A lightweight detection method for small targets in UAV images
A small target detection and unmanned aerial vehicle technology, applied in neural learning methods, computer parts, instruments, etc., can solve the problems of low calculation and parameter amount of target detection algorithm, limited storage and computing resources, etc., to reduce errors. Inspection, optimization of inspection results, accurate inspection effect
Active Publication Date: 2022-05-27
BEIHANG UNIV
View PDF5 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
However, the storage and computing resources of the UAV platform are limited, and the target detection algorithm needs to have a low amount of calculation and parameters.
Method used
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0018] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments.
[0019] The method of the present invention is designed based on the typical framework of the center point prediction method, such as figure 1 shown. The input image (Image) is scaled (Resize) and then input to the Feature Extractor (Feature Extractor) for feature extraction. The extracted features are up-sampling (Up-sampling) and then input to the Detection Head (Detection Head) for the target center point (Center Point). ) and Scale predictions. The feature extractor is usually based on the structure of the classification network, and the feature extractor and the upsampling module together constitute the backbone network based on the center point prediction framework. The detection head is responsible for target prediction based on the extracted features, and includes three branches: center point branch, center point offset branch and scale...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More PUM
Login to View More Abstract
The invention provides a lightweight UAV image small target detection method, which belongs to the technical field of UAV image processing. The present invention processes each frame of images according to the time sequence of the input image and video of the unmanned aerial vehicle to be tested, including: scaling the image and inputting it into a Revised MobileNetV2 feature extractor to output a feature map; inputting the feature map into a synchronous upsampling and detection module to obtain The position of the target center point and the corresponding scale, and all the predicted target bounding boxes in the frame are obtained; after all the frames of the video to be tested are processed, the fast sequence non-maximum suppression processing is performed on the prediction results of all frames, and the target detection results of the video to be tested are output. . The invention uses a lightweight backbone network to detect small targets in UAV images, reduces false detection of small targets, improves detection efficiency, and can realize fast and accurate detection of small targets in UAV images and videos.
Description
technical field [0001] The invention belongs to the technical field of UAV image processing, and in particular relates to a lightweight UAV image small target detection method. Background technique [0002] With the maturity of UAV technology and the increase in the number of UAV suppliers, the cost of UAVs has gradually decreased. In recent years, UAVs have attracted extensive attention in many fields such as geology, agriculture and forestry, and monitoring of people / vehicle flow. . The drone itself can carry a variety of peripheral sensors, including infrared image sensors, visible light image sensors, acceleration sensors, air pressure sensors, etc. Among them, visible light image sensors can provide rich environmental information. Therefore, the visible light image understanding technology of drones is One of the hottest areas of UAV application research. Among them, the target detection technology can locate the target of interest category in the image, which can und...
Claims
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More Application Information
Patent Timeline
Login to View More Patent Type & Authority Patents(China)
IPC IPC(8): G06V20/13G06V10/32G06V10/44G06V10/764G06V10/82G06K9/62G06N3/04G06N3/08
CPCG06N3/08G06V20/13G06V10/32G06V10/44G06N3/045G06F18/241
Inventor 李红光王蒙丁文锐
Owner BEIHANG UNIV