【摘要】 针对地震数据在压缩与重建过程中部分高频和峰值信息丢失的问题，结合小波变换（WT）在多分辨率分析中的优势和卷积自编码器（CAE）在特征提取和数据重建方面的高效能力，提出了一种基于WT改进CAE的地震道数据压缩方法。该方法构建了两个改进的CAE模型：低压缩比模型WTCAE-L，高压缩比模型WTCAE-H，实现了对地震数据的高效压缩，同时保持了较高的重建质量。实验结果表明，两者在各自压缩比范围内展现最佳性能。更多还原

【Abstract】 To address the issue of partial loss of high-frequency and peak information during the compression and reconstruction of seismic data, a seismic trace data compression method based on an improved convolutional autoencoder(CAE) enhanced by wavelet transform(WT) was proposed, leveraging the advantages of WT in multi-resolution analysis and the efficient capabilities of CAE in feature extraction and data reconstruction. Two improved CAE models were constructed using the method: a lowcompression-ratio model, WTCAE-L, and a high-compression-ratio model, WTCAE-H, which enabled efficient compression of seismic data while maintaining high reconstruction quality. Experimental results show that both models exhibit optimal performance within their respective compression ratio ranges.更多还原