Traditional Raman spectroscopy is highly susceptible to fluorescence and Raman scattering of the container wall when detecting unknown samples in containers, which often limits its commercial applications to transparent plastic or glass packaging. Since the photon migration direction inside the medium is random, the Raman scattered photons generated at the inner deep layer are more likely to migrate laterally during the diffusion process. Therefore, the Raman spectrum at different distances from the laser incident point contains different Raman spectral information of depth layers. The spatially offset Raman spectroscopy (SORS) can suppress the fluorescence and Raman scattering interference of the container wall by deviating the Raman light collection point from the laser incident point, there by realizing effective detection of the sample in the colored and opaque package. By designing a SORS experimental device, the offset distance of -1.0~10.0 mm can be adjusted. A cyan, opaque 1 mm thick PMMA plate was used to simulate the container wall, and calcium carbonate (CaCO3) powder was used as the internal sample to be tested. The samples were measured by the conventional method (zero offset) and the spatially offset method. The acquired raw spectra were first averaged and fitted by a 7th order polynomial to remove the baseline. Then the average of the three largest spectral peaks was used as the spectral intensity, and the variation law of the SORS signal with the offset distance was analyzed. It was found that: as the spatial offset distance increases, the Raman scattering intensity of the container wall decreases rapidly, while the Raman scattering intensity of the internal sample first rises and then decreases slowly; for samples of uniform thickness and isotropic, the trend of change is symmetrical about the zero offset, and the oblique incidence of the laser beam causes a slight asymmetry; at a certain offset distance, the ratio of the spectral intensity of the sample to the container wall reaches a maximum value, and there is an optimal detection offset distance (for this sample, the optimal offset distance is 1.2 mm). In the case where the material of the container and the sample is unknown, the clean Raman spectrum of each layer can still be obtained by the method of proportional subtraction. By calculating the spectrum at the zero offset and the optimal offset, the clean Raman spectra of the container wall and the internal sample are obtained respectively, which can be used in later spectral analysis and identification processes. This work demonstrates the potential of SORS for the detection of samples in opaque, colored containers, and provides a basis for further research on SORS and data processing methods.