Abstract
The CNN (convolutional neural network)-based small target detection techniques for static complex scenes have been applied in many fields, but there are still certain technical challenges. This paper proposes a novel high-resolution small-target detection network named the IIHNet (information interworking high-resolution network) for complex scenes, which is based on information interworking processing technology. The proposed network not only can output a high-resolution presentation of a small target but can also keep the detection network simple and efficient. The key characteristic of the proposed network is that the target features are divided into three categories according to image resolution: high-resolution, medium-resolution, and low-resolution features. The basic features are extracted by convolution at the initial layer of the network. Then, convolution is carried out synchronously in the three resolution channels with information fusion in the horizontal and vertical directions of the network. At the same time, multiple utilizations and augmentations of feature information are carried out in the channel convolution direction. Experimental results show that the proposed network can achieve higher reasoning performance than the other compared networks without any compromise in terms of the detection effect.
Highlights
With the development of convolutional neural network (CNN), their application in the field of target detection has gradually increased, and new methods have been emerging one after another
Since a high detection performance is required in real-world application scenarios, it is a common practice to use a method developed based on a faster RCNN [8] for performance improvement and optimization
The cascade RCNN mainly analyzes the influence of the IoU threshold selection on results at different stages; the HRNet network achieves a high-resolution presentation by using parallel multiresolution networks and multiscale information fusion on the basis of learning the underlying information exchange; the DenseNet network [9] provides a high-performance network by performing information fusion between channels and employs the ResNext [10] and Xception [11] models to use the convoluted data of every channel
Summary
With the development of CNNs (convolutional neural networks), their application in the field of target detection has gradually increased, and new methods have been emerging one after another. Since a high detection performance is required in real-world application scenarios, it is a common practice to use a method developed based on a faster RCNN [8] for performance improvement and optimization. The cascade RCNN mainly analyzes the influence of the IoU threshold selection on results at different stages; the HRNet network achieves a high-resolution presentation by using parallel multiresolution networks and multiscale information fusion on the basis of learning the underlying information exchange; the DenseNet network [9] provides a high-performance network by performing information fusion between channels and employs the ResNext [10] and Xception [11] models to use the convoluted data of every channel. It is necessary to test multiple schemes before an effective strategy and a network type are selected On this basis, a particular network is built according to the requirements of a specific scene and good detection performance can be obtained through repeated network optimization and enhancement. The research and development process is complex and time-consuming and it is hard to achieve a good balance between detection performance and reasoning performance
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
