Nondestructive Detection Method Research Articles

Hyperspectral Imaging and Machine Learning for Huanglongbing Detection on Leaf-Symptoms

Huanglongbing is one of the most destructive diseases of citrus worldwide. Infected trees die due to the absence of practical cures. Thus, the removal of HLB-infected trees is one of the principal HLB managements for the regulation of disease spread. Here, we propose a non-destructive HLB detection method based on hyperspectral leaf reflectance. In total, 72 hyperspectral leaf images were collected in an HLB-invaded citrus orchard in Thailand and each image was visually distinguished into either any HLB symptom appearance (symptomatic) or no symptoms (asymptomatic) on the leaf. Principal component analysis was applied on the hyperspectral data and revealed 16 key wavelengths at red-edge to near-infrared regions (715, 718, 721, 724, 727, 730, 733, 736, 930, 933, 936, 939, 942, 945, 957, and 997 nm) that were characteristically differentiated in the symptomatic group. Seven models learnt on the spectral data at these 16 wavelengths were examined for the potential to separate these two image groups: random forest, decision tree, support vector machine, k-nearest neighbor, gradient boosting, logistic regression, linear discriminant. F1-score was employed to select the best-fit model to distinguish the two categories: random forest achieved the best score of 99.8%, followed by decision tree and k-nearest neighbor. The reliability of the visual grouping was evaluated by nearest neighbor matching and permutation test. These three models separated the two image categories as precisely as PCR results, indicating their potential as alternative tool instead of PCR.

Deep neural network for electromechanical impedance analysis and damage detection of laminated rubber bearings

Abstract Environmental factors and loads cause potential small cracks in the rubber layers of laminated bearings. Conventional methods for identifying structural damage based on global vibration characteristics are ineffective, and current nondestructive structural damage detection methods based on Piezoelectric ceramic lead Zirconate Titanate (PZT) cannot fully capture deep damage-sensitive features from PZT signals. This study proposes a novel deep learning approach using a convolutional neural network with residual connections to conduct raw electromechanical admittance analysis and automatically extract damage-sensitive features to identify minor cracks in laminated rubber bearings. Numerical models are used to generate a sufficient amount of PZT data under various damage conditions for training the deep neural network, which overcomes the sparsity of damage states in actual laminated rubber bearings. A transfer learning method based on a model fine-tuning strategy was adopted to address the feature differences between the numerical models and actual rubber bearing structure. The fine-tuned model could effectively identify the damage location of actual laminated rubber bearings. The proposed method does not require a complex data preprocessing process when analyzing PZT signals and has a relatively high accuracy in damage localization. The promising results from this work offer a potential paradigm for data-driven PZT signal analysis, which is an appealing approach for identifying minor damage in real-life laminated rubber bearings.

Utilizing linear and nonlinear ultrasound for improved estimation of concrete strength

ABSTRACT This study investigated non-destructive concrete strength evaluation using in-situ cored samples. Data from various mix proportions and design strengths revealed inconsistent correlations between compressive strength and linear ultrasonic wave velocity (V p). This inconsistency arises because V p captures only linear elastic behaviour, whereas concrete strength reflects both linear and nonlinear behaviours. In contrast, the acoustic nonlinearity parameter (β re) exhibits superior sensitivity to microstructural conditions owing to its direct link to material nonlinearity. We proposed a new integrity index (M/β*) by combining the constrained modulus (M) derived from V p with β re to capture the entire elastic behaviour. This index demonstrated statistically significant correlation with the factor of safety (FS) across different design strengths and mix proportions. Next, we developed a risk stratification system based on M/β* using fuzzy c-means clustering, categorising concrete into three risk zones: critical (M/β* <40), moderate (40 < M/β* <75) and low (M/β* >75), providing a quantitative tool for assessing concrete quality. The results highlight the importance of integrating linear and nonlinear ultrasonic parameters for accurate strength prediction, particularly in field conditions where traditional V p–strength correlations are unreliable. This approach offers a new method for non-destructive concrete strength detection, thereby enhancing structural safety assessments.

