The invention discloses a self-adaptive fluorescent immune layer quantitative detection
feature extraction method and
system, and relates to the technical field of fluorescent immune detection.The method comprises the steps that a fluorescent speckle
image sequence is obtained, an
image gradient difference value between adjacent frames is calculated, a three-dimensional interference disturbance
tensor is constructed, and a main disturbance
modal map is obtained through sparse
principal component analysis; constructing a self-supervised triple training sample based on the main disturbance
modal map, and training an image block
encoder by taking a disturbance value as a pixel-level weighting factor; extracting image block features by using the trained
encoder, and matching the image block features with the response feature dictionary to generate a response probability map; calculating perturbation spectrum entropies of the
time sequence maps, and generating a fusion response probability map according to reciprocal weighted fusion of the perturbation spectrum entropies; and through
mask screening and response integral calculation, outputting through a fitting method. According to the method, the local immune response recognition capability can be effectively enhanced, the adaptivity of
feature extraction and the stability of response fusion under interference are improved, and high-precision quantification of
fluorescence immunodetection is realized.