簡易檢索 / 詳目顯示

回結果列表
研究生： 毛致元
Mao, Chih-Yuan
論文名稱： 結合 CNN-Transformer 雙路徑與頻域去噪之豬隻超音波妊娠辨識輕量化模型
A Lightweight Hybrid CNN-Transformer Architecture with Frequency-Domain Denoising for Sow Pregnancy Classification in Ultrasound Imaging
指導教授： 劉任修
Liu, Ren-Shiou
學位類別： 碩士
Master
系所名稱： 管理學院 - 資訊管理研究所
Institute of Information Management
論文出版年： 2026
畢業學年度： 114
語文別： 中文
論文頁數： 72
中文關鍵詞： SqueezeNet 、Vision Transformer 、豬隻妊娠辨識 、超音波影像 、多尺度特徵融合 、離散餘弦轉換
外文關鍵詞： Sow Pregnancy Detection, Ultrasound Image, SqueezeNet, Vision Transformer, Discrete Cosine Transform, Multi-scale Feature Fusion
相關次數： 點閱：5  下載：0 
分享至:
查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報

    • 隨著全球人口增長與開發中國家肉品需求持續攀升，豬肉作為增量主力之一，養殖產業亟需擴大生產規模並提升繁殖管理效率。在此背景下，豬隻妊娠辨識是影響繁殖效率與經濟效益的關鍵技術。然而，產業同時面臨技術人力斷層與生物安全風險的雙重挑戰：傳統妊娠檢測高度仰賴資深獸醫與技術人員的經驗判讀，不僅培訓成本高昂，且在獸醫人力短缺的情況下，頻繁依賴外聘技術人員跨場服務，更增加場際間交叉感染的防疫風險。因此，發展一套可由非專業人員操作、具備高準確率且計算成本低的智慧化妊娠辨識系統，對於提升養豬產業的生產效率與永續發展具有重要意義。

    本研究提出一種結合 SqueezeNet 與 Vision Transformer 的深度學習模型 DFTNet，專門針對低頻超音波影像進行豬隻妊娠辨識。其中利用單向特徵注入單元結合局部紋理與全域空間關聯，實現跨架構特徵融合。此外，針對超音波影像固有的散斑噪聲特性，本研究設計頻紋融合模組，整合 DCT 於特徵層進行頻域去噪，並透過保留組織邊界微結構等資訊，結合自適應加權融合機制動態平衡去噪與特徵保留。

    實驗結果顯示，DFTNet 在豬隻妊娠超音波影像分類任務中具有良好表現，準確率達 95.54%，並在維持 59.10M 參數量與 4.35 ms 推論延遲之條件下，有效應對低頻超音波影像中邊界模糊與紋理等挑戰。結果顯示，本研究方法能在實務場域限制下，輔助非專業人員進行妊娠診斷，為養豬產業智慧化管理提供可參考的技術方案。

    As global meat demand rises and sow farms face skilled workforce shortages and inter-farm bio-security risks, automated pregnancy detection systems operable by non-specialists are essential for sustainable sow farm management. This study formulates sow pregnancy assessment from low-frequency ultrasound as binary image classification between Pregnant and Non-Pregnant. To address speckle noise, boundary ambiguity, and texture similarity inherent to low-frequency ultrasound, we propose DFTNet, a lightweight SqueezeNet-ViT hybrid with FIU for cross-architecture feature fusion and FTFM integrating DCT-based denoising with a texture branch and adaptive weighting. On 10,500 labeled images from 42 sows, DFTNet achieved $95.54%$ accuracy and $95.55%$ F1-score under an $80{:}20$ sow-identity split, with $4.35$ ms inference latency and a $224.73$ MB model size. These results show that DFTNet balances recognition performance and computational efficiency for on-farm sow pregnancy monitoring.

