The invention relates to a seismic acquisition data compression and transmission method so that lossless data compression is realized, transmission quantity of the data is reduced and the data is ensured not to be lost. The method comprises the steps: 1) when the number of sampling points in each frame of data is certain, data lengths of different acquisition units are variable; 2) the sampling points of each frame are variable, but the sampling points contained by data frames corresponding to all the acquisition units at the same time in the same system are certain; 3) the first six bytes in each frame are fixed, and the seventh byte is a mark byte which records a transmission frame mark; and a sampling input retaining circuit ensures that 0 and the maximum value do not appear adjacently; 4) the bytes are marked, three bits are utilized to express transmission digits, and corresponding data are 8-24 bits; 5) each sampling point is at least 8 bits, with a progressive step of 4 bits, and 24 bits at most; 6) a non-volatile data memory is additionally arranged in an acquisition unit so that the acquisition data is ensured not to be lost when the data cannot be transmitted in real time; and 7) the top digit of each stored sampling point must be a sign digit so that polarity of the acquisition data is ensured not to be changed during compression.