Research On Neutron Track Recognition And Application Based On Machine Vision

A small amount of neutrons escaping from the environment around the neutron source may cause damage to the human body and equipment.Therefore,it is particularly important to monitor the neutron irradiation intensity in the neutron irradiation environment and the neutron irradiation dose statistics of related workers..In order to realize the rapid and accurate detection of neutron radiation dose,two neutron track detection methods based on machine vision are proposed.The main research contents are as follows:(1)Through experiments,the preprocessing effect of common filtering algorithms on neutron track images is explored,and a set of preprocessing procedures for neutron track images are proposed.The neutron track detection algorithm.The detection speed of the neutron detection method based on morphological features reaches 2s/sheet,far exceeding the manual detection method.The regression coefficient R~2 of the irradiation intensity-net track density curve is 0.9866,which exceeds the manual detection of 0.9523.The detection error is 19.82%,which is less than 20%of the engineering requirements and meets the engineering needs.(2)For higher detection accuracy,we continue to explore the neutron track detection scheme based on deep learning.Through experimental comparison,it is determined that the Faster-RCNN model is more suitable for neutron track detection,and an intermediate feature layer is proposed for fusion,k-means clustering generates an anchor box size that is more suitable for the size of the neutron track,and the global pooling layer replaces the fully connected layer in the classification and regression network.Three-point improvement measures to improve the detection accuracy,and the improved Faster-RCNN detection accuracy increases The detection speed is increased by 6.39%,the detection speed is increased by 58.92%,the regression coefficient R~2 in the detection of neutron dose is 0.9441,and the error of neutron dose detection is 16.00%,which is reduced by 3.82%on the basis of neutron detection based on morphological features.(3)In order to facilitate the use of radiation protection personnel,the above two algorithms are integrated into a neutron track detection software,and the track image is rapidly detected through a graphical interface.It can be seen from the test results that the detection method based on morphological features proposed in this paper has obtained higher detection stability,and the method based on the improved Faster-RCNN has higher detection accuracy and smaller detection error.This paper provides a technical reference for the neutron track detection technology based on machine vision.