Abstract: An action recognition method includes: obtaining original feature submaps of each of temporal frames on a plurality of convolutional channels by using a multi-channel convolutional layer; calculating, by using each of the temporal frames as a target temporal frame, motion information weights of the target temporal frame on the convolutional channels according to original feature submaps of the target temporal frame and original feature submaps of a next temporal frame, and obtaining motion information feature maps of the target temporal frame on the convolutional channels according to the motion information weights; performing temporal convolution on the motion information feature maps of the target temporal frame to obtain temporal motion feature maps of the target temporal frame; and recognizing an action type of a moving object in image data of the target temporal frame according to the temporal motion feature maps of the target temporal frame on the convolutional channels.

Abstract: A multisource volumetric spectral computed tomography imaging device includes an x-ray source array with multiple spatially distributed x-ray focal spots, an x-ray beam collimator with an array of apertures, each confining the radiation from a corresponding x-ray focal spot to illuminate a corresponding segment of an object, a digital area x-ray detector, and a gantry to rotate the x-ray source array and the detector around the object. An electronic control unit activates the radiations from the x-ray focal spots to scan the object multiple times as the gantry rotates around the object. The images are used to reconstruct a volumetric CT image of the object with reduced scattered radiation. For dual energy and multi energy imaging, radiation from each focal spot is filtered by a corresponding spectral filter to optimize its energy spectrum.