隨著物聯網 (IoT) 的發展，在具有IoT功能的攝影機上利用既有的計算資源來實現邊緣智能 (以下稱為邊緣攝影機) ，藉此提供各種以人為中心的服務。但是，環境干擾 (例如暴雨) 會嚴重降低圖像進入邊緣攝影機中的影像品質，從而影響整體圖像和服務品質。為了維持圖像穩定性，目前已經廣泛地使用複雜的深度網路做影像強化。由於邊緣攝影機的有限計算能力，所以無法有效運行複雜的神經網路。既有研究中所提出的輕量化解決方案容易減損模型的影像品質，因此本研究提出輕量級的遞迴注意力殘差強化網路，該網路將高效的卷積運算作為基本構建模組。透過在遞迴神經網路 (RNN) 訓練框架上進行模組化設計，可以根據給定邊緣攝影機的硬體規格來優化所得的神經網路。此外，為了適應實際和動態環境中部署的邊緣攝影機的各種天氣條件，我們進一步設計了一種序列化環境感知模型部署器，可以檢測連續的天氣變化以便部署最合適的影像強化模型，以減少不良干擾所造成的影響，從而盡可能保持高的影像品質。實驗結果顯示，在降雨的影像強化方面，於類似的圖像品質下，速度提高了約60%，並且影像品質沒因此降低，PSNR提升5.3%，SSIM提升2.3%。而在霧霾的影像強化方面，本研究之速度提升至少約18%，PSNR提升平均約17%，SSIM提升8.75%。在降雪的影像強化方面，本研究在PSNR以及SSIM的表現上與既有研究相當，PSNR提升約0.8%，從30.39小幅提升到30.64，速度為20 FPS，且本研究是既有研究中第一篇列出運行速度的研究。而本研究之影像強化結果在物件偵測上可提升4%的辨識準確度。最後，在動態的環境中，本研究提出序列環境感知技術，可在環境中偵測天氣的概念飄移，在過去研究的比較當中，辨識錯誤率由1%下降至0.03%。如此，可說明本研究所提出之影像遞迴強化技術在各式場景上具有應用的高度彈性，所提出之序列環境感知技術在動態環境中可更準確偵測概念飄移，並在過去研究的比較中取得更好表現。

With the development of the Internet of Things (IoT) and artificial intelligence (AI), existing computing resources are used to enable edge intelligence (hereafter referred to as edge cameras) on cameras with IoT capabilities to provide a variety of people-centric services. However, environmental interference, such as heavy rain, can significantly reduce the image quality of images entering the edge camera, affecting the overall image and service quality. In order to maintain image quality, complex deep networks have been widely used for image enhancement. Due to the restricted computing power of edge cameras, it very challenging to effectively run complex neural networks. The existing lightweight solutions are prone to degrade the image quality of an edge camera, so we incorporate recursive attention enhancement into a lightweight residual network that uses high-efficient convolution operations as a basic building block. With this modular design on a Recurrent Neural Networks (RNN), the resulting neural network can be easily optimized according to hardware specifications of a given edge camera. In addition, in order to adapt to the various weather conditions of the edge cameras deployed in real and dynamic environments, we further designed a serialized environment-aware model deployer that detects continuous weather changes and deploy the most appropriate image-enhanced model to reduce the impact of adverse interference and thus maintain as high image quality as possible. The experimental results show that in terms of image enhancement in rainy scenes, the speed is increased by about 60% under similar image quality and PSNR is increased by 5.3%. SSIM is increased by 2.3%. In terms of image enhancement in hazy scenes, the speed is increased by at least about 18%, and PSNR is increased by 17%. SSIM is increased by 8.75%. In terms of image enhancement in snowy scenes, our study is comparable to existing research in PSNR and SSIM. PSNR is increased by 0.8%, from 30.39 to 30.64, and the speed is 20 FPS. Our study is the first of the existing studies to list the running speed.
