Abstract: The moisture and oil content of Xanthoceras sorbifolia Bunge affects the results of its breeding and processing and storage. This study aims to explore a non-destructive, rapid and accurate method for the detection of moisture and oil content of Xanthoceras sorbifolia Bunge, provide excellent seeds for oil extraction and breeding of Xanthoceras sorbifolia Bunge. During the detection process of low-field nuclear magnetic resonance (LF-NMR), the peak overlapping phenomenon of Xnthoceras sorbifolia Bunge was found. There are two main ways to validate and solve the problem of overlapping moisture and oil signals in food products. One is to use LF-NMR combined with dry sampling for removing water to eliminate the water signal by drying to evaporate the water to eliminate the overlapping area to measure the oil content. The other is to use low field two-dimensional nuclear magnetic resonance (LF-2D-NMR) to separate the oil and moisture signals. A series of experiments were conducted to determine the feasibility of the two techniques in detecting the moisture and oil content of Xanthoceras sorbifolia Bunge. The national standard oven drying method was used to detect moisture and Soxhlet extraction method was used to detect oil, and the results were used as benchmark values to verify the accuracy of the results obtained by the nondestructive testing methods. The test results show that the data obtained by LF-NMR in combination with dry sampling is anomalous compared to the results of the national standard method. Matlab was used to extract the eigenvalues of each color of Lab in the MRI pseudo-color map. The analysis revealed that the signals were shifted from outside to inside during the drying process, and the comparative analysis with the physicochemical test revealed that abnormal changes occurred inside the Xanthoceras sorbifolia Bunge. LF-NMR encountered difficulties in trying to differentiate and quantify overlapping NMR signals in the samples, and the use of dry sampling to solve the problem resulted in anomalous variations and the inability to reasonably account for the anomalous variations behind these signals using LF-NMR, which limited its applicability in the Xanthoceras sorbifolia Bunge applicability in moisture and oil content detection, so the method was not suitable for detecting the oil and moisture content of Xanthoceras sorbifolia Bunge, while the LF-2D-NMR was able to analyze the overlapping signals in the one-dimensional spectra qualitatively, and successfully explained the problem of the overlapping signals in the one-dimensional spectrum. It was found that drying induced molecular rupture in the internal molecules of Xanthoceras sorbifolia Bunge, the saturated fatty acid and carbonyl values rose, and new peaks were formed by generating hazardous substances, which explained the anomalies in the LF-NMR data. Meanwhile, there was a significant linear correlation between the peak areas of the T₁-T₂ spectra of Xanthoceras sorbifolia Bunge measured by LF-2D-NMR and the moisture and oil contents obtained by the oven drying method and Soxhlet extraction method, correlation coefficient of R² after verification with the true value being 0.9209 and 0.9424, respectively, and therefore this technique can be used for quantitative analysis of the moisture and oil containing contents of Xanthoceras sorbifolia Bunge. Thus, the LF-2D-NMR technique shows great potential for quantitative analysis of the water and oil content of Xanthoceras sorbifolia Bunge. This study expands the theoretical cognitive depth of the LF-NMR technique, and improves the detection accuracy and the guiding value in practical production.