Adaptive target detection method based on lightweight residual learning and deconvolution cascading

Abstract

The invention discloses an adaptive target detection method based on a lightweight residual network and deconvolution cascading. The method comprises the following steps: obtaining an image training data set and a test data set; extracting the deep features of a to-be-detected image through a lightweight residual network in combination with deep separable convolution and residual learning to obtain the deep expression of a target; using 1 * 1 convolution to fixedly output feature map dimensions for the extracted feature maps of different levels; increasing the resolution of the deep-level feature map by using a deconvolution cascading structure so as to realize that the spatial size of the deep-level feature map is consistent with the spatial size of a previous-level feature map; guiding the candidate region generation network to adaptively generating a target candidate box which is more matched with the real target on the multi-scale feature map by utilizing the semantic features; andfinally, correcting the generated target candidate box Anchor. According to the method, the accuracy of target detection is effectively improved, the target can be rapidly and accurately detected under complex conditions, and the real-time performance of target detection is effectively improved.

Description

technical field [0001] The invention relates to a target detection method, which belongs to the fields of digital image processing, deep learning and artificial intelligence, and especially designs an adaptive target detection method based on lightweight residual learning and deconvolution cascade. Background technique [0002] With the rapid development of computer vision technology, target detection technology has become a research hotspot in the field of artificial intelligence and computer vision, and is widely used in military and civilian fields. Target detection is mainly aimed at identifying and locating one or more specific targets in a video image sequence. In most cases, video image acquisition equipment contains rich visual content. Although it can provide more comprehensive scene information, the target to be detected usually has a large scale change in the image or video, the distribution is concentrated, there is occlusion, and there is not enough Detect deta...

AI Technical Summary

Problems solved by technology

[0004] The shortcomings of the existing methods: on the one hand, the traditional classical target detection algorithm is limited by artificially designed manual features and selective search algorithms, resulting in low target detection accuracy, slow detection speed, and poor algorithm robustness; on the other hand, Although the accuracy of target detection based on deep learning has been improved, the convolutional neural network has a large number of parameters, the algorithm structure is complex, and the amount of calculation is large, which makes it difficult to meet real-time requirements.

Images