Abstract: An interframe registration and adaptive step size-based non-uniformity correction method for an infrared image, comprising: first calculating a normalized cross-power spectrum of n-th and (n?1)-th original infrared images with the non-uniformity, and then calculating a horizontal relative displacement and a vertical relative displacement of the n-th and (n?1)-th original infrared images with the non-uniformity; calculating a space variance and a time variance of each pixel of the n-th original infrared image with the non-uniformity, using the obtained space variance and time variance to calculate an adaptive iterative step size of each pixel of the n-th original infrared image with the non-uniformity, and using the iterative step size to update a gain correction coefficient and a bias correction coefficient; finally, performing non-uniformity correction on the pixel in an overlapping area of the n-th and (n?1)-th original infrared images with the non-uniformity.

Abstract: A detection and positioning method for a train water injection port includes: acquiring train water injection port video images, and performing threshold segmentation on the train water injection port video images to obtain binarized train water injection port video images; processing the binarized train water injection port video images and matching the processed binarized train water injection port video images with a train water injection port template image; detecting a position of the water injection port in the train water injection port video image and comparing the position with a pre-set position range where the water injection port is located; and if the position and the pre-set position range have been matched, then transmitting a matching valid signal to a mechanical device control module to control a mechanical device to move or stop and start or stop water injection.

Abstract: A feature fusion and dense connection-based method for infrared plane object detection includes: constructing an infrared image dataset containing an object to be recognized, calibrating a position and class of the object to be recognized in the infrared image dataset, and obtaining an original known label image; dividing the infrared image dataset into a training set and a validation set; performing image enhancement preprocessing on images in the training set, performing feature extraction and feature fusion, and obtaining classification results and bounding boxes through a regression network; calculating a loss function according to the classification results and the bounding boxes in combination with the original known label image, and updating parameter values of a convolutional neural network; repeating the steps to iteratively update the parameters of the convolutional neural network; and processing images in the validation set through the parameters to obtain a final object detection result map.

Abstract: A detection and positioning method for a train water injection port includes: acquiring train water injection port video images, and performing threshold segmentation on the train water injection port video images to obtain binarized train water injection port video images; processing the binarized train water injection port video images and matching the processed binarized train water injection port video images with a train water injection port template image; detecting a position of the water injection port in the train water injection port video image and comparing the position with a pre-set position range where the water injection port is located; and if the position and the pre-set position range have been matched, then transmitting a matching valid signal to a mechanical device control module to control a mechanical device to move or stop and start or stop water injection.

Abstract: A feature fusion and dense connection-based method for infrared plane object detection includes: constructing an infrared image dataset containing an object to be recognized, calibrating a position and class of the object to be recognized in the infrared image dataset, and obtaining an original known label image; dividing the infrared image dataset into a training set and a validation set; performing image enhancement preprocessing on images in the training set, performing feature extraction and feature fusion, and obtaining classification results and bounding boxes through a regression network; calculating a loss function according to the classification results and the bounding boxes in combination with the original known label image, and updating parameter values of a convolutional neural network; repeating the steps to iteratively update the parameters of the convolutional neural network; and processing images in the validation set through the parameters to obtain a final object detection result map.

Abstract: An interframe registration and adaptive step size-based non-uniformity correction method for an infrared image, comprising: first calculating a normalized cross-power spectrum of n-th and (n?1)-th original infrared images with the non-uniformity, and then calculating a horizontal relative displacement and a vertical relative displacement of the n-th and (n?1)-th original infrared images with the non-uniformity; calculating a space variance and a time variance of each pixel of the n-th original infrared image with the non-uniformity, using the obtained space variance and time variance to calculate an adaptive iterative step size of each pixel of the n-th original infrared image with the non-uniformity, and using the iterative step size to update a gain correction coefficient and a bias correction coefficient; finally, performing non-uniformity correction on the pixel in an overlapping area of the n-th and (n?1)-th original infrared images with the non-uniformity.

Abstract: A non-uniformity correction method for an infrared image based on guided and high-pass filtering includes: assigning high-frequency component of first image frame of original image sequence to first fixed pattern noise f1; successively loading N-th image frame of the original image sequence with non-uniformity as current image frame, determining a difference between the current image frame and (N?1)-th image frame to obtain (N?1)-th differential image frame, and obtaining a relative change amplitude of each image element of (N?1)-th image frame according to the (N?1)-th differential image frame; and performing high-pass filtering based on a combination of a high-frequency component of the current image frame obtained through guided filtering and the relative change amplitude to obtain n-th fixed pattern noise fn, and performing non-uniformity correction on the current image frame according to fn to obtain a correction result of the current image frame, where N?2, n?2.

Abstract: A non-uniformity correction method for an infrared image based on guided and high-pass filtering includes: assigning high-frequency component of first image frame of original image sequence to first fixed pattern noise f1; successively loading N-th image frame of the original image sequence with non-uniformity as current image frame, determining a difference between the current image frame and (N-1)-th image frame to obtain (N-1)-th differential image frame, and obtaining a relative change amplitude of each image element of (N-1)-th image frame according to the (N-1)-th differential image frame; and performing high-pass filtering based on a combination of a high-frequency component of the current image frame obtained through guided filtering and the relative change amplitude to obtain n-th fixed pattern noise fn, and performing non-uniformity correction on the current image frame according to fn to obtain a correction result of the current image frame, where N?2, n?2.