Finally, in a dynamic environment, this study proposes a sequential environment-aware method that can detect the concept shift in weather. In the comparison of latest studies, our detection error rate has dropped from 1% to 0.03%. The image-enhanced results also prove their effectiveness in ob-ject detection by at most 4% improvement in accuracy. In this way, it can be explained that the Recursive Lightweight Enhancement Network has a higher degree of flexibility in various weather scenarios, and the proposed sequential environment-aware method can more accurately detect the concept drift in a dynamic environment.

[1] S. Pouyanfar et al., "A Survey on Deep Learning: Algorithms, Techniques, and Applications," ACM Computing Surveys (CSUR), vol. 51, no. 5, p. 92, 2018.
[2] G. Ananthanarayanan et al., "Real-Time Video Analytics: The Killer App for Edge Computing," Computer, vol. 50, no. 10, pp. 58-67, 2017.
[3] C.-H. Lu and B.-E. Shao, "Environment-Aware Multiscene Image Enhancement for Internet of Things Enabled Edge Cameras," IEEE Systems Journal, 2020.
[4] X. Fu, B. Liang, Y. Huang, X. Ding, and J. Paisley, "Lightweight Pyramid Networks for Image Deraining," arXiv preprint arXiv:1805.06173, 2018.
[5] D. Ren, W. Zuo, Q. Hu, P. Zhu, and D. Meng, "Progressive image deraining networks: a better and simpler baseline," in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2019, pp. 3937-3946.
[6] L. Shen, Z. Yue, Q. Chen, F. Feng, and J. Ma, "Deep joint rain and haze removal from a single image," in 2018 24th International Conference on Pattern Recognition (ICPR), 2018, pp. 2821-2826: IEEE.
[7] X. Fu, J. Huang, X. Ding, Y. Liao, and J. Paisley, "Clearing the skies: A deep network architecture for single-image rain removal," IEEE Transactions on Image Processing, vol. 26, no. 6, pp. 2944-2956, 2017.
[8] X. Fu, J. Huang, D. Zeng, Y. Huang, X. Ding, and J. Paisley, "Removing rain from single images via a deep detail network," in IEEE Conference on Computer Vision and Pattern Recognition, 2017, pp. 1715-1723.
[9] B. Cai, X. Xu, K. Jia, C. Qing, and D. Tao, "Dehazenet: An end-to-end system for single image haze removal," IEEE Transactions on Image Processing, vol. 25, no. 11, pp. 5187-5198, 2016.
[10] W. Ren, S. Liu, H. Zhang, J. Pan, X. Cao, and M.-H. Yang, "Single image dehazing via multi-scale convolutional neural networks," in European conference on computer vision, 2016, pp. 154-169: Springer.
[11] B. Li, X. Peng, Z. Wang, J. Xu, and D. Feng, "Aod-net: All-in-one dehazing network," in Proceedings of the IEEE International Conference on Computer Vision, 2017, pp. 4770-4778.
[12] S. Li, W. Ren, J. Zhang, J. Yu, X. J. C. V. Guo, and I. Understanding, "Single image rain removal via a deep decomposition–composition network," vol. 186, pp. 48-57, 2019.
[13] W. Yang, R. T. Tan, J. Feng, J. Liu, Z. Guo, and S. Yan, "Deep joint rain detection and removal from a single image," in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017, pp. 1357-1366.
[14] Z. Fan, H. Wu, X. Fu, Y. Hunag, and X. Ding, "Residual-Guide Feature Fusion Network for Single Image Deraining," arXiv preprint arXiv:1804.07493, 2018.
[15] G. Li, X. He, W. Zhang, H. Chang, L. Dong, and L. Lin, "Non-locally Enhanced Encoder-Decoder Network for Single Image De-raining," arXiv preprint arXiv:1808.01491, 2018.
[16] X.-J. Mao, C. Shen, and Y.-B. J. a. p. a. Yang, "Image restoration using convolutional auto-encoders with symmetric skip connections," arXiv preprint arXiv:1606.08921, 2016.
