A bullet hole recognition method based on deep learning

Abstract

The invention discloses a bullet hole identification method based on deep learning. The method comprises the following steps: Step 1, model building; Step 2, data collection; Step 3, data processing and labeling; Step 4, model training; Step 5, Bullet hole identification. The identification system used in this method has very low requirements on the target surface; the probability of false detection, missed detection, and re-inspection is below 1%, and the accuracy of bullet hole identification is high, meeting the requirements for use; the identification system is used in this method to detect The delay is less than 40ms, and the detection results are displayed synchronously on the real-time video stream; in the case of dense bullet holes, the overlap of bullet holes is less than 50%, that is, bullet holes and overlapping bullet holes can be distinguished; when the target paper shakes relative to the camera, the center point alignment algorithm is used , so that the relative position of the bullet hole and the center remains unchanged, which solves the influence of the target paper shaking on the detection; uses the fuzzy detection algorithm to filter out the blurred picture frames when the bullet hits, thus effectively solving the misidentification caused by the blurred picture.

Description

technical field [0001] The invention relates to the field of shooting training, and particularly provides a method for identifying bullet holes based on deep learning. Background technique [0002] Live ammunition shooting is a basic training and assessment item for public security, armed police, military and other departments. At present, the target reporting method of these departments mainly adopts manual target reporting, that is, after the shooting is completed, the impact point of the target paper on the target surface is visually observed, and the impact point is counted one by one. Make a target. This method of target reporting affects the efficiency of shooting training, requires higher quality of target reporting personnel, and cannot report the number of rings in real time. [0003] In the prior art, most of the identification methods of bullet holes use traditional image processing methods to calculate the difference between two adjacent frames of images, use an...

AI Technical Summary

Problems solved by technology

However, this method will cause many false detections and missed detections in the case of large virtual targets, blurred images, and continuous holes caused by large target position shaking.

Images