Search

Abstract In the underwater domain, small object detection plays a crucial role in the protection, management, and monitoring of the environment and marine life. Advancements in deep learning have led to the development of many efficient detection techniques. However, the complexity of the underwater environment, limited information available from small objects, and constrained computational resources make small object detection challenging. To tackle these challenges, this paper presents an efficient deep convolutional network model. First, a CSP for small object and lightweight (CSPSL) module is introduced to enhance feature retention and preserve essential details. Next, a variable kernel convolution (VKConv) is proposed to dynamically adjust the convolution kernel size, enabling better multi-scale feature extraction. Finally, a spatial pyramid pooling for multi-scale (SPPFMS) method is presented to preserve the features of small objects more effectively. Ablation experiments on the UDD dataset demonstrate the effectiveness of the proposed methods. Comparative experiments on the UDD and DUO datasets demonstrate that the proposed model delivers the best performance in terms of computational cost and detection accuracy, outperforming state-of-the-art methods in real-time underwater small object detection tasks.

Abstract With the advances in technology, humans tend to explore the world underwater in a more constructive way than before. The appearance of an underwater object varies depending on depth, biological composition, temperature, ocean currents, and other factors. This results in colour distorted images and hazy images with low contrast. To address the aforesaid problems, in proposed approach, initially White balance algorithm is carried out to pre-process original underwater image. Contrast enhanced image is achieved by applying the Contrast Limited Adaptive Histogram Equalization algorithm (CLAHE). In CLAHE, tile size and clip limit are the major parameters that control the enhanced image quality. Hence, to enhance the contrast of images optimally, Firefly algorithm is adopted for CLAHE. Dark Channel Prior algorithm (DCP) is modified with guided filter correction to get the sharpened version of the underwater image. Multiscale fusion strategy was performed to fuse CLAHE enhanced and dehazed images. Finally, the restored image is treated with optimal CLAHE to improve visibility of enhanced underwater image. Experimentation is carried out on different underwater image datasets such as U45 and RUIE and resulted in UIQM = 5.1384, UCIQE = 0.6895 and UIQM = 5.4875, UCIQE = 0.6953 respectively which shows the superiority of proposed approach.

Limited by the imaging capabilities of sensors, research based on single modality is difficult to cope with faults and dynamic perturbations in detection. Effective multispectral object detection, which can achieve better detection accuracy by fusing visual information from different modalities, has attracted widespread attention. However, most of the existing methods adopt simple fusion mechanisms, which fail to utilize the complementary information between modalities while lacking the guidance of a priori knowledge. To address the above issues, we propose a novel background-aware cross-attention multiscale fusion network (BA-CAMF Net) to achieve adaptive fusion in visible and infrared images. First, a background-aware module is designed to calculate the light and contrast to guide the fusion. Then, a cross-attention multiscale fusion module is put forward to enhance inter-modality complement features and intra-modality intrinsic features. Finally, multiscale feature maps from different modalities are fused according to background-aware weights. Experimental results on LLVIP, FLIR, and VEDAI indicate that the proposed BA-CAMF Net achieves higher detection accuracy than the current State-of-the-Art multispectral detectors.

Ocean exploration has always been an important strategic direction for the joint efforts of all mankind. Many countries in the world today are developing their own underwater autonomous explorers to better explore the seabed. Vision, as the core technology of autonomous underwater explorers, has a great impact on the efficiency of exploration. Different from traditional tasks, the lack of ambient light on the seabed makes the visual system more demanding. In addition, the complex terrain on the seabed and various creatures with different shapes and colors also make exploration tasks more difficult. In order to effectively solve the above problems, we combined the traditional models to modify the structure and proposed an algorithm for the super-resolution fusion of enhanced extraction features to perform semantic segmentation of seabed scenes. By using a structurally reparameterized backbone network to better extract target features in complex environments, and using subpixel super-resolution to combine multiscale feature semantic information, we can achieve superior ocean scene segmentation performance. In this study, multiclass segmentation and two-class segmentation tests were performed on the public datasets SUIM and DeepFish, respectively. The test results show that the mIoU and mPA indicators of our proposed method on SUIM reach 84.52% and 92.33%mPA, respectively. The mIoU and mPA on DeepFish reach 95.26% and 97.38%, respectively, and the proposed model achieves SOTA compared with state-of-the-art methods. The proposed model and code are exposed via Github1.

Recently, the isolation forest (IF) methods have received increasing attention for their promising performance in hyperspectral anomaly detection (HAD). However, limited by the ability of exploiting spatial-spectral information, existing IF-based methods suffer from a lot of false alarms and disappointing performance of detecting local anomalies. To overcome the two problems, a multiscale superpixel guided discriminative forest method is proposed for HAD. First, the multiscale superpixel segmentation is employed to generate some homogeneous regions, and it can effectively extract spatial information to guide anomaly detection for the discriminative forest in local areas. Then, a novel discriminative forest (DF) model with the gain split criterion is designed, which enhances the sensitivity of the DF to local anomalies by the utilization of multi-dimension spectral bands for node division; meanwhile, the acceptable range of hyperplane attribute values is introduced to capture any unseen anomaly pixels that are out-of-range in the evaluation stage. Finally, for the high false alarm rate situation in the existing IF-based algorithms, the multiscale fusion with guided filtering is put forward to refine the initial detection results from the DF. In addition, the extensive experimental results on four real hyperspectral datasets demonstrate the effectiveness of the proposed method.