    摘要 i EXTENDED ABSTRACT ii 誌謝 x 目錄 xi 表目錄 xiv 圖目錄 xv 1 緒論 1 1.1 背景及動機 1 1.2 研究目的 4 1.3 研究貢獻 4 1.4 論文架構 5 2 相關文獻探討 6 2.1 妊娠辨識 6 2.1.1 傳統妊娠辨識方法 6 2.1.2 機器學習用於妊娠辨識 8 2.1.3 深度學習用於妊娠辨識 9 2.2 多尺度融合機制 10 2.3 離散餘弦轉換 12 2.4 小結 13 3 研究方法 15 3.1 SqueezeNet 與 Vision Transformer 17 3.1.1 SqueezeNet 架構 18 3.1.2 ViT 架構 19 3.2 雙路徑並行架構 20 3.2.1 特徵注入單元 21 3.2.2 多尺度融合機制 23 3.3 頻紋融合模組 24 3.3.1 DCT 分支 24 3.3.2 紋理分支 26 3.3.3 Adaptive Fusion Unit 27 3.4 模型訓練 29 4 實驗與分析 32 4.1 實驗流程 32 4.2 實驗資料集概述 33 4.3 實驗之 Base Model 概述 34 4.4 實驗環境與參數設定 35 4.5 實驗評估指標 35 4.6 實驗結果與分析 37 4.6.1 實驗一：雙路徑架構驗證 37 4.6.2 實驗二：頻紋融合模組組件分析 38 4.6.3 實驗三：去噪處理時機比較（Input-level vs. Feature-level） 39 4.6.4 實驗四：多尺度特徵融合驗證 40 4.6.5 實驗五：DCT 低頻保留比例分析 42 4.6.6 實驗六：與基準模型全面比較 43 5 結論與未來發展 46 參考文獻 47

    Ahmed, N., Natarajan, T., and Rao, K. R. (1974). Discrete cosine transform. IEEE Transactions on Computers, C-23(1):90–93.

    Alexandratos, N. and Bruinsma, J. (2012). World agriculture towards 2030/2050: the 2012 revision. ESA Working Paper No. 12-03, Food and Agriculture Organization of the United Nations (FAO), Rome.

    Almond, G. W. and Dial, G. D. (1986). Pregnancy diagnosis in swine: a comparison of the accuracies of mechanical and endocrine tests with return to estrus. J Am Vet Med Assoc, 189(12):1567–1571.

    Andrade, A. C., da Silva, L. O., Souza, V. F., de Almeida Rufino, L. M., da Silva, T. E., dos Santos, A. D. F., de Azevedo Gomes, D., and Rodrigues, J. P. P. (2023). Identifying pregnancy in cows using ovarian ultrasound images and convolutional neural networks-a proof-of-concept study. Computers and Electronics in Agriculture, 206:107674.

    Barukab, O., Ali, F., Alghamdi, W., Bassam, Y., and Khan, S. A. (2022). Dbp-cnn: Deep learning-based prediction of dna-binding proteins by coupling discrete cosine transform with two-dimensional convolutional neural network. Expert Systems with Applications, 197:116729.

    Basak, J. K., Paudel, B., Deb, N. C., Kang, D. Y., Moon, B. E., Shahriar, S., and Kim, H. T. (2023). Prediction of body composition in growing-finishing pigs using ultrasound based back-fat depth approach and machine learning algorithms. Computers and Electronics in Agriculture, 213:108269.

    Burgos-Artizzu, X. P., Coronado-Gutiérrez, D., Valenzuela-Alcaraz, B., Bonet-Carne, E., Eixarch, E., Crispi, F., and Gratacós, E. (2020). Evaluation of deep convolutional neural networks for automatic classification of common maternal fetal ultrasound planes. Scientific Reports, 10(1):10200.

    Burt, P. J. and Adelson, E. H. (1987). The laplacian pyramid as a compact image code. In Readings in computer vision, pages 671–679. Elsevier.

