Path Optimization of Target Detection Method Based on Deep Learning Feature Fusion

Abstract

With the development of network informatization, the rapid growth of multimedia information such as voice, image, video, etc., for the massive images containing rich visual information, how can the target detection method accurately and quickly retrieve the images that users need in the large-scale image library? Become a research hotspot in the field of multimedia information retrieval. The purpose of this paper is to study the path optimization of the target detection method based on deep learning feature fusion. In this paper, the two most important parts in the path optimization of the target detection method based on deep learning feature fusion are feature extraction and training classifiers. How to design a robust feature extraction method that is not affected by the environment, and choose a high discrimination Sexual classifiers are the key to determining the pros and cons of pedestrian detection methods. Among them, feature extraction is mainly divided into extraction methods based on manual design and learning. This article mainly explores the feature extraction part, researches new manual design features and deep features based on deep networks, and introduces a pedestrian from rough to fine Detection scheme and new data expansion technology. Experiments show that the method in this paper can further improve the detection performance. Compared with the LDCF method on the Caltech data set, the missed detection rate dropped from 24.80% to 11.82%, a 12.98% drop. Although compared with the current best deep learning methods, there are still shortcomings, but the method in this paper has advantages in detection speed, which can better balance detection accuracy and speed.