The invention discloses a sequence image relative
pose estimation method, device and equipment and a medium, and relates to the technical field of
image processing, the method accurately estimates inter-frame poses through
deep learning, so that ultrasonic three-dimensional reconstruction can be realized under the condition of not using external tracking equipment such as
electromagnetic tracking and the like; therefore, various possible complex medical environments can be adapted more flexibly, and medical workers can be helped to perform examination and diagnosis better. And meanwhile, the method eliminates the dependence of ultrasonic three-dimensional reconstruction on external equipment such as optical /
magnetic tracking equipment, so that the
system architecture is simpler and more compact, the hardware cost and complexity are remarkably reduced, and the method is favorable for adapting to various different clinical application scenes. Besides, through rotation isovariant
convolution and multi-scale cross attention, the method can better adapt to factors such as angle change and
tissue deformation in the acquisition process, and the robustness of the method is improved.