Soil body configuration has a significant impact on soil moisture, solute transport process, and crop growth. Generally, measurements are performed using manual digging of the soil profile and sample collection, and then taking the samples to the laboratory for testing and analysis, etc., which has a low cycle length and low efficiency. To solve the aforementioned problems, by taking the ground penetrating radar (GPR) waveform and its image as the research object, and starting from the detection of the level, layer thickness, and soil quality of the soil structure, a nondestructive detection method is proposed to rapidly measure the soil body configuration using GPR. In this study, the longitudinal gradient information from the GPR waveform image can reflect soil stratification by using an envelope detection method to extract the envelope signal from the GPR echo and a Hilbert transform to analyze its instantaneous phase and determine the layer position. Given the relationship between the soil dielectric constant and the radar echo amplitude, this paper used the echo amplitude of the GPR to invert the dielectric constant of each layer, and then to calculate the propagation speed of the radar wave in the soil from the dielectric constant, so as to obtain the thickness of each layer of the soil profile. Based on the relationship between the image noise of the GPR waveform map and the soil sand to soil ratio, the principal component analysis method is used to evaluate the image noise for each layer of soil and obtain the sand content of each layer of soil, and combined with the support vector machine to identify the soil quality of each layer. The paper establish a knowledge base of soil structure covering regional information, soil indexes, detection information, and image multi-feature fusion information, and compile an information system for the rapid identification of soil and soil configuration, and use this method for field detection and verification in two test bases, six sampling points, and four types of soil and soil configurations around Hohhot City, Inner Mongolia Autonomous Region, China. Studies have shown that: (1) the correct rate of soil configuration recognition within 1 m below the surface of the above areas has reached more than 94%; (2) the relative measurement error of each layer of soil thickness detected is less than 10%. This research method provides a new mesoscopic detection concept for the rapid detection of soil configuration.