Unsupervised defect detection for solar photovoltaic cells based on convolutional autoencoder

ABSTRACT Solar photovoltaic (PV) cells are inevitably subject to defects during the production process, affecting their power generation efficiency and life. Electroluminescence (EL) imaging is the mainstream non-destructive method for PV cell defect detection. Aiming at PV cell EL images, an unsupervised defect detection method was proposed. Specifically, an unsupervised convolutional autoencoder (CAE), the scale structure perception convolutional autoencoder (SSP_CAE), was constructed, whose Squeeze-and-Excitation Attention (SE Attention) and skip connections avoid the blurring of image structure information and the loss of pixel-level details in the encoding and decoding process. Furthermore, to balance the global and local information of the image, a scale perception loss function called SP_SSIM was proposed for model training. The defect segmentation was achieved by using Otsu thresholding method to binarize the obtained Mean Absolute Error (MSE) residual heat image in the testing stage of the model. Finally, the experiments were performed on the test dataset and the experimental results showed that the proposed SSP_CAE can effectively detect PV cell defects. The experimentally obtained defect detection performance metrics Precision, Recall, IoU, F1-score and AUROC values were 0.739, 0.886, 0.723, 0.764 and 0.841, respectively. Compared with other classical methods, the proposed SSP_CAE had a better comprehensive performance for defect detection.

Non-Destructive Detection of Silage pH Based on Colorimetric Sensor Array Using Extended Color Components and Novel Sensitive Dye Screening Method

Non-destructive detection of maize silage quality is essential. The aim is to propose a fast and non-destructive silage pH detection method based on a colorimetric sensor array (CSA). Extended color components, a novel sensitive dye screening method, and a feature screening method were integrated and applied to enhance pH detection. Fifty color components were constructed from five color spaces and used to extract information about the response of CSA to silage. Forward and backward stepwise selection and support vector regression (SVR) were combined to create a sensitive dye screening method, which was used to determine the optimal sensitive dye. The variable combination population analysis–iteratively retains informative variables algorithm was iterated to optimize effective features. Consequently, six hundred variables were extracted from the twelve dyes, which were able to comprehensively and finely characterize the CSA response. Four sensitive dyes were screened out from the twelve dyes, which were sensitive to silage volatile compounds and accurately reflected the odor changes. Twenty-eight effective features were preferred, based on which the SVR model had Rp2, RMSEP and RPD scores of 0.9533, 0.4186, and 4.4186, respectively; the pH prediction performance was substantially improved. This study provides technical support for the scientific evaluation of silage quality.

Noninvasive, Presymptomatic Detection of Potato Cyst Nematode Infection in Tomato Using Chlorophyll Fluorescence Analysis.

Potato cyst nematodes (PCNs) are notorious pathogens in all major potato production areas worldwide. Mainly due to the low mobility of this soil pathogen, PCN infestations are mostly observed as patches ("foci") that only cover a fraction of the acreage. In-field presymptomatic localization of these pathogens is valuable, as it would allow for the localized application of control measures. Although the mapping of foci is technically feasible, it is unpractical, as it would require the analysis of numerous soil samples. We investigated whether chlorophyll fluorescence (Chl-F) could be suitable as a rapid, nondestructive method for early PCN detection. To this end, the impact of four Globodera pallida densities on the Chl-F of tomato was investigated in a phenotyping greenhouse for 26 days. Furthermore, the classical plant performance indicators of biomass and root surface area were compared with Chl-F. Thermal dissipation (NPQ) and an estimate of the photosynthetic rate (ΦPSII) responded at 1 day postinoculation, and ΦPSII was most sensitive to low PCN infection levels. Chl-F parameters responded more readily to PCN infection than biomass and root surface area. The maximum quantum yield of photosystem II (Fv/Fm) and the potential activity of photosystem II (Fv/F0) initially increased at low PCN infection levels, whereas a sharp decrease was observed at higher infestation levels. Hence, our data suggest that low PCN levels promoted plant performance before becoming detrimental at higher levels. Although Chl-F allowed for early and sensitive PCN detection, it remains to be investigated whether these signals can be distinguished from those produced by other belowground stressors in the field.