Hyperspectral image classification (HSIC) methods usually require more training samples for better classification performance. However, a large number of labeled samples are difficult to obtain because it is cost- and time-consuming to label an HSI in a pixel-wise way. Therefore, how to overcome the problem of insufficient accuracy and stability under the condition of small labeled training sample size (SLTSS) is still a challenge for HSIC. In this paper, we proposed a novel multiple superpixel graphs learning method based on adaptive multiscale segmentation (MSGLAMS) for HSI classification to address this problem. First, the multiscale-superpixel-based framework can reduce the adverse effect of improper selection of a superpixel segmentation scale on the classification accuracy while saving the cost to manually seek a suitable segmentation scale. To make full use of the superpixel-level spatial information of different segmentation scales, a novel two-steps multiscale selection strategy is designed to adaptively select a group of complementary scales (multiscale). To fix the bias and instability of a single model, multiple superpixel-based graphical models obatined by constructing superpixel contracted graph of fusion scales are developed to jointly predict the final results via a pixel-level fusion strategy. Experimental results show that the proposed MSGLAMS has better performance when compared with other state-of-the-art algorithms. Specifically, its overall accuracy achieves 94.312%, 99.217%, 98.373% and 92.693% on Indian Pines, Salinas and University of Pavia, and the more challenging dataset Houston2013, respectively.

Real-time acquisition and intelligent classification of pole-like street-object point clouds are of great significance in the construction of smart cities. Efficient point cloud processing technology in road scenes can accelerate the development of intelligent transportation and promote the development of high-precision maps. However, available algorithms have the problems of incomplete extraction and the low recognition accuracy of pole-like objects. In this paper, we propose a segmentation method of pole-like objects under geometric structural constraints. As for classification, we fused the classification results at different scales with each other. First, the point cloud data excluding ground point clouds were divided into voxels, and the rod-shaped parts of the pole-like objects were extracted according to the vertical continuity. Second, the regional growth based on the voxel was carried out based on the rod part to retain the non-rod part of the pole-like objects. A one-way double coding strategy was adopted to preserve the details. For spatial overlapping entities, we used multi-rule supervoxels to divide them. Finally, the random forest model was used to classify the pole-like objects based on local- and global-scale features and to fuse the double classification results under the different scales in order to obtain the final result. Experiments showed that the proposed method can effectively extract the pole-like objects of the point clouds in the road scenes, indicating that the method can achieve high-precision classification and identification in the lightweight data. Our method can also bring processing inspiration for large data.

Urban land cover classification for high-resolution images is a fundamental yet challenging task in remote sensing image analysis. Recently, deep learning techniques have achieved outstanding performance in high-resolution image classification, especially the methods based on deep convolutional neural networks (DCNNs). However, the traditional CNNs using convolution operations with local receptive fields are not sufficient to model global contextual relations between objects. In addition, multiscale objects and the relatively small sample size in remote sensing have also limited classification accuracy. In this paper, a relation-enhanced multiscale convolutional network (REMSNet) method is proposed to overcome these weaknesses. A dense connectivity pattern and parallel multi-kernel convolution are combined to build a lightweight and varied receptive field sizes model. Then, the spatial relation-enhanced block and the channel relation-enhanced block are introduced into the network. They can adaptively learn global contextual relations between any two positions or feature maps to enhance feature representations. Moreover, we design a parallel multi-kernel deconvolution module and spatial path to further aggregate different scales information. The proposed network is used for urban land cover classification against two datasets: the ISPRS 2D semantic labelling contest of Vaihingen and an area of Shanghai of about 143 km2. The results demonstrate that the proposed method can effectively capture long-range dependencies and improve the accuracy of land cover classification. Our model obtains an overall accuracy (OA) of 90.46% and a mean intersection-over-union (mIoU) of 0.8073 for Vaihingen and an OA of 88.55% and a mIoU of 0.7394 for Shanghai.

The convolutional neural network (CNN) can automatically extract hierarchical feature representations from raw data and has recently achieved great success in the classification of hyperspectral images (HSIs). However, most CNN based methods used in HSI classification neglect adequately utilizing the strong complementary yet correlated information from each convolutional layer and only employ the last convolutional layer features for classification. In this paper, we propose a novel fully dense multiscale fusion network (FDMFN) that takes full advantage of the hierarchical features from all the convolutional layers for HSI classification. In the proposed network, shortcut connections are introduced between any two layers in a feed-forward manner, enabling features learned by each layer to be accessed by all subsequent layers. This fully dense connectivity pattern achieves comprehensive feature reuse and enforces discriminative feature learning. In addition, various spectral-spatial features with multiple scales from all convolutional layers are fused to extract more discriminative features for HSI classification. Experimental results on three widely used hyperspectral scenes demonstrate that the proposed FDMFN can achieve better classification performance in comparison with several state-of-the-art approaches.