Image super-resolution reconstruction (SR) aims to obtain high-resolution images from one or more low-resolution images. Recently, SR has been developing and widely applied in different fields. This survey retrospects the history of SR technique and provides a comprehensive overview of representative SR methods, with an emphasis on recent deep learning-based approaches. We elaborate the details of various deep learning-based SR methods, including their strengths and weakness, in terms of the deep learning model, architecture, and message pass. Finally, we discuss the possible research directions on SR technique.

The computer vision based scene classification technology is widely developed and applied in different fields.In this paper,the typical scene classification technology is analyzed and compared from the different directions.First,the background,application and development situation is introduced.Then,the related researches both at home and overseas are analyzed,compared and summarized from the perspectives of feature extraction,semantic analysis and machine learning.Finally,the problems that the current researches are facing and potential future development are discussed.

Object detection is a hot topic in the field of computer vision, and has been widely used in robot navigation, intelligent video surveillance, aerospace, and other fields. The research background, significance and challenges of object detection were introduced. Then the object detection algorithms based on deep learning were reviewed according to two categories: candidate region-based and regression-based. For the candidate region-based algorithms, we first introduced the R-CNN (Region with Convolutional Neural Network) based series of algorithms, and then the R-CNN based methods were overviewed from four dimensions: the research of feature extraction networks, the region of interesting pooling researches, improved works based on region proposal networks, and some improved approaches of non maximum suppression algorithms. Next, the regression-based algorithms were surveyed in terms of YOLO (You Only Look Once) series and SSD (Single Shot multibox Detector) series. Finally, according to the current trend of object detection algorithms that are developing more efficient and reasonable detection frameworks, the future research focuses of unsupervised and unknown category object detection directions were prospected.

In order to improve the efficiency of high-resolution image segmentation and solve the problem of incomplete segmentation caused by small discrimination of foreground and background in the complex pattern near the edge of the target to be segmented, we propose a fast image segmentation algorithm based on superpixel multi-feature fusion (SMFF). Firstly, the most effective superpixel algorithm is used for superpixel processing, and then the superpixel-based HOG feature, laboratory color feature and spatial position feature are extracted. Lastly, by designing a multi-feature measurement algorithm, the fast image segmentation algorithm based on superpixel multi-feature fusion is implemented. Experimental results verify the effectiveness of the proposed algorithm, which is close to the most classical image segmentation algorithm, and the time performance of the proposed algorithm is significantly better than other comparison algorithms.

The bilinear convolutional neural network(B-CNN) has been widely used in computer vision. B-CNN can capture the linear correlation between different channels by performing the outer product operation on the features of the convolutional layer output, thus enhancing the representative ability of the convolutional network. Since the non-linear relationship between the channels in the feature map is not taken account of, this method cannot make full use of the richer information contained between the channels. In order to solve this problem, this paper proposes a kernelized bilinear convolutional neural network employing the kernel function to effectively capture the non-linear relationship between the channels in the feature map, and further enhancing the representative ability of the convolutional network. In this paper, the method is evaluated on three common fine-grained benchmarks CUB-200-2011, FGVC-Aircraft and Cars. Experiments show that our method is superior to its counterparts on all three benchmarks.

The image database of crime scene investigation (CSI)has the characteristics of high confidentiality,rare image content and so on.Aiming at the complexity of the content and the ambiguity of the target object,the DCT-DCT wave texture feature is proposed,which is fused with HSV color histogram feature and GIST feature to form the fusion feature.Compared with the commonly used image features,DCT-DCT wave texture feature can get higher retrieval efficiency,and the average retrieval precision rate of the fused features is higher than that of the three features.Finally,the semantic analysis technology is introduced into the retrieval process,and an image retrieval algorithm based on the optimization of retrieval results is proposed.Support vector machine (SVM)classifier was used to extract the semantic of the query image.The semantic analysis of the results of the first retrieval is carried out,and the second retrieval scheme is selected according to the proportion of semantic categories in the initial retrieval results.The algorithm can further improve the average retrieval accuracy based on case-by-case query.

To solve the problem of slow speed and low accuracy in the surface defect detection of hot rolled strips, an improved YOLOv3 algorithm is proposed. Firstly, the weighting K-means clustering algorithm is put forward to optimize priors anchor's parameters, which can improve the match between priors anchor and feature map. Secondly, the improved network structure of the YOLOv3 algorithm is proposed to improve the detection accuracy, whose shallow features and deep features are combined to form the new large-scale inspection layer. The experiments are carried out on the NEU-DET dataset, the results show that the average accuracy of the improved YOLOv3 algorithm is 80%, which is 11% higher than that of the original algorithm; the detection speed is 50fps, which is faster than other strip surface defect detection algorithms based on deep learning.

Most of current crop-disease recognition approaches mainly focus on improving the recognition accuracy on public datasets, while ignoring the recognition robustness.When dealing with real-world recognition problem, the actual recognition accuracy of those approach are often unsatisfactory because of noise interference and environmental influence. To address these issues, we propose a high-order residual and parameter-sharing feedback convolutional neural network (HORPSF) for crop-disease recognition. The high-order residual subnetwork is helpful for improving the recognition accuracy of crop disease. The parameter-sharing feedback subnetwork can effectively depress the background noises and enhance the robustness of the model. Extensive experiment results demonstrate that the proposed HORPSF approach significantly outperforms other competing methods in terms of recognition accuracy and robustness, especially demonstrating superior performance when dealing with the real-world examples of crop-disease recognition.