Method for detecting dangerous goods by explosive-handling robot based on self-adaptive spatial feature fusion

Abstract

The invention discloses a method for detecting dangerous goods by an explosive-handling robot based on self-adaptive spatial feature fusion, and the method comprises the steps: capturing data from the Internet to obtain an initial hazardous article data set, cleaning the initial hazardous article data set, then marking the initial hazardous article data set, and finally carrying out the data enhancement of the marked data set, thereby obtaining a hazardous article data set; using a dangerous object data set to training a dangerous object detection model, wherein the training process is the process of taking an object image of a known category as input, taking the probability of the corresponding category of the object as theoretical output and continuously adjusting model parameters, wherein the training termination condition is that the upper limit of the training frequency is reached, and the dangerous goods detection model takes MobileNetV3 as a backbone extraction network; adopting a YOLOV4 model of an ASFF feature fusion strategy; and obtaining a picture of to-be-detected goods collected by the explosive ordnance disposal robot, and inputting the picture into the dangerous goods detection model to obtain the category of the to-be-detected goods. The method of the invention has the advantages of high accuracy and efficiency of detecting the dangerous goods and good robustness.

Description

technical field [0001] The invention belongs to the technical field of visual detection, and relates to a dangerous goods detection method of an explosion-proof robot based on adaptive spatial feature fusion and its application. Background technique [0002] In recent years, robot technology has advanced by leaps and bounds, and robots have been used in industrial production, rescue and disaster relief and other fields. EOD robots, as robots in the security field, have attracted a lot of attention. The EOD robots currently in use mainly rely on the video images returned by the robot's computer vision system, and then the EOD personnel remotely control the EOD robots through keyboards or remote controls to complete operations such as grasping, transferring, and detonating explosives. The efficiency and efficiency will be affected by factors such as the quality of the returned video image, the proficiency of the EOD personnel on the equipment, and the time error between the re...

AI Technical Summary

Problems solved by technology

[0007] The purpose of the present invention is to overcome the defects that the existing deep learning algorithm used for the detection of dangerous objects in explosive robots has no dangerous object data sets available, the detection accuracy is low, and the detection rate is slow, and to provide a method that can obtain a large number of dangerous object data sets and detect High precision and fast detection speed detection method for dangerous goods of EOD robot

Images