Vehicle detection method based on YOLO v3 improved algorithm

Abstract

The invention discloses a vehicle detection method based on a YOLO v3 improved algorithm. The method comprises the steps: S1, collecting a vehicle data set, wherein the vehicle data set comprises a plurality of images I of an object GT frame; S2, clustering the width and the height of the GT frame by using a dimension clustering method to generate the sizes of the width and the height of K prior frames; S3, inputting the image I into a skeleton network Darknet-53 of the YOLO v3 for feature extraction, and outputting feature maps of different scales; S4, averagely distributing the sizes of theK prior boxes to the feature maps of different scales for prediction and generating corresponding candidate boxes; S5, selecting a final prediction box according to the generated candidate box; S6, mapping the final prediction box to the original image according to the relationship between the feature map and the original image, and positioning the vehicle information in the image. According to the method, the distance calculation formula of the GT frame and the clustering center is improved, the sensitivity of the formula to the IOU value is reduced, the priori frame size obtained through clustering better conforms to the true value, the priori frame quality is improved, and therefore the performance of the YOLO v3 detection method is improved.

Description

technical field [0001] The invention belongs to the field of image target detection, in particular to a vehicle detection method based on an improved algorithm of YOLO v3. Background technique [0002] Object detection is an important building block in the field of computer vision. Among them, vehicle detection is a hot research issue in the field of target detection, and has important applications in assisted driving, road monitoring, and remote sensing images. The goal of vehicle detection is to quickly detect vehicle targets and their related feature information from pictures or videos. [0003] Traditional machine learning-based vehicle detection usually includes two steps: first, feature extraction is performed, and then based on the extracted feature vectors, it is classified using a classifier. Commonly used feature extraction methods include Local Binary Pattern (LBP), Histogram of Oriented Gradient (HOG), Scale Invariant Feature Transform (SIFT), etc. Commonly us...

AI Technical Summary

Problems solved by technology

Compared with the two-stage method, the single-stage target detection method has a faster detection speed, but the detection accuracy needs to be strengthened

[0005] For example, the Chinese patent whose application number is CN201910377894.9 discloses a method for constructing a pedestrian detection model. When the method is performing a clustering analysis step, the distance calculation formula is d(box, centroid)=1-IoU(box, centroid), but the formula is too sensitive to the IoU value, so the final pedestrian detection result is still not accurate enough

Images