mb-ssd method and mb-ssd feature extraction network suitable for object detection

Abstract

The invention discloses a MB‑SSD method suitable for target detection, which includes: extracting multi-small target images, using a generative confrontation network to enhance the extracted multi-small target images; constructing a MB‑SSD feature extraction network, the MB‑SSD The SSD feature extraction network includes the main branch feature extraction network, the branch feature extraction network and the positioning network. The enhanced data is input into the MB-SSD feature extraction network, and the classification and positioning results of the main branch feature extraction network and the branch feature extraction network are respectively obtained; and then according to The IoU coincidence degree of different candidate frames in the same area on the positioning network adjusts the output features of the classification network; the classification and positioning results of the main branch feature extraction network and branch feature extraction network are merged and dimensionality is reduced; the model loss is calculated, the model is trained, and the model parameters are optimized . The invention can add a method of adjusting the classification result by relative coincidence degree in the classification layer, so as to improve the classification effect of the SSD algorithm, and effectively improve the detection accuracy of the small target at the same time.

Description

Technical field[0001]The present invention relates to the field of computer vision technology, in particular to an MB-SSD method and MB-SSD feature extraction network suitable for target detection.Background technique[0002]Target detection is a popular research direction in the field of computer vision, which can be applied to areas such as unmanned driving, video surveillance, pedestrian detection, and remote sensing image detection. Traditional target detection algorithms first manually extract features, such as SIFT (Scale Invariant Feature Transform), HOG (Histogram of Oriented Gradient), SURF (Speeded Up RobustFeatures), etc. , And then combine these artificially extracted features with the classifier for target recognition, and finally combine the corresponding strategies to locate the target location, but for scenes with complex and changeable backgrounds and complex and changeable targets, it is difficult for people to summarize the abstraction of the image Characteristics, ...

AI Technical Summary

Problems solved by technology

For the detection of large targets in sparse scenes, the detection accuracy of the SSD algorithm can exceed that of the more accurate Faster R-CNN, but the detection effect for small targets is not good, and there are cases of missed detection and false detection of small targets.

Images