    Chae, J. W., Choi, Y. H., Lee, J. N., Park, H. J., Jeong, Y. D., Cho, E. S., Kim, Y. S., Kim, T. K., Sa, S. J., and Cho, H. C. (2023). An intelligent method for pregnancy diagnosis in breeding sows according to ultrasonography algorithms. Journal of Animal Science and Technology, 65(2):365.

    Chen, C. J. and Liu, R.-S. (2026). A multi-scale attention network for steel surface defect recognition. Soft Computing, pages 1–17.

    Choi, H., Kiesenhofer, E., Gantner, H., Hois, J., and Bamberg, E. (1987). Pregnancy diagnosis in sows by estimation of oestrogens in blood, urine or faeces. Animal Reproduction Science, 15(3):209–216.

    Chollet, F. (2017). Xception: Deep learning with depthwise separable convolutions. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 1251–1258.

    Cubuk, E. D., Zoph, B., Mane, D., Vasudevan, V., and Le, Q. V. (2018). Autoaugment: Learning augmentation policies from data. arXiv preprint arXiv:1805.09501.

    Delgado, C. L., Rosegrant, M. W., Steinfeld, H., Ehui, S. K., and Courbois, C. (1999). Livestock to 2020: The Next Food Revolution. Number 28 in Food, Agriculture, and the Environment Discussion Paper. International Food Policy Research Institute (IFPRI), Washington, D.C.

    Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., and Fei-Fei, L. (2009). Imagenet: A large-scale hierarchical image database. In 2009 IEEE Conference on Computer Vision and Pattern Recognition, pages 248–255. IEEE.

    Dixon, L. K., Sun, H., and Roberts, H. (2019). African swine fever. Antiviral research, 165:34–41.

    Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., Dehghani, M., Minderer, M., Heigold, G., Gelly, S., et al. (2020). An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929.

    Einarsson, S., Madej, A., and Tsuma, V. T. (1996). The influence of stress on early pregnancy in the pig. Animal Reproduction Science, 42(1-4):165–172.

    Engvall, E. and Perlmann, P. (1971). Enzyme-linked immunosorbent assay (elisa) quantitative assay of immunoglobulin g. Immunochemistry, 8(9):871–874.

    Feng, Y., Gao, R., Yang, W., Lu, H., Sun, W., Zhang, Y., Ren, Y., Gao, L., Li, M., Li, Q., et al. (2026). Study on early pregnancy diagnosis of sows based on body fluid metabolite detection combined with machine learning models. Veterinary Sciences, 13(5):409.

    Feng, Y., Gao, R., Zhang, Y., Yang, W., Li, M., Li, Q., Pu, G., Zhang, Y., Ren, Y., Ai, Z., et al. (2025). Study on early pregnancy diagnosis in sows based on joint analysis of vaginal secretion metabolomics and machine learning. bioRxiv, pages 2025–09.

    Gonzalez, R. C. and Woods, R. E. (2018). Digital Image Processing. Pearson, 4th edition.

    Graves, A. (2012). Long short-term memory. Supervised sequence labelling with recurrent neural networks, pages 37–45.

    Hu, J., Shen, L., and Sun, G. (2018). Squeeze-and-excitation networks. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 7132–7141.

    Huang, Q., Zhang, F., and Li, X. (2018). Machine learning in ultrasound computer-aided diagnostic systems: a survey. BioMed research international, 2018(1):5137904.

    Iandola, F. N., Han, S., Moskewicz, M. W., Ashraf, K., Dally, W. J., and Keutzer, K. (2016). Squeezenet: Alexnet-level accuracy with 50x fewer parameters and 0.5 mb model size. arXiv preprint arXiv:1602.07360.

    Jaskowski, J., Zduńczyk, S., Zielińska-Górska, M., and Czarnik, U. (2024). Non-invasive methods for diagnosing pregnancy in cows and their real value. Polish Journal of Veterinary Sciences, 27(1):25–34.

