Method and device for detecting leakage target by integrating temporal information

A system for detecting a leakage target by integrating temporal information, comprising:

• an acquisition module used for acquiring multiple frames of images of a target position and arranging the multiple frames of images based on temporal information corresponding to the multiple frames of images to obtain a target image sequence;
• a multi-dimensional tensor generation module used for merging, in a channel dimension, the multiple frames of images in the target image sequence to obtain a multi-dimensional tensor;
• a convolution processing module used for performing convolution processing on the multi-dimensional tensor to obtain a first output result, wherein the first output result is used for representing a frame difference between every two adjacent ones of the multiple frames of images;
• an accumulated frame difference calculation module used for calculating an accumulated frame difference corresponding to the multiple frames of images based on the first output result;
• and
• a target detection module used for inputting the accumulated frame difference to a trained target detection model as input data to obtain a second output result integrated with the temporal information, wherein the second output result is used for reflecting a probability of the presence of the leakage target at the target position and a class of the leakage target;
• wherein
• when calculating the accumulated frame difference corresponding to the multiple frames of images based on the first output result, the accumulated frame difference calculation module is specifically used for calculating an accumulated channel difference corresponding to each of channels based on the frame difference between every two adjacent ones of the multiple frames of images;
• and obtaining accumulated frame difference by summing the accumulated channel differences corresponding to the channels, wherein the accumulated channel difference is used for representing a sum of absolute values of differences of every two adjacent ones of the multiple frames of images in a channel.