To address the challenges of insufficient feature extraction capability and class imbalance in datasets faced by traditional convolutional neural networks when classifying spectrum images of impact echo signals,we propose a model of Multi-scale Hybrid Attention and Spatial Channel Reconstruction Convolutional Neural Network(MHA-SCConvNet).The model initiates with a Multi-scale Hybrid Attention(MHA) module that extracts spectral features at diverse scales,significantly enhancing the focus on essential waveform information.This is followed by the integration of a Spatial and Channel Reconstruction Convolution(SCConv) module,designed to optimize the representation of image features and effectively reduce redundancy.Furthermore,we introduce the Gradient and Distribution Harmonized Margin Loss(GDHM Loss),a dynamic hybrid loss function tailored to address the challenges of classifying both hard-to-classify and minority class samples.Rigorous evaluation of the proposed MHA-SCConvNet on a proprietary dataset demonstrates a remarkable accuracy of 94.58%,outperforming established models such as AlexNet,VGGNet,and GoogLeNet.These experimental results validate the MHA-SCConvNet's superior capability in enhancing the accuracy and efficiency of impact echo spectral image classification.