    Kaufman, S., Rosset, S., Perlich, C., and Stitelman, O. (2012). Leakage in data mining: Formulation, detection, and avoidance. ACM Transactions on Knowledge Discovery from Data (TKDD), 6(4):1–21.

    Kim, T. K., Choi, Y. H., Hong, J. S., Park, H. J., Kim, Y. M., Kim, J. E., Lee, J. H., Sa, S. J., Jeong, Y. D., Kim, J. S., and Cho, H. c. (2025). Deep learning-enhanced diagnosis of sow pregnancy through low-frequency ultrasound imaging. Animals, 15:318.

    Kingma, D. P. and Ba, J. (2014). Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980.

    Knox, R. and Flowers, W. (2008). Using real-time ultrasound for pregnancy diagnosis in swine. FactSheet. Pork Information Gateway. North Carolina State University, Raleigh, NC, USA.

    Koketsu, Y., Tani, S., and Iida, R. (2017). Factors for improving reproductive performance of sows and herd productivity in commercial breeding herds. Porcine health management, 3(1):1.

    Krishnan, K. B., Farrokhi, Y., and Roy, R. (2020). Design of a dct/idct filter structure and its application for ultrasound image filtering and reconstruction.

    Li, X., Navab, N., and Jiang, Z. (2025). Speckle2self: Self-supervised ultrasound speckle reduction without clean data. Medical Image Analysis, page 103755.

    Lin, T.-Y., Dollár, P., Girshick, R., He, K., Hariharan, B., and Belongie, S. (2017). Feature pyramid networks for object detection. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 2117–2125.

    Litjens, G., Kooi, T., Bejnordi, B. E., Setio, A. A. A., Ciompi, F., Ghafoorian, M., Van Der Laak, J. A., Van Ginneken, B., and Sánchez, C. I. (2017). A survey on deep learning in medical image analysis. Medical image analysis, 42:60–88.

    Liu, S., Huang, D., and Wang, Y. (2019). Learning spatial fusion for single-shot object detection. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 3658–3667.

    Liu, S., Qi, L., Qin, H., Shi, J., and Jia, J. (2018). Path aggregation network for instance segmentation. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 8759–8768.

    Liu, Z., Mao, H., Wu, C.-Y., Feichtenhofer, C., Darrell, T., and Xie, S. (2022). A convnet for the 2020s. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 11976–11986.

    Loshchilov, I. and Hutter, F. (2017). Sgdr: Stochastic gradient descent with warm restarts. In International Conference on Learning Representations.

    Lou, A., Guan, S., and Loew, M. (2021). Dc-unet: rethinking the u-net architecture with dual channel efficient cnn for medical image segmentation. In Medical Imaging 2021: Image Processing, volume 11596, pages 758–768. SPIE.

    Michailovich, O. and Tannenbaum, A. (2006). Despeckling of medical ultrasound images. IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, 53(1):64–78.

    Niederwerder, M. C. and Hesse, R. A. (2018). Swine enteric coronavirus disease: a review of 4 years with porcine epidemic diarrhoea virus and porcine deltacoronavirus in the united states and canada. Transboundary and Emerging Diseases, 65(S1):660–675.

    Peng, Z., Huang, W., Gu, S., Xie, L., Wang, Y., Jiao, J., and Ye, Q. (2021). Conformer: Local features coupling global representations for visual recognition. In Proceedings of the IEEE/CVF international conference on computer vision, pages 367–376.

    Pham, T. M. H. et al. (2024). Characterizing veterinary service needs in vietnam: A mixed-methods study of animal health workers’ perspectives in thai nguyen province. PLOS Global Public Health, 4(2):e0000485.

    Pork Information Gateway (2021). Estrus or heat detection. [https://porkgateway.org/resource/estrus-or-heat-detection/](https://porkgateway.org/resource/estrus-or-heat-detection/). Accessed: 2025-11-27.

