The frame rate and exposure of an industrial camera are core parameters that inherently have a mutually restrictive relationship—both share the camera’s “time resources.” The central goal of balancing these two parameters is to ensure that the frame rate can keep pace with the motion speed of the detection target (without missing any details or producing motion blur) while also guaranteeing that the exposure time provides the sensor with sufficient light (allowing clear images with full detail). An imbalance will directly lead to detection failure—either “capturing the image but lacking clarity” or “achieving clear images but failing to capture the target at all.” Therefore, it’s essential to combine theoretical principles, specific application scenarios, and practical operational methods to achieve dynamic adaptation.

I. First, let’s clarify the core issue: the “time battle” between frame rate and exposure.

To achieve balance, we must first clarify the definitions and inherent conflicts between the two, thereby avoiding blind parameter tuning.

1. Core Definition (Each Has Its Own Role)

Frame rate (fps)The number of images captured per second—its core function is to “keep up with the target.” For example, when parts are being conveyed along a production line, insufficient frame rate can lead to missed captures or cause images to appear “broken” or “ghosted” due to excessively high motion speeds. Particularly in high-speed inspection scenarios, the frame rate directly determines the completeness of the inspection.

Exposure time (μs/ms)The exposure time—the duration for which a camera sensor receives light—is crucial for capturing fine details. The longer the exposure time, the more light enters the sensor, resulting in a brighter image and sharper contrast between defects (such as scratches or missed coatings). Conversely, an excessively short exposure can lead to a dark image, increased noise, and the obscuring of fine details..

2. Core Conflict (Time-Limited)

For each frame captured by the camera, the total time required is equal to the exposure time plus the data transmission time plus the sensor readout time. This total time must be less than or equal to 1 divided by the frame rate (i.e., the interval between consecutive frames); otherwise, stuttering or frame drops will occur. Simply put: the higher the frame rate, the shorter the available time per frame, and thus the exposure time becomes more constrained. Conversely, the longer the exposure time, the harder it is to increase the frame rate—these two factors exhibit a trade-off relationship where one increases as the other decreases.

Here’s a straightforward example: If you set the frame rate to 30 fps, the interval between frames is about 33 ms. If you set the exposure time to 20 ms, you’ll have only 13 ms left for data transmission and readout. If you forcibly increase the exposure time to 30 ms, the total time available for transmission and readout shrinks to just 3 ms. Once the interface bandwidth falls short of this requirement, you’ll encounter issues such as frame drops and image stuttering.

II. Three Common Consequences of Imbalance (Avoiding Misconceptions in Hyperparameter Tuning)

The imbalance between frame rate and exposure directly affects detection performance; 90% of industrial vision defects are related to this issue, which can be broadly categorized into three types.:

Frame rate too high, exposure too short.It manifests as “images that can be captured but are hard to see clearly.” The overall image appears dark, with defect details becoming blurred (e.g., scratches turn into faint gray, and impurities become difficult to distinguish from the background), or even completely black. This phenomenon is commonly observed when high-speed production lines forcibly push the frame rate to its maximum—such as 500 fps—reducing the exposure time to below 10 μs, leaving the sensor insufficient time to collect enough light. For example, in battery electrode sheet inspection, at a frame rate of 200 fps and an exposure time of 3 μs, the grayscale difference between missed coating areas and normal areas significantly diminishes, causing the algorithm’s misjudgment rate to rise sharply.

Overexposure, too low frame rateIt manifests as “clear visibility but inability to capture effectively.” The image details are sharp and clear, yet there may be missed captures (some parts fail to be recorded) or severe motion blur (caused by the movement of parts, resulting in a blurred image). This issue is commonly encountered in low-light environments (such as dark-field inspection). When the exposure time is extended—for instance, to 50 milliseconds—the frame rate is inevitably reduced to below 20 fps, making it impossible to keep up with the production line’s speed. For example, in automotive part inspection, although a 30-millisecond exposure can highlight recessed shadows, insufficient frame rate will lead to an increased rate of missed captures.

Ignore transmission/readout timeThis manifests as “theoretical compliance, but actual stuttering.” Although the frame rate and exposure time calculations appear to satisfy the condition “total elapsed time ≤ frame interval,” the omission of data transmission time—such as the 8 ms required to transmit a 20-megapixel image over a GigE interface—causes the total elapsed time to exceed the limit, resulting in frequent frame drops. For example, with a frame rate of 15 fps (frame interval of 66 ms) and an exposure time of 30 ms, it might seem that there are 36 ms remaining for data transmission. However, transmitting a high-pixel image actually takes 40 ms, ultimately leading to one frame being dropped every 10 frames.

III. The 3-Step Balancing Method (Practical and Easy to Implement—Simply Apply It Directly)

The core logic behind balancing is “first set the bottom line, then adjust the upper limit, and finally address the shortcoming.” No complicated calculations are needed—simply follow the steps to achieve a balanced outcome.

Step 1: Set a “minimum frame rate”—prioritize ensuring that nothing is missed.

The core function of frame rate is to capture the motion of the target. First, the “minimum frame rate” must be calculated based on the detection scenario, and then a safety margin should be reserved to prevent missed frames.

Minimum frame rate calculation formula Minimum frame rate = Pipeline speed ÷ Shortest side length of the part (applicable to area-array cameras); for line-scan cameras, refer to the line frequency—higher line frequencies allow for faster motion speeds.