[17] I. Goodfellow et al., "Generative adversarial nets," in Advances in neural information processing systems, 2014, pp. 2672-2680.
[18] H. Zhang, V. Sindagi, and V. M. Patel, "Image de-raining using a conditional generative adversarial network," arXiv preprint arXiv:1701.05957, 2017.
[19] R. Li, J. Pan, Z. Li, and J. Tang, "Single image dehazing via conditional generative adversarial network," in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018, pp. 8202-8211.
[20] J. Johnson, A. Alahi, and L. Fei-Fei, "Perceptual losses for real-time style transfer and super-resolution," in European Conference on Computer Vision, 2016, pp. 694-711: Springer.
[21] A. Graves, A.-r. Mohamed, and G. Hinton, "Speech recognition with deep recurrent neural networks," in 2013 IEEE international conference on acoustics, speech and signal processing, 2013, pp. 6645-6649: IEEE.
[22] S. Hochreiter and J. Schmidhuber, "Long short-term memory," Neural computation, vol. 9, no. 8, pp. 1735-1780, 1997.
[23] S. Xingjian, Z. Chen, H. Wang, D.-Y. Yeung, W.-K. Wong, and W.-c. Woo, "Convolutional LSTM network: A machine learning approach for precipitation nowcasting," in Advances in neural information processing systems, 2015, pp. 802-810.
[24] W.-S. Lai, J.-B. Huang, N. Ahuja, and M.-H. Yang, "Deep laplacian pyramid networks for fast and accurate superresolution," in IEEE Conference on Computer Vision and Pattern Recognition, 2017, vol. 2, no. 3, p. 5.
[25] J. Hu, L. Shen, and G. Sun, "Squeeze-and-excitation networks," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 7132-7141.
[26] J. Gama, I. Žliobaitė, A. Bifet, M. Pechenizkiy, and A. Bouchachia, "A survey on concept drift adaptation," ACM Computing Surveys (CSUR), vol. 46, no. 4, p. 44, 2014.
[27] I. Žliobaitė, "Learning under concept drift: an overview," arXiv preprint arXiv:1010.4784, 2010.
[28] S. Muthukrishnan, E. van den Berg, and Y. Wu, "Sequential change detection on data streams," in Data Mining Workshops, 2007. ICDM Workshops 2007. Seventh IEEE International Conference on, 2007, pp. 551-550: IEEE.
[29] E. Ikonomovska, J. Gama, and S. Džeroski, "Learning model trees from evolving data streams," Data mining and knowledge discovery, vol. 23, no. 1, pp. 128-168, 2011.
[30] R. Klinkenberg and I. Renz, "Adaptive information filtering: Learning in the presence of concept drifts," Learning for Text Categorization, pp. 33-40, 1998.
[31] J. B. Gomes, E. Menasalvas, and P. A. Sousa, "Learning recurring concepts from data streams with a context-aware ensemble," in Proceedings of the 2011 ACM symposium on applied computing, 2011, pp. 994-999: ACM.
[32] A. Dries and U. Rückert, "Adaptive concept drift detection," Statistical Analysis and Data Mining: The ASA Data Science Journal, vol. 2, no. 5‐6, pp. 311-327, 2009.
[33] I. Adä and M. R. Berthold, "EVE: a framework for event detection," Evolving systems, vol. 4, no. 1, pp. 61-70, 2013.
[34] J. Gama, P. Medas, G. Castillo, and P. Rodrigues, "Learning with drift detection," in Brazilian symposium on artificial intelligence, 2004, pp. 286-295: Springer.
[35] R. Klinkenberg and T. Joachims, "Detecting Concept Drift with Support Vector Machines," in ICML, 2000, pp. 487-494.
[36] J. M. Carmona-Cejudo, M. Baena-García, J. del Campo-Avila, R. Morales-Bueno, and A. Bifet, "Gnusmail: Open framework for on-line email classification," Frontiers in Artificial Intelligence and Applications, pp. 1141-1142, 2011.