    Rao, K. R. and Yip, P. (2014). Discrete cosine transform: algorithms, advantages, applications. Academic press.

    Rose, D. C. and Chilvers, J. (2018). Agriculture 4.0: Broadening responsible innovation in an era of smart farming. Frontiers in Sustainable Food Systems, 2:87.

    Singla, B., Jain, A. K., Tiwari, R. G., Kukreja, V., and Sharma, V. (2023). Classification model using cnn and lstm for cow pregnancy. In 2023 10th International Conference on Signal Processing and Integrated Networks (SPIN), pages 457–461. IEEE.

    Svoboda, M., Nemeckova, M., Medkova, D., Sardi, L., and Hodkovicova, N. (2024). Non-invasive methods for analysing pig welfare biomarkers. Vet Med (Praha), 69(5):137–155.

    Szegedy, C., Ioffe, S., Vanhoucke, V., and Alemi, A. (2017). Inception-v4, inception-resnet and the impact of residual connections on learning. In Proceedings of the AAAI conference on artificial intelligence, volume 31.

    Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., Erhan, D., Vanhoucke, V., and Rabinovich, A. (2015). Going deeper with convolutions. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

    Tan, M. and Le, Q. (2021). Efficientnetv2: Smaller models and faster training. In International conference on machine learning, pages 10096–10106. PMLR.

    Tan, M., Pang, R., and Le, Q. V. (2020). Efficientdet: Scalable and efficient object detection. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 10781–A survey on deep learning in medical image analysis.–10790.

    USDA Foreign Agricultural Service (2026). Production, supply and distribution: Pork. [https://fas.usda.gov/data/production/commodity/0113000](https://fas.usda.gov/data/production/commodity/0113000). Accessed: 2026-05-04.

    Vafaeezadeh, M., Behnam, H., and Gifani, P. (2024). Ultrasound image analysis with vision transformers. Diagnostics, 14(5):542.

    Vallet, J. L., Miles, J. R., Rempel, L. A., Lents, C. A., and Freking, B. A. (2021). Nutrition and conceptus development: The pig as a model. Animal Frontiers, 11(3):61–70.

    Wallace, G. K. (1991). The jpeg still picture compression standard. Communications of the ACM, 34(4):30–44.

    Williams, S. I., Piñeyro, P., and de la Sota, R. L. (2008). Accuracy of pregnancy diagnosis in swine by ultrasonography. Canadian Veterinary Journal, 49(3):269–273.

    Woo, S., Park, J., Lee, J.-Y., and Kweon, I. S. (2018). Cbam: Convolutional block attention module. In Proceedings of the European Conference on Computer Vision (ECCV), pages 3–19.

    Wróbel, M. and Szymczyk, A. (2023). The transfer of agricultural knowledge in the context of sustainability: A literature review. Sustainability, 15(4):3226.

    Xiao, Y., Zhuang, Y., Ling, W., Jiang, S., Chen, K., Liao, G., Xie, Y., Hou, Y., Han, L., Hua, Z., et al. (2025). Lgf-net: A multi-scale feature fusion network for thyroid nodule ultrasound image classification. Journal of Applied Clinical Medical Physics, 26(8):e70149.

    Xu, K., Qin, M., Sun, F., Wang, Y., Chen, Y.-R., and Ren, F. (2020). Learning in the frequency domain. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 1740–1749.

    Yalow, R. S. and Berson, S. A. (1960). Immunoassay of endogenous plasma insulin in man. The Journal of clinical investigation, 39(7):1157–1175.

    Zoli, A. P., Guilbault, L. A., Delahaut, P., Ortiz, W. B., and Beckers, J.-F. (1992). Radioimmunoassay of a bovine pregnancy-associated glycoprotein in serum: its application for pregnancy diagnosis. Biology of Reproduction, 46(1):83–92.

    QR CODE