Wild mushrooms are popular for their taste and nutritional value; however, non-experts often struggle to distinguish between toxic and non-toxic species when foraging in the wild, potentially leading to poisoning incidents. To address this issue, this study proposes a compact bilinear neural network method based on Transformer and multi-scale feature fusion. The method utilizes a dual-stream structure that integrates multiple feature extractors, enhancing the comprehensiveness of image information capture. Additionally, bottleneck attention and efficient multi-scale attention modules are embedded to effectively capture multi-scale features while maintaining low computational costs. By employing a compact bilinear pooling module, the model achieves high-order feature interactions, reducing the number of parameters without compromising performance. Experimental results demonstrate that the proposed method achieves an accuracy of 98.03%, outperforming existing comparative methods. This proves the superior recognition performance of the model, making it more reliable in distinguishing wild mushrooms while capturing key information from multiple dimensions, enabling it to better handle complex scenarios. Furthermore, the development of public-facing identification tools based on this method could help reduce the risk of poisoning incidents. Building on these findings, the study suggests strengthening the research and development of digital agricultural technologies, promoting the application of intelligent recognition technologies in agriculture, and providing technical support for agricultural production and resource management through digital platforms. This would provide a theoretical foundation for the innovation of digital agriculture and promote its sustainable development.

Rational utilization of soil total nitrogen is one of the keys to achieving sustainable agricultural development. By accurately measuring the content of total nitrogen in the soil, the utilization efficiency of nitrogen in the soil can be improved, and the scientific use of chemical fertilizers can reduce the pressure of agriculture on natural resources and realize the sustainable development of agriculture. In order to measure soil total nitrogen content simply and accurately, combined with the method of artificial olfactory systems, a new method of soil total nitrogen content detection based on convolutional noise reduction autoencoder (CDAE)–whale optimization algorithm (WOA)–deep residual shrinkage network (DSRN) is proposed. In order to obtain more salient features for fusion, the channel mechanism of the DSRN is improved by adding global Max pooling. The model uses a CDAE for the first filtering stage to automatically obtain data that filters simple noise and uses the WOA to automatically optimize hyperparameters. Finally, the optimized hyperparameters were used to train the DRSN for secondary filtering and predict the soil total nitrogen content. Experimental results show that the R² of CAE-WOA-DSRN test set is 0.968, which is significantly better than the R² of a traditional algorithm (0.873) and a simple BP network (0.877), and it can more accurately measure soil total nitrogen content.