Remote sensing image special vehicle target detection method based on transfer learning

A target detection and special vehicle technology, applied in the field of remote sensing image vehicle target detection, can solve the problems of low detection accuracy, time-consuming and labor-intensive, and high cost, and achieve the effects of improving detection accuracy, improving convergence speed, and improving success rate.

CN111898523AInactive Publication Date: 2020-11-06UNIV OF ELECTRONICS SCI & TECH OF CHINA
9 Cites 22 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Publication Date
2020-11-06
Estimated Expiration
Not applicable · inactive patent

Smart Images

  • Figure 1
    Figure 1
  • Figure 2
    Figure 2
  • Figure 3
    Figure 3
Patent Text Reader

Abstract

The invention provides a remote sensing image special vehicle target detection method based on transfer learning, and the method comprises the following steps: firstly obtaining vehicle image data, and obtaining training data according to the vehicle image data; inputting the training data to a ResNet network for transfer learning to obtain a target detection network; and finally, inputting to-be-detected data to the target detection network, and outputting a detection result by the target detection network. According to the invention, the convergence speed of the residual network is improved,the target detection network outputs the detection result, the detection precision is effectively improved, and the cost is reduced.
Need to check novelty before this filing date? Find Prior Art

Description

technical field

[0001] The invention belongs to the field of earth vision target recognition, and relates to a vehicle target detection method of a remote sensing image. Background technique

[0002] Object recognition has always been a research hotspot in the field of computer vision. Currently, deep learning is widely used in the field of object recognition. The characteristic of deep learning is that the larger the sample set, the more accurate the data set can be trained. However, in the process of actual research, it is difficult to obtain a large amount of data, and the method of manual labeling of images is time-consuming and laborious. The training set and test set must meet Defects such as the same distribution will make it difficult to obtain large data sets. The expiration of training samples and the lack of training data often make it necessary to re-collect labeled data, but if the amount of data is large, it will be time-consuming, labor-intensive, and costly....

Examples

Embodiment Construction

[0039] The present invention will be further described below in conjunction with the accompanying drawings and embodiments, and the present invention includes but not limited to the following embodiments.

[0040] The present invention comprises the following steps:

[0041] S1: Acquire vehicle image data. The present invention collects a total of 207 images of special vehicles and common vehicle images as reference;

[0042] S2: Obtain training data according to the vehicle image data;

[0043]S3: Input the training data to the ResNet network for transfer learning to obtain a target detection network;

[0044] S4: Input the data to be detected to the target detection network, and the target detection network outputs a detection result.

[0045] Wherein, step S2 specifically includes:

[0046] S201: Filter the vehicle image data to obtain a first data set, where the first data set is a plurality of image data with a single vehicle type;

[0047] S202: Perform preprocessing...