State Key Laboratory of Extreme Environment Optoelectronic Dynamic Measurement Technology and Instrument,North University of China,Taiyuan 030051,China
Vision-based measurement has been widely applied in industrial liquid level monitoring. However, in weak-texture scenarios such as liquid surfaces, existing methods often rely on floaters, markers, or structured light to enhance matching features. These approaches are unsuitable in environments like propellant tanks of rocket engines, where high purity and airtightness are strictly required. To improve the accuracy and robustness of stereo vision in weak-texture liquid surface scenarios, a confidence-guided fusion method of monocular and stereo depth estimation is proposed. The absolute depth map from stereo matching is used as the primary reference and the relative depth map predicted by a monocular neural network is also incorporated in this method. A multi-dimensional confidence map is constructed to align the monocular depth to the stereo scale, and pixel-wise weighted fusion is performed under the guidance of confidence. And the physical scale accuracy is preserved in high-confidence stereo regions, while monocular structural information is used to complement details in weak-texture regions, thereby enhancing the overall accuracy and completeness of depth estimation. An experimental platform is built to measure liquid level height, and the performance of SGBM, IGEV++, and the proposed fusion algorithm is compared under different liquid levels. Experimental results show that the proposed method achieves an average error of 1.505%, which represents a reduction of approximately 90.69% compared to the SGBM method. These results demonstrate the effectiveness of the proposed approach in improving the accuracy and stability of liquid level measurement.
ISLAMT, MAURYAO P. Design and fabrication of non-contact fringing field capacitive sensor for liquid level measurement[C]//2019 IEEE 16th India Council International Conference (INDICON), 2019: 1-4.
HUFalu. Application research of radar level meter in chemical tank area storage tank level measurement[J]. Chemical Industry Management, 2019(9): 149-150. (in Chinese)
DINGFenglin, WEIYanming. Application analysis of ultrasonic flowmeter in the spacecraft propulsion system[J]. Aerospace Control and Application, 2010, 36(6): 47-50. (in Chinese)
[11]
İNANS A, AKSOYB. Phosphorescence marking method for liquid level detection and control using computer vision in dark environments[J]. Precision Engineering, 2025, 93: 397-405.
WANGKai, LIDerui, XIANGSheng, et al. Dynamic liquid surface wave height field measurement method based on speckle projection 3D-DIC[J]. Chinese Journal of Theoretical and Applied Mechanics, 2023, 55(10): 2427-2438. (in Chinese)
ZHANGZong, WANGJianli. Design of liquid level detection and control system based on machine vision[J]. Automation & Instrumentation, 2017, 32(7): 49-51. (in Chinese)
[22]
WANGX, SUNJ, QINH, et al. Accurate unsupervised monocular depth estimation for ill-posed region[J]. Frontiers in Physics, 2023, 10: 1115764.
[23]
RANFTLR, LASINGERK, HAFNERD, et al. Towards robust monocular depth estimation: mixing datasets for zero-shot cross-dataset transfer[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022, 44(3): 1623-1637.
[24]
SUC, TANG, LUOY. Research on stereo matching technology based on binocular vision[J]. OALib, 2019, 6(9): 1-10.
[25]
ZHANGZ. Flexible camera calibration by viewing a plane from unknown orientations[C]//Proceedings of the Seventh IEEE International Conference on Computer Vision, 1999(1): 666-673.
[26]
ZHANGL, CAIF, WANGJ, et al. The SGM algorithm based on census transform for binocular stereo vision[C]//2022 International Conference on Machine Learning and Knowledge Engineering (MLKE), 2022: 50-54.
[27]
ZAMA RAMIREZP, POGGIM, TOSIF, et al. Geometry meets semantics for semi-supervised monocular depth estimation[C]//Lecture Notes in Computer Science(LNIP), 2019, 11363: 298-313.
[28]
CHENGJ, LIUL, XUG, et al. MonSter: Marry monodepth to stereo unleashes power [C]//2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2025: 6273-6282.
[29]
XUG, WANGX, ZHANGZ, et al. IGEV++: Iterative multi-range geometry encoding volumes for stereo matching[C]//IEEE Transactions on Pattern Analysis and Machine Intelligence, 2025, 47(8): 7108-7122.