Small target detection method and device based on improved Fast RCNN, and storage medium

Abstract

The invention relates to a small target detection method based on an improved Fast RCNN, which is realized by executing an improved Fast RCNN algorithm instruction by a processor, and comprises the following steps: receiving a scene picture containing a small target and extracting a first feature map F; obtaining a predicted anchor point frame a (x, y, w, h) according to the first feature map F1; obtaining a second feature map F2 with the same size as the first feature map F1 according to the first feature map F1 and the predicted anchor frame a (x, y, w, h); and obtaining a detection result of the scene picture according to the second feature map F2 and the predicted anchor frame a (x, y, w, h). According to the invention, modification is carried out based on a Fast RCNN algorithm framework, and the RPN network under the Faster RCNN algorithm framework is replaced with the adaptive anchor frame network, so that the generated anchor frame can be better matched with targets with different scales, thereby avoiding missing detection caused by unreasonable size of the anchor frame, and improving the detection accuracy.

Description

technical field [0001] The invention relates to the field of target recognition, in particular to a small target detection method based on improved Faser RCNN. At the same time, the present invention also relates to a small target detection device and storage medium based on the improved Faser RCNN. Background technique [0002] Object detection, also called object extraction, is an image segmentation based on the geometric and statistical characteristics of the object. It combines target segmentation and recognition into one, and its accuracy and real-time performance are an important capability of the whole system. [0003] Object detection is a popular direction in computer vision and digital image processing. It is widely used in robot navigation, intelligent video surveillance, industrial inspection, aerospace and many other fields. It has important practical significance to reduce the consumption of human capital through computer vision. Therefore, target detection h...

AI Technical Summary

Problems solved by technology

[0005] However, in real-world applications, the results of target detection often show that the detection effect of small targets is far worse than that of large and medium targets.

This is because there are two problems in small target detection: ① lack of information, that is, the proportion of the target in the image is very small, and the amount of information reflected by the pixels in the corresponding area is very limited; ② the amount of data is scarce, that is, the data set contains small There are few images of the target, resulting in an imbalance in the categories of the entire training set, resulting in the detection accuracy of small target objects being much lower than that of medium and large objects

However, this method is simple and rude, the operation is complicated, the amount of calculation is too large, and the practical significance is not strong.

[0007] ②Use the GAN model to enlarge the small target and then detect it. This method is consistent with the idea of ​​image data amplification, but it also has the disadvantage of complicated operation.

[0008] ③Modify the parameters of model training, such as setting the parameter stride to 1, but the effect of this method is also general

However, if the setting size of the proposals is unreasonable, it is easy to cause the problem of missed detection, especially in the case of a large difference between the two target scales, it is very easy to cause the problem of missed detection of small-scale targets, thereby reducing the detection accuracy.

Images