Long-term accumulation of carbon deposits in automobile engines can easily accelerate the ageing of automobiles. Timely detection and cleaning can effectively prolong the service life of automobiles. In this paper, a carbon deposits degree recognition method based on a visual image is proposed to automatically recognitze the degree of carbon deposits and provide guidance for carbon deposits cleanup. Firstly, to address the issue of small data volume and uneven category distribution in carbon deposits images dataset, data preprocessing is conducted. Secondly, aimed at the wide range of feature distribution and fine-grained characteristics of carbon images, a feature resampling module is designed to improve the feature expression from both spatial and channel directions. Finally, a lightweight carbon deposits degree recognition model is designed to facilitate the detection of deployed. The experimental results demonstrate that compared to other methods, the method proposed in this thesis achieves the highest inference speed of 179 frames/s with a testing accuracy of 84.5%, meeting the needs of the industry.
APEAGYEIA, ADEMOLAKET E, ADOM-ASAMOAHM. Evaluation of deep learning models for classification of asphalt pavement distresses[J]. International Journal of Pavement Engineering, 2023, 24(1): 2180641.
LIANGMeiyan, ZHANGQiannan, RENZhuyun, et al. Research on identification of colon pathology image based on attention mechanism[J]. Journal of Test and Measurement Technology, 2022, 36(2): 93-100. (in Chinese)
[10]
SHARMAV, TRIPATHIA K, MITTALH. DLMC-Net: deeper lightweight multi-class classification model for plant leaf disease detection[J]. Ecological Informatics, 2023, 75: 102025.
[11]
FANS, WANGY, CAOS, et al. A deep residual neural network identification method for uneven dust accumulation on photovoltaic(PV) panels[J]. Energy, 2022, 239: 122302.
[12]
HEK, ZHANGX, RENS, et al. Deep residual learning for image recognition[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016: 770-778.
[13]
MEHTAS, RASTEGARIM, SHAPIROL, et al. Espnetv2: a light-weight, power efficient, and general purpose convolutional neural network[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019: 9182-9192.
[14]
QINZ, ZHANGP, WUF, et al. FcaNet: frequency channel attention networks[C]// Proceedings of the IEEE/CVF International Conference on Computer Vision, Virtual, 2021: 763-772.
[15]
HUJ, SHENL, SUNG. Squeeze-and-excitation networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018: 7132-7141.
[16]
WOO S, PARKJ, LEEJ Y, et al. CBAM: convolutional block attention module[C]// Proceedings of the European Conference on Computer Vision (ECCV). Cham: Springer, 2018: 3-19.
[17]
LEEH J, KIMH E, NAM H. SRM: a style-based recalibration module for convolutional neural networks[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision, Virtual, 2019: 1854-1862.
[18]
SIMONYANK, ZISSERMANA. Very deep convolutional networks for large-scale image recognition[EB/OL]. Computer Science, 2014. DOI:10.48550/arXiv.1409.1556 .
[19]
HUANGG, LIUZ, VAN DER MAATENL, et al. Densely connected convolutional networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017: 2261-2269.
[20]
TANM, LEQ V. EfficientNet: rethinking model scaling for convolutional neural networks[EB/OL].
[21]
DOSOVITSKIYA, BEYERL, KOLESNIKOVA, et al. An image is worth 16x16 words: transformers for image recognition at scale[EB/OL].