Lightweight remote sensing target detection method based on SE-YOLOv3

Abstract

The invention relates to a lightweight remote sensing target detection method based on SEYOLOv3, which belongs to the technical field of target detection, and comprises the following steps: 1, takinga YOLOv3 algorithm as a basic model framework, and in order to reduce network parameters and improve network reasoning speed, designing a lightweight trunk feature extraction network; 2, in order to improve the scale invariance of the features and reduce the over-fitting risk, a spatial pyramid pooling (SPP) algorithm is provided, and pooling of three scales is carried out to obtain an output feature vector with a fixed length; a spatial attention model SE module is introduced, useless information is further compressed, and useful information is enhanced; and 3, updating parameters through iterative training to obtain a final network model, adopting multi-scale prediction by utilizing the model, and predicting a final result through detection heads of three scales. According to the method,while the reasoning speed of the network is effectively improved, the precision is ensured, the feature expression capability of the network is enhanced, and the scale invariance is improved.

Description

technical field [0001] The invention belongs to the field of computer vision and deep learning, in particular to a light-weight framework remote sensing image target detection method based on SE-YOLOv3. Background technique [0002] With the rapid development of aerospace technology and deep learning, high-resolution and large-scale remote sensing image data are constantly enriched. Remote sensing images usually have problems such as large scale changes, high resolution, and sparse distribution of targets. Artificial neural networks are widely used in the field of target detection in remote sensing images, but most of the algorithms are based on the prior frame method, and perform all-round scanning detection in remote sensing images. For images of large scenes and large feature extraction networks, this One way requires a lot of computing resources. In order to balance detection speed and detection accuracy, realizing fast remote sensing target detection has become a resea...

AI Technical Summary

Problems solved by technology

The One-stage method is much faster than the Two-stage method in terms of speed, but relatively low in accuracy

[0004] Due to the characteristics of single imaging angle of view, dense distribution of targets, and large changes in target scales in remote sensing images, the direct application of natural scene target detection methods to remote sensing image target detection tasks cannot achieve satisfactory results.

Moreover, its high resolution and large image size will increase the calculation cost of the algorithm

In recent years, the One-stage algorithm has been comparable to the Two-stage algorithm in terms of accuracy. The YOLO algorithm series is a representative One-stage algorithm. The YOLOv3 algorithm is a target detection network with balanced speed and accuracy, but compared to RCNN A series of object detection methods have poor accuracy in identifying object positions and low recall rates

Images