication hub between industrial cameras and computers. Its core responsibilities fall into two categories:

1. Signal “Translation”: Ensuring Seamless Communication

Industrial cameras output two types of signals, both requiring conversion:

Analog signals (traditional CCD cameras):
They output continuous electrical waveforms—similar to analog broadcasting—while computers can only process digital data. The acquisition board uses A/D conversion to transform 0–5 V signals into 8-bit or 12-bit digital values (e.g., 0–255 or 0–4095), enabling the PC to interpret the image.

Digital signals (mainstream industrial cameras):
These often use proprietary or specialized protocols such as Camera Link. The operating system cannot decode them directly. The acquisition board acts as a decoder, converting them into standardized formats such as BMP or RAW.

Example:
A bottle-cap manufacturer uses an analog CCD camera to inspect scratches. When connected directly to the PC, only voltage fluctuations appear. After adding an acquisition board, the signal is converted into 12-bit digital images, revealing 0.1 mm scratches with clarity for automated defect marking.

2. Camera “Control”: Improving Capture Accuracy

Beyond signal conversion, acquisition boards also orchestrate camera behavior:

• Trigger control: Captures images only when parts reach the inspection point—preventing missed or empty shots.

• Parameter adjustment: Remotely modifies exposure, gain, and other settings.

• Multi-camera synchronization: Critical for 3D inspection or panoramic stitching. Sync precision can reach 0.1 μs, avoiding spatial misalignment.

2. Working Principle: A 3-Step Process—Receive, Process, Transmit

The operation of an acquisition board consists of three fundamental stages:

1. Signal Reception

After power-up, the board performs protocol handshakes with the camera—confirming resolution, pixel clock, frame rate, bit depth, and initializing buffers. For a GigE camera, for example, it ensures that the receiver’s speed matches the output to prevent overflow.

2. Signal Processing

Different modules are used depending on the signal type:

• Analog signals:
  A/D converters (e.g., ADI AD9288) digitize incoming pixels, while filtering circuits suppress EMI noise to eliminate snow or false edges that could affect measurement.

• Digital signals:
  FPGA chips (e.g., Xilinx Artix series) decode the protocol, extract pixel and sync signals, cache frames, and decompress compressed data to maintain raw image integrity.

3. Data Transmission

Processed images are sent to the computer via interfaces such as PCIe or USB 3.0:

• High-speed interfaces:
  PCIe x4 offers up to 8 GB/s, supporting 20 MP at 60 fps. USB 3.0 provides ~5 GB/s for lower-speed applications.

• Buffering:
  On-board DDR3/DDR4 (1–4 GB) buffers hold 1–2 frames before burst transfer, avoiding interruptions when the PC is temporarily busy.

3. Key Parameters: Three Indicators That Determine Your Selection

Choosing an acquisition board solely by interface type often leads to misconfiguration or overspecifying. Instead, evaluate the following three criteria:

1. Number of Channels

Defines how many cameras can connect simultaneously:

• Single-channel: For individual inspection points.

• Multi-channel: Four-channel boards support panoramic PCB inspection, 3D systems, etc.

Ensure support for multi-channel synchronization to avoid inter-camera timing errors.

2. Sampling Frequency

Defines the maximum processing speed.
Formula:

Sampling Frequency ≥ Resolution × Frame Rate

Example:
A 20 MP camera at 60 fps requires at least 120 MHz. A 100 MHz board will drop frames.

Typical ranges:

• Camera Link: up to 850 MHz

• GigE: ~200 MHz
  Plan a 10–20% safety margin.

3. Resolution & Bit Depth

Determines imaging accuracy:

• The board must support the camera’s maximum resolution. Otherwise, the image will be downsampled, losing detail (critical for semiconductor pin inspection).

• A/D bit depth:

  • 8-bit (256 levels): Suitable for high-contrast scenes

  • 12-bit (4096 levels): Captures subtle brightness variations (e.g., scratches on transparent plastics)

4. Avoiding Common Mistakes: Understanding “Frame” vs. “Field”

In analog or interlaced-scan systems, confusion often arises:

• Frame: A complete image. Progressive-scan cameras output full frames—ideal for static or slow-moving objects.

• Field: Half of a frame. Interlaced cameras transmit odd and even fields separately to increase motion smoothness.

Avoid this pitfall:
For fast-moving objects, choose progressive-scan cameras + compatible acquisition boards. If using interlaced cameras, ensure the board supports field merging to avoid half-missing images.

5. Application-Based Selection

There is no universal acquisition board. Choose based on application needs:

High-precision (Semiconductor inspection)

• High resolution

• Multi-camera sync

• Camera Link/CoaXPress

• ≥200 MHz sampling

• ≥12-bit A/D

High-speed (Beverage bottle inspection)

• High frame rate

• Real-time performance

• PCIe (single or dual channel)

• ≥300 MHz sampling

• Hardware trigger support

Cost-sensitive (Education, small-scale testing)

• USB 3.0

• Single channel

• ≥100 MHz sampling

• Compatible with common USB industrial cameras

Conclusion: The Image Acquisition Board—The Hidden Foundation of Machine Vision

In a machine vision system, the camera is the “eye,” the algorithm is the “brain,” and the acquisition board is the neural backbone. Without it, even the best cameras and algorithms cannot function reliably.

Selection logic:

1. Identify application requirements (camera count, frame rate, accuracy).

2. Match the acquisition board by channel count, sampling frequency, and resolution support.

The right acquisition board unlocks the full performance of a machine vision system—delivering precise inspection and high-efficiency production, and safeguarding quality in automated manufacturing.

For project inquiries, please contact: chenguo@mindvision.com.cn