Automatic detection method for multi-scale polymorphic target in two-dimensional image sequence

Abstract

The invention discloses an automatic detection method for a multi-scale polymorphic target in a two-dimensional image sequence. The method comprises the following steps: changing the size of an original two-dimensional image and calibrating a target; forming a three-dimensional array according to stacking of the two-dimensional image sequence, and then an image of another plane being obtained anda target of the image being calibrated; respectively training data sets of two planes by using a target detection neural network, ensuring a relatively high recall rate as much as possible, and obtaining a detection result of two-dimensional image prediction; projecting a detection result on the target plane to the auxiliary plane to obtain a frame position, comparing the frame position with a detection frame predicted by the auxiliary plane network, judging an intersection degree of the frame position and the detection frame, and removing a false positive example appearing in two-dimensionalimage prediction detection according to a prediction detection result of the auxiliary plane; and finally, further removing false positive examples by using the gray continuity of the target in the image, recovering the image to the original size, and synthesizing the information of the two planes through the steps to obtain a target detection result with relatively high accuracy.

Description

technical field [0001] The invention relates to the technical field of image analysis and target detection, in particular to an automatic detection method for multi-scale and multi-morphological targets in a two-dimensional image sequence. Background technique [0002] Object detection widely exists in face recognition, gesture recognition, medical detection and other tasks, and is one of the most basic and important technologies in the field of machine vision. Target detection includes two steps: candidate area extraction and target category discrimination, among which the target category judgment is the main difference between various detection methods. At present, for the target detection of two-dimensional image sequences, there are mainly traditional detection methods that artificially construct features and use methods such as rule matching and support vector machines for classification and recognition; and deep learning methods such as convolutional neural networks. ...

AI Technical Summary

Problems solved by technology

[0006] Purpose of the invention: The technical problem to be solved by the present invention is to solve the problem of how to effectively use the feature of image continuity in multi-size and multi-morphological target detection in two-dimensional image sequences. Based on the convolutional neural network in deep learning, a combination of Another dimensional information is used to assist in judging the multi-size and multi-morphological target on the two-dimensional image sequence, which realizes more accurate detection of the target

Images