Light field saliency target detection method based on generative adversarial convolutional neural network

Abstract

The invention discloses a light field saliency target detection method based on a generative adversarial convolutional neural network. The method comprises the following steps: 1, converting light field data into a refocusing sequence; 2, performing data enhancement on the refocusing sequence; 3, constructing a generative adversarial convolutional neural network on the basis of a U-Net network anda GAN network structure, taking a refocusing sequence as network input, and performing training by using the light field data set; and 4, using the trained generative adversarial convolutional neuralnetwork to carry out significance target detection on the to-be-processed light field data. According to the method, a deep learning method and light field refocusing information can be fully utilized, so that the accuracy of salient target detection of a complex scene image can be effectively improved.

Description

technical field [0001] The invention belongs to the field of computer vision, and specifically relates to a light field salient target detection method based on generating an anti-convolutional neural network. Background technique [0002] Salient object detection is an attention mechanism of the human visual system. When we face a scene, humans automatically process the regions of interest and selectively ignore the regions that are not of interest. These regions of interest to people are called salient regions. Salient target detection is to select a part of the data that the observer is most interested in from the input visual information for processing, such as target recognition, target tracking and image segmentation. At present, saliency detection has become one of the hot research directions in the field of computer vision. [0003] The current methods of light field salient object detection can be roughly divided into two categories: statistics-based methods and l...

AI Technical Summary

Problems solved by technology

[0007] 1. Statistical-based methods usually make assumptions about significant targets, which have a small scope of application and are suitable for simple scene predictions. When dealing with complex scenes, the error is relatively large

[0008] 2. Early learning-based methods can only manually extract some low-level visual features and build simple learning models

[0009] 3. In some methods based on deep learning, the connection between feature information such as color, depth, and position is separated, and the correlation and complementarity between them are not fully considered.

[0010] 4. In the salient target detection method, for methods based on deep learning, the update of network parameters mainly comes from training data, so a large amount of training data is usually required, and the convergence is relatively slow

However, the current research on light field saliency is still in its infancy, and the number of data sets is limited, which cannot achieve higher prediction accuracy.

Images