Embedded platform-oriented dual-mode target detection method and system

A target detection and embedded technology, which is applied in the field of computer communication, can solve the problems of difficulty in implementing embedded platform transplantation algorithms, and cannot adapt to the requirements of real-time and portability of target detection algorithms, achieving a good balance between accuracy and speed and reducing requirements. , the effect of improving the robustness

Pending Publication Date: 2021-08-13
青岛以萨数据技术有限公司
View PDF0 Cites 3 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] Although some visible light target detection models can achieve high accuracy, due to the limitation of computing power and memory resources of the embedded platform, it is

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 more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Embedded platform-oriented dual-mode target detection method and system
  • Embedded platform-oriented dual-mode target detection method and system
  • Embedded platform-oriented dual-mode target detection method and system

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0085] 1. MNYOLO target detection network

[0086] 1.1 Detection principle

[0087] The MNYOLO network continues to use the idea of ​​YOLO target detection. The entire image is used as the input of the entire network. It does not need to generate a suggestion area. In the output layer, the regression idea is used to obtain the position and category of the bounding box, and then the non-maximum value suppression algorithm is used to remove it. The redundant bounding box is used to get the final prediction result. The whole process is that the detection network directly performs end-to-end prediction, and the detection speed is relatively high.

[0088] The YOLOv4 algorithm optimizes the YOLOv3 model from the perspectives of data preprocessing, backbone network, training method, activation function, etc., so that the detection model achieves a good balance between detection speed and detection accuracy. The YOLOv4 backbone network CSPDarkNet absorbs the advantages of CSPNet (C...

Embodiment 2

[0136] Based on the same technical idea, the present invention also provides a dual-mode target detection system for embedded platforms, including:

[0137] An acquisition module, configured to acquire an image to be identified;

[0138] The feature extraction module is used to segment the image to be recognized according to a preset segmentation algorithm to obtain at least one image frame; perform feature extraction on the image frame through the depth separable convolution in the improved YOLOv4-Tiny network, obtaining the image features corresponding to the image frame;

[0139] The target detection module is used to determine the infrared image and the visible light image according to the image characteristics of the image to be recognized, register the collected infrared image and the visible light image, input the predefined visible light target detection model and the infrared target detection model respectively, and output the visible light and the detection results ...

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

No PUM Login to view more

Abstract

The invention provides an embedded platform-oriented dual-mode target detection method and system, and the method comprises the steps: carrying out the segmentation processing on a to-be-recognized image according to a preset segmentation algorithm, and obtaining at least one image box; performing feature extraction on the image box through depth separable convolution in the improved YOLOv4-Tiny network to obtain image features corresponding to the image box; determining an infrared image and a visible light image according to the image features, respectively inputting the same a predefined visible light target detection model and an infrared target detection model after registration, and outputting detection results of the visible light target detection model and the infrared target detection model; and performing decision-level image fusion on the detection results through a pre-established decision-level fusion detection model, so that infrared and visible-light multi-band information complementation is realized, and a dual-mode target detection result is obtained. According to the scheme, real-time detection of an embedded platform for a multi-source image is realized, and the method has more obvious advantages in detection precision compared with a single-band target detection model while the requirements of real-time performance and effectiveness are met.

Description

technical field [0001] The invention relates to the technical field of computer communication, in particular to an embedded platform-oriented dual-mode target detection method and system. Background technique [0002] Target detection is an important research content in the field of computer vision. With the rapid development of deep learning, new detection algorithms are emerging in the visible light environment, which are mainly divided into two-stage detection models and one-stage detection models. The Two-stage detection model mainly includes the R-CNN series algorithm, which greatly improves the detection accuracy by generating suggested regions; the One-stage detection model mainly includes the SSD (Single Shot MultiBox Detector) series, YOLO (YouOnly Look Once) series, etc., using A one-step framework for global regression and classification, while sacrificing certain accuracy, the detection speed is greatly improved. The above two detection models are based on prese...

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
no application Login to view more
IPC IPC(8): G06K9/00G06K9/32G06K9/46G06K9/62G06N3/04G06N3/08G06T5/50G06T7/11G06T7/33
CPCG06T5/50G06T7/11G06T7/33G06N3/08G06T2207/10048G06T2207/20081G06T2207/20084G06T2207/20104G06T2207/20221G06V20/00G06V10/25G06V10/40G06V2201/07G06N3/045G06F18/214G06F18/25G06F18/241
Inventor 孙海铭李凡平石柱国
Owner 青岛以萨数据技术有限公司
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap