Inclined target detection method and device based on yolv3 algorithm, and storage medium

Abstract

The invention provides an inclined target detection method and device based on a yolv3 algorithm and a storage medium, and the method comprises the following steps: obtaining a picture containing a detection target, and respectively generating a training set A and a verification set B; labeling the detection targets in the training set A and the verification set B one by one by using rectangular frames at any angle; training an improved yolov3 algorithm by using the training set A, wherein the improved yolov3 algorithm is an algorithm obtained by improving an anger frame, IOU calculation and frame regression on the basis of an original yolov3 algorithm so as to adapt to inclined target detection; then selecting a weight model with the best effect and the best generalization ability throughthe verification set B; and transplanting the improved yolo v3 algorithm to the NVIDIA Jetson TX2, and loading the trained optimal weight to carry out real-time detection on the target in the picture. According to the invention, common vertical target detection can be processed, and detection of an inclined target can also be processed.

Description

technical field [0001] The present invention relates to the fields of computer vision and machine learning, in particular to a method, device and storage medium for detecting an inclined target based on the yolo v3 algorithm. Background technique [0002] Target detection is a relatively practical and challenging computer vision task, which can be seen as a combination of image classification and positioning. Given a picture, the target detection system must be able to identify the target of the picture and give its location. Since the number of objects in the picture is uncertain and the precise location of the object must be given, object detection is more complicated than classification tasks. [0003] In recent years, object detection algorithms have made great breakthroughs. The more popular algorithms can be divided into two categories. One is the R-CNN algorithm based on Region Proposal (R-CNN, Fast R-CNN, Faster R-CNN, Mask R-CNN). They are two-stage networks that r...

AI Technical Summary

Problems solved by technology

[0005] Due to the actual situation, the above-mentioned algorithms are used to detect horizontal or vertical targets, not all detected targets are horizontal or vertical. If such algorithms are applied to the detection of inclined targets, the detection results will be inaccurate. The detection accuracy we need cannot be achieved, so we design a method of oblique target detection based on the yolo v3 network

Images