Intelligent detection method for beacon light quality based on machine vision

Abstract

The invention discloses an intelligent detection method for beacon light quality based on machine vision. The method is characterized by comprising the following steps of: firstly, performing complex environment simulation on image frame data of a beacon light video based on gamma correction and a median filtering method; then subjecting the whole image data to target detection under a Yolov4 deep learning algorithm and to retraining based on transfer learning to obtain a network model with accurate detection under a small amount of data; and then conducting training by using a SiamFC deep learning method to obtain a twin neural network model which is used for real-time tracking on a target area, carrying out automatic correction and judgment regularly in a tracking process, then extracting corresponding beacon light signals for color acquisition mapping, color change time recording and calculation of corresponding cycle, frequency and light intensity characteristics. According to the invention, the working requirements of on-site detection of long-distance beacon light equipment and corresponding light quality parameter acquisition and result analysis can be met.

Description

technical field [0001] The present invention relates to the technical fields of digital image processing, machine learning, and deep learning, and specifically relates to a machine vision-based intelligent detection method for light quality of navigation lights, which can combine the color space of video images with a recognition and tracking network model, and in In a complex environment, it can dynamically and autonomously track and detect the light quality of navigation lights in real time. Background technique [0002] Navigation marks are very important signs to help guide ships to navigate, locate and mark obstructions and express warnings. At present, the detection of navigation marks mainly relies on laboratory testing, and the staff need to go to the sea regularly to bring the navigation lights back to the shore laboratory for testing, so as to determine whether there is a problem with the navigation lights. Due to the large size of the navigation light itself and ...

AI Technical Summary

Problems solved by technology

There are the following deficiencies: the amount of experimental training data is small, and only more than 70 image data of navigation lights have been collected and produced; and for the time being, only two known light qualities of navigation lights can be detected, specifically: 1s and B lights are on for 1s and off for 3s, and finally only the effect of visual beacon light monitoring is realized

There are the following deficiencies: only a module for light quality acquisition photoelectric conversion is proposed, by obtaining the complete flash signal of the beacon light, and passing through the light splitting unit, it is divided into two channels according to a certain ratio, a small part enters the electronic observation unit, and most of the rest pass through Optical fiber transmission enters the spectral analysis unit for detection

However, in the specific design, the needs of identification, collection, tracking and detection of offshore beacon lights in different environments have not been addressed in a targeted manner

Images