Human behavior recognition is an important basic technology in the fields such as intelligent monitoring, human-computer interaction and robotics. Graph Convolutional Neural Network (GCN) achieve excellent performance in skeleton-based human behavior recognition. The following problems exist in the research of human behavior recognition using GCNs:1) the human skeleton points are represented by coordinates, which lacks detailed information about the movement of the skeleton points; 2) in some videos, the motion amplitude of the human skeleton is too small, so that the representation information of the key skeleton points is not obvious. Aiming at the above problems, firstly, a temporal divergence model of skeleton points was designed to describe the movement states of the skeleton points, which amplified the between-class variances of different human behaviors. In addition, the attention mechanism of temporal divergence features was designed to highlight the key skeleton points and further expand the between-class variances. Finally, a two-stream fusion model was constructed based on the complementarity between the spatial data characteristics of the original skeleton and the temporal divergence characteristics. The proposed algorithm achieved the accuracy of 82.9% and 83.7% under two partitioning strategies of authoritative human behavior dataset NTU-RGB+D respectively, which were 1.3 percentage points and 0.5 percentage points higher than those of Adaptive Graph Convolutional Network (AGCN) respectively. The improvement of the accuracy of the proposed algorithm on the dataset proves the effectiveness of this algorithm.

Liver is the main organ of human metabolic function. At present, the main problems of machine learning in the semantic segmentation of liver images are as follows:1) there are inferior vena cava, soft tissue and blood vessels in the middle of the liver, and even some necrosis or hepatic fissures; 2) the boundary between the liver and some adjacent organs is blurred and difficult to distinguish. In order to solve the problems mentioned above, the Edge Attention Model (EAM)and the Edge Attention Net (EANet) were proposed by using Encoder-Decoder framework. In the encoder, the residual network ResNet34 pre-trained on ImageNet and the EAM were utilized, so as to fully obtain the detailed feature information of liver edge; in the decoder, the deconvolution operation and the proposed EAM were used to perform the feature extraction to the useful information, thereby obtaining the semantic segmentation diagram of liver image. Finally, the smoothing was performed to the segmentation images with a lot of noise. Comparison experiments with AHCNet were conducted on three datasets, and the results showed that:on 3Dircadb dataset, the Volumetric Overlap Error (VOE) and Relative Volume Difference (RVD) of EANet were decreased by 1.95 percentage points and 0.11 percentage points respectively, and the DICE accuracy was increased by 1.58 percentage points; on Sliver07 dataset, the VOE, Maximum Surface Distance (MSD) and Root Mean Square Surface Distance (RMSD) of EANet were decreased approximately by 1 percentage points, 3.3 mm and 0.2 mm respectively; on clinical MRI liver image dataset of a hospital, the VOE and RVD of EANet were decreased by 0.88 percentage points and 0.31 percentage points respectively, and the DICE accuracy was increased by 1.48 percentage points. Experimental results indicate that the proposed EANet has good segmentation effect of liver image.

Anchor-free deep learning based object detection is a mainstream single-stage object detection algorithm. An hourglass network structure that incorporates multiple layers of supervisory information can significantly improve the accuracy of the anchor-free object detection algorithm, but its speed is much lower than that of a common network at the same level, and the features of different scale objects will interfere with each other. In order to solve the above problems, an object detection algorithm based on asymmetric hourglass network structure was proposed. The proposed algorithm is not constrained by the shape and size when fusing the features of different network layers, and can quickly and efficiently abstract the semantic information of network, making it easier for the model to learn the differences between various scales. Aiming at the problem of object detection at different scales, a multi-scale output hourglass network structure was designed to solve the problem of feature mutual interference between different scale objects and refine the output detection results. In addition, a special non-maximum suppression algorithm for multi-scale outputs was used to improve the recall rate of the detection algorithm. Experimental results show that the AP50 index of the proposed algorithm on Common Objects in COntext (COCO) dataset reaches 61.3%, which is 4.2 percentage points higher than that of anchor-free network CenterNet. The proposed algorithm surpasses the original algorithm in the balance of accuracy and time, and is particularly suitable for real-time object detection in industry.

In order to solve the detection of ships with different types and sizes under complex marine environment, a real-time object detection algorithm based on deep learning was proposed. Firstly, a discriminant method between sharp and fuzzy such as rainy and foggy images was proposed. Then a multi-scale object detection algorithm based on deep learning framework of You Only Look Once (YOLO) v2 was proposed. Finally, concerning the character of remote sensing images of ships, an improved non-maximum supression and saliency partitioning algorithm was proposed to optimize the final detection results. The experimental results show that, on the dataset of ship detection in an open competition under complex sea conditions and meteorological conditions, the precision of the proposed method is increased by 16% compared with original YOLO v2 algorithm.