An Image Detection Method for Image Stabilization Deviation of the Tank Gunner's Primary Sight

Area of Science:

• Military Engineering
• Computer Vision
• Robotics

Background:

• Image stabilization is crucial for tank gunner primary sight control systems.
• Accurate measurement of image stabilization deviation is key to assessing system performance.
• Existing methods may lack the precision required for effective evaluation.

Purpose of the Study:

• To develop an advanced image detection algorithm for measuring sight stabilization deviation in tank gunner systems.
• To enhance the accuracy and effectiveness of image stabilization evaluation.
• To contribute to the development of parameter measurement systems for gunner sights.

Main Methods:

• An enhanced You Only Look Once version 5 (YOLOv5) algorithm was developed.
• A novel loss function, delta-SIOU (δ-SIOU), replaced CIoU.
• Enhancements included integrating a dynamic weight factor into SIOU, improving the Spatial Pyramid Pool module, and incorporating the C3CA module with Coordinate Attention (CA).
• The Bi-directional Feature Pyramid (BiFPN) network was integrated into the YOLOv5 Neck.

Main Results:

• The enhanced YOLOv5 algorithm demonstrated improved multi-scale feature fusion and target location information learning.
• Experimental results showed a 2.1% increase in detection accuracy.
• The proposed method effectively measures image stabilization deviation in the aiming line.

Conclusions:

• The developed image detection method offers a more accurate approach to measuring image stabilization deviation.
• This research facilitates the advancement of parameter measurement systems for tank gunner primary sights.
• The findings support improved operational status evaluation of tank gunner control systems.