1. Introduction
Spray uniformity of injector nozzles plays a critical role in ensuring stable combustion, improving fuel efficiency, and reducing exhaust emissions in modern internal combustion engines. Non-uniform spray distribution may lead to incomplete combustion, local over-rich zones, increased soot formation, and higher fuel consumption. Therefore, accurate evaluation of spray uniformity is essential for injector performance assessment and quality control. Traditional evaluation methods mainly rely on weighing methods or empirical observation, which are time-consuming and lack spatial resolution. With the rapid development of machine vision and image processing technology, image recognition has become an effective tool for the quantitative analysis of injector spray uniformity.
This study proposes an image-based recognition and quantitative evaluation method for injector nozzle spray uniformity, integrating high-speed imaging, digital image processing, and statistical analysis to achieve accurate and repeatable assessment of spray distribution characteristics.
2. Experimental System and Image Acquisition
A high-speed spray visualization test system is established, consisting of a common-rail fuel supply unit, injector driving module, constant-volume spray chamber, high-speed camera, pulsed LED light source, and synchronization trigger unit. The injector is mounted in the spray chamber, and fuel is injected under controlled rail pressure and injection duration. The high-speed camera captures the transient spray process with a frame rate above 20,000 frames per second to ensure sufficient temporal resolution.
The original spray images are stored as grayscale sequences. To reduce interference from background noise and lighting non-uniformity, background subtraction and grayscale normalization are applied before further processing.
3. Image Recognition and Processing Method
The spray region is extracted using adaptive threshold segmentation based on grayscale intensity. Morphological operations such as erosion and dilation are applied to eliminate isolated noise pixels and refine spray boundaries. After segmentation, the spray image is divided into multiple sectors according to the nozzle hole distribution and spray cone geometry.
Key image features, including spray area, grayscale intensity distribution, penetration length, and sector-wise fuel distribution, are extracted. The grayscale value is treated as an indirect indicator of local fuel concentration. By integrating grayscale intensity over each sector, the relative fuel distribution is quantified.
To improve recognition accuracy, edge detection algorithms and contour fitting are used to identify spray boundaries precisely. The spray cone angle and radial diffusion characteristics are then calculated based on the extracted contours.
4. Quantitative Evaluation of Spray Uniformity
Spray uniformity is quantitatively evaluated using statistical indicators derived from image feature parameters. A spray uniformity index (SUI) is defined based on the coefficient of variation of sector-wise fuel distribution:
SUI=1−μσ
where
σ is the standard deviation and
μ is the mean grayscale-integrated intensity of all sectors. A higher SUI indicates better spray uniformity.
In addition, the radial uniformity of spray penetration is evaluated by analyzing the variance of penetration lengths in different angular directions. The spatial distribution entropy is also introduced to characterize the dispersion degree of spray fuel.
By combining these indicators, a comprehensive spray uniformity evaluation system is established, enabling both axial and radial distribution assessment.
5. Results and Discussion
Experimental results show that injector nozzles with good manufacturing consistency exhibit high grayscale symmetry and small sector-wise intensity deviation. The calculated SUI under stable operating conditions remains above 0.95, indicating excellent spray uniformity. In contrast, partially blocked or worn nozzles show明显 non-uniform spray patterns, with SUI dropping below 0.85, accompanied by明显 directional bias in spray penetration.
The influence of injection pressure and pulse width on spray uniformity is also analyzed. As rail pressure increases, the atomization quality and radial diffusion improve, leading to higher uniformity. However, excessively short injection duration causes increased non-uniformity due to incomplete needle lift and unstable initial flow.
Comparative experiments demonstrate that the proposed image-based method is highly sensitive to small structural defects and early-stage clogging of nozzle holes. Compared with conventional weighing-based flow distribution methods, the proposed technique provides higher spatial resolution and better real-time visualization capability.
6. Conclusions
An image recognition and quantitative evaluation method for injector nozzle spray uniformity is proposed and experimentally validated. By integrating high-speed imaging with advanced image processing algorithms, the transient spray morphology and spatial fuel distribution can be accurately extracted. A set of uniformity evaluation indices, including sector-wise distribution uniformity and radial penetration consistency, is established to achieve comprehensive quantitative assessment.
The experimental results confirm that the proposed method has high accuracy, good repeatability, and strong sensitivity to nozzle defects and operating condition variations. This technology provides a powerful tool for injector nozzle performance evaluation, quality inspection, and spray optimization, and shows broad application prospects in fuel injection system development and engine calibration.
