In the field of capacitance tomography reconstruction (ECT) post-processing, aiming at the over-segmentation problem of traditional watershed algorithm in image segmentation, an improved watershed algorithm based on morphological markers and MSRM was proposed. Through the open and closed operation reconstruction of the graph, Laplace is used to sharpen the image boundary, and then the target and background are marked to modify the amplitude image. The image is roughly segmented by watershed transformation, and then the MSRM region combination method is used to effectively segment the target and background. Compared with the traditional segmentation algorithm, Otsu algorithm and the iterative algorithm based on the minimum error threshold method, it is found that the segmentation accuracy of the algorithm has obvious advantages.