Floating object target detection method and system fusing attention mechanism
A target detection and floating object technology, which is applied in the field of floating object detection integrating attention mechanism, can solve the problems of separation of target area and background area, inability to apply floating object detection, large parameters and computational overhead, etc., to improve detection accuracy , Enhance the characteristics of the target to be inspected, and improve the effect of attention
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0034] This embodiment provides a floating object detection method incorporating an attention mechanism, including:
[0035] Obtain the image information of the water flow environment; it can be understood that the obtained image information can be the static image information of the water flow surface or the new type of dynamic image; the dynamic image information here refers to the constantly switching images, or the continuous images extracted from the video information. The image can be understood as the collected water flow at different times in the same collection area, or multiple continuous / non-continuous images of water flowing through different areas;
[0036] According to the image information and the preset floating object detection model, the floating object detection result is obtained; it is understandable that the detection result and the output result when training the floating object detection model may be the floating object in the image. Location, judgment ...
Embodiment 2
[0059] In order to further illustrate Embodiment 1, this embodiment provides a floating object detection method fused with an attention mechanism, such as figure 1 As shown, this embodiment proposes a lightweight target detection algorithm CBAM-YOLOv4-tiny fused with attention mechanism, specifically:
[0060] The algorithm traverses the entire input sample image, extracts the feature information of the sample image through the constructed CBAM-YOLOv4-tiny model, and regresses the category and bounding box of the sample target;
[0061] The constructed CBAM-YOLOv4-tiny network mainly includes three parts: backbone network (backbone), neck structure (stem) and prediction end (head), wherein the backbone network (backbone) is a CSP-DarkNet53-tiny structure, the The stem is the FPN structure embedded in the CBAM module, and the prediction head is YOLO-Head.
[0062] The convolutional attention module CBAM introduced in this embodiment is as follows figure 2 shown, specifically...
Embodiment 3
[0085] This embodiment provides a floating object detection system incorporating an attention mechanism, including:
[0086] a data acquisition module, configured to: acquire image information of the water flow environment;
[0087] a detection module, configured to: obtain a floating object detection result according to the image information and a preset floating object target detection model;
[0088] Among them, the floating object detection model is trained by the YOLOv4-tiny model. In the YOLOv4-tiny model, CSPDarknet53-tiny is used as the feature extraction network, and the convolutional attention module is embedded in the feature pyramid structure; The mean clustering algorithm clusters and analyzes the size information of the floating objects. In the K-means clustering algorithm, the area intersection ratio is used as the criterion function.
[0089] The working method of the system is the same as that of the floating object detection method fused with the attention m...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More