[37] G. Hulten, L. Spencer, and P. Domingos, "Mining time-changing data streams," in Proceedings of the seventh ACM SIGKDD international conference on Knowledge discovery and data mining, 2001, pp. 97-106: ACM.
[38] X. Li, J. Wu, Z. Lin, H. Liu, and H. Zha, "Recurrent squeeze-and-excitation context aggregation net for single image deraining," in Proceedings of the European Conference on Computer Vision (ECCV), 2018, pp. 254-269.
[39] M. Sandler, A. Howard, M. Zhu, A. Zhmoginov, and L.-C. Chen, "MobileNetV2: Inverted Residuals and Linear Bottlenecks," in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018, pp. 4510-4520.
[40] K. He, X. Zhang, S. Ren, and J. Sun, "Deep residual learning for image recognition," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 770-778.
[41] K. He, X. Zhang, S. Ren, and J. Sun, "Identity mappings in deep residual networks," in European conference on computer vision, 2016, pp. 630-645: Springer.
[42] A. Vaswani et al., "Attention is all you need," in Advances in neural information processing systems, 2017, pp. 5998-6008.
[43] L. Itti and C. Koch, "Computational modelling of visual attention," Nature reviews neuroscience, vol. 2, no. 3, pp. 194-203, 2001.
[44] A. M. J. B. m. b. Treisman, "Selective attention in man," British medical bulletin, pp. 12-16, 1964.
[45] X. Qi, C. Liu, and S. Schuckers, "IoT edge device based key frame extraction for face in video recognition," in 2018 18th IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing (CCGRID), 2018, pp. 641-644: IEEE.
[46] A. Howard et al., "Searching for mobilenetv3," in Proceedings of the IEEE International Conference on Computer Vision, 2019, pp. 1314-1324.
[47] H. Zhang and V. M. Patel, "Density-aware single image de-raining using a multi-stream dense network," arXiv preprint arXiv:1802.07412, 2018.
[48] B. Li et al., "Benchmarking Single-Image Dehazing and Beyond," IEEE Transactions on Image Processing, vol. 28, no. 1, pp. 492-505, 2019.
[49] D. P. Kingma and J. Ba, "Adam: A method for stochastic optimization," arXiv preprint arXiv:1412.6980, 2014.
[50] Y.-F. Liu, D.-W. Jaw, S.-C. Huang, and J.-N. J. I. T. o. I. P. Hwang, "DesnowNet: Context-aware deep network for snow removal," IEEE Transactions on Image Processing, vol. 27, no. 6, pp. 3064-3073, 2018.
[51] W. Liu et al., "Ssd: Single shot multibox detector," in European conference on computer vision, 2016, pp. 21-37: Springer.
[52] A. Bochkovskiy, C.-Y. Wang, and H.-Y. M. J. a. p. a. Liao, "YOLOv4: Optimal Speed and Accuracy of Object Detection," arxiv preprint arXiv:2004.10934, 2020.
[53] W. Yang et al., "Joint rain detection and removal from a single image with contextualized deep networks," IEEE transactions on pattern analysis and machine intelligence, vol. 42, no. 6, pp. 1377-1393, 2019.
[54] Y. Du, J. Xu, X. Zhen, M.-M. Cheng, and L. Shao, "Conditional Variational Image Deraining," IEEE Transactions on Image Processing, 2020.
[55] K. Jiang et al., "Multi-scale progressive fusion network for single image deraining," in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020, pp. 8346-8355.
[56] S. Mehta, M. Rastegari, A. Caspi, L. Shapiro, and H. Hajishirzi, "Espnet: Efficient spatial pyramid of dilated convolutions for semantic segmentation," in Proceedings of the European Conference on Computer Vision (ECCV), 2018, pp. 552-568.
[57] Z. Li et al., "Single Image Snow Removal via Composition Generative Adversarial Networks," IEEE Access, vol. 7, pp. 25016-25025, 2019.