Redundancy settings On this basis, reserve a safety margin of 20% to 30% to prevent missed captures due to transmission delays. For example: If the conveyor speed is 0.8 m/s and the shortest side length of the part is 0.08 m, the minimum frame rate would be 0.8/0.08 = 10 fps. After accounting for the reserved margin, set the frame rate to 12–13 fps. In high-speed scenarios (>2 m/s), a safety margin of more than 30% should be reserved; it is recommended to prioritize cameras with global shutter (to avoid motion blur). In low-speed scenarios (<0.5 m/s), the safety margin can be appropriately increased to allow more time for exposure.

Step 2: Calculate the “exposure ceiling”—while ensuring that it’s “clearly visible.”

The exposure time must not exceed “frame interval – transmission time – redundancy time.” It needs to be adjusted in conjunction with lighting conditions to avoid overexposure or underexposure.

Maximum Exposure Time Calculation FormulaMaximum exposure time = 1/Frame rate – Transmission time – 5% redundancy (the transmission time can be referenced from the camera interface specifications; for example, transmitting a 20-megapixel image via USB3.0 takes approximately 5 ms, while via GigE interface it takes about 8 ms).

Actual Adjustment PrinciplesFor bright-field inspection (e.g., PCB circuit inspection), set the exposure time to 70%-80% of the maximum exposure time to avoid overexposure and subsequent loss of detail. For dark-field inspection (e.g., metal scratch detection), the exposure time can be close to the maximum exposure time (but not exceeding 90%) to ensure sufficient light for highlighting defects. If the image appears too dark, prioritize adjusting the fill light rather than blindly extending the exposure time (to avoid compromising the frame rate).

Step 3: Address “hardware shortcomings”—resolve irreconcilable conflicts.

If, after tuning according to the first two steps, “underexposure” or “insufficient frame rate” still occurs, it indicates that there is a hardware bottleneck. To overcome this limitation, hardware optimization is required, and the specific solutions are outlined in the table below:

IV. Case Studies for Adapting to Different Scenarios (Just Copy the Homework)

Based on actual industrial scenarios, provide specific parameter tuning solutions tailored to different inspection requirements:

Case 1: High-speed battery electrode sheet inspection (speed: 3 m/s, defect size: 0.1 mm)

Fixed frame rate: The minimum frame rate = 3 m/s ÷ 0.1 m = 30 fps. To account for a 30% margin of safety in high-speed scenarios, set the frame rate to 40 fps.

Exposure calculation: Use a USB 3.0 interface (transmitting a 20-megapixel image takes 5 ms). Frame interval = 1/40 ≈ 25 ms; maximum exposure time = 25 - 5 - 1.25 ≈ 18.75 ms; set the actual exposure time to 15 ms.

Addressing Weaknesses: We’ve increased the brightness of the bar-shaped light source (doubling its luminance) to ensure that 15ms exposure times can clearly capture any missed coating defects, ultimately achieving zero missed detections and zero blurring.

Case 2: Detection of Metal Dents in Low-Light Conditions (speed: 0.6 m/s; shadows must be highlighted)

Fixed frame rate: Minimum frame rate = 0.6 m/s ÷ 0.05 m (shortest side length of the part) = 12 fps; with a 20% margin reserved, set it to 15 fps.

Exposure calculation: Select the GigE interface (transmitting a 10-megapixel image takes 6 ms). Frame interval = 1/15 ≈ 66 ms. Maximum exposure time = 66 - 6 - 3.3 ≈ 56.7 ms. Set the actual exposure time to 50 ms (close to the upper limit to ensure sufficient light intake).

Addressing Weaknesses: Add a dark-field light source to reduce interference from ambient stray light, and simultaneously select a low-noise sensor to prevent noise levels from rising due to long exposures.

V. Pitfall Guide and Balancing Tips

The Most Common Parameter-Tuning Mistakes

Blindly pursuing high frame rates while neglecting exposure time results in dark images where defects cannot be identified.

Simply prolonging the exposure time without considering the transfer/readout time leads to stuttering and frame loss.

Ignore fill-light optimization and rely solely on extending the exposure, at the expense of frame rate.

Confusing global shutter with rolling shutter—using a rolling shutter for high-speed scenes can lead to image distortion and motion blur (since a rolling shutter exposes the image line by line, fast-moving objects are prone to deformation; in contrast, a global shutter uses simultaneous exposure, making it better suited for high-speed scenarios).

Balance Mnemonic (Quick Memory)

Set the lower limit first for the frame rate, then calculate the upper limit for exposure; prioritize fill light over extension, and promptly address hardware shortcomings; use a global shutter for high-speed applications, and a rolling shutter for low-speed ones; fine-tune based on actual testing to ensure optimal results—only when there’s no blurring or leakage can it meet the standard.

VI. Summary

The balance between frame rate and exposure in industrial cameras essentially comes down to the "rational allocation of time resources." The key is not to maximize both parameters simultaneously, but rather to tailor them to specific inspection requirements: In high-speed scenarios, prioritize maintaining a high frame rate first, then compensate for insufficient exposure through supplementary lighting and hardware optimization; in precision inspection scenarios, prioritize ensuring adequate exposure first, then adjust the frame rate and optimize data transmission to guarantee clear details. In actual parameter tuning, it’s crucial to combine formula-based calculations, hardware optimization, and on-site measurements to achieve detection results that are “not blurry, not missed, and not stuttering,” enabling industrial cameras to truly function as “electronic eyes.”