Integrated Extraction of Root Diameter and Location in Ground-Penetrating Radar Images via CycleGAN-Guided Multi-Task Neural Network

The diameter of roots is pivotal for studying subsurface root structure geometry. Yet, directly obtaining these parameters is challenging due their hidden nature. Ground-penetrating radar (GPR) offers a reproducible, nondestructive method for root detection, but estimating diameter from B-Scan images remains challenging. To address this, we developed the CycleGAN-guided multi-task neural network (CMT-Net). It comprises two subnetworks, YOLOv4-Hyperbolic Position and Diameter (YOLOv4-HPD) and CycleGAN. The YOLOv4-HPD is obtained by adding a regression header for predicting root diameter to YOLOv4-Hyperbola, which achieves the ability to simultaneously accurately locate root objects and estimate root diameter. The CycleGAN is used to solve the problem of the lack of a real root diameter training dataset for the YOLOv4-HPD model by migrating field-measured data domains to simulated data without altering root diameter information. We used simulated and field data to evaluate the model, showing its effectiveness in estimating root diameter. This study marks the first construction of a deep learning model for fully automatic root location and diameter extraction from GPR images, achieving an “Image Input–Parameter Output” end-to-end pattern. The model’s validation across various dataset scales opens the way for estimating other root attributes.

An intelligent model approach for leakage detection of modified atmosphere pillow bags

Modified atmosphere pillow bags have been widely used to package various food products due to their advantages for preservation and shipment. Sealing defects are statistically inevitable, although modern packaging machinery and manual inspection utilized by manufacturers continue reducing the leakage probability. Hence the bag contents may spoil if the seal is broken. Instead of manual inspection and various destructive methods utilized by factories, this study introduces non-destructive leakage detection using deep learning methods. Firstly, a squeezing method is developed to aggravate the feature difference between positive samples and negative samples without destroying the bag content, thus 2160 images of three different pillow bags are acquired to establish dataset. Secondly, the deep learning model Vision Transformer (ViT) is deployed and studied so that feasibility of computer vision method is verified. Then the Semantic segmentation and Contour Extraction model combining ViT (SCE-ViT) is proposed and improved to the Multi-dimensional Fusion model (SCE-MdF). The accuracies of SCE-MdF reached 97.5%, 97.5%, and 97.5%, respectively. The F1-scores of SCE-MdF reached 97.6%, 97.6%, and 97.4%, respectively. Compared to averaged accuracies of SCE-ViT, accuracies introduced in the ultimate model SCE-MdF improved by 19.17%, 5.84%, and 11.67%, respectively. Therefore, combination of unique squeezing method and Semantic segmentation Contour Extraction with Multi-dimensional Fused ViT, is eventually validated viable on leakage detection of modified atmosphere pillow bags. Hence a cost-effective, efficient and non-destructive leakage detection method for modified atmosphere pillow bags in relevant industry is introduced, filling a gap between artificial intelligence and food packaging industry.

An efficient nondestructive detection method of rapeseed varieties based on hyperspectral imaging technology

An efficient nondestructive detection method of rapeseed varieties based on hyperspectral imaging technology

Performance Analysis Using CNN for Detecting Wood Knots

This study proposes a Convolutional Neural Network (CNN) model to quickly and accurately detect wood deformations. The performance of the CNN was enhanced by extracting structural deformation features, optimizing training parameters, and improving datasets. Experimental analyses demonstrate that the CNN achieved high accuracy rates and is an effective method for deformation detection. The CNN model was designed to identify various wood deformations. Its layered architecture was optimized to analyze deformations at different scales and levels of detail. Minimal preprocessing was applied to the images used during training, and data augmentation techniques were employed to enhance dataset diversity. The model was trained on a training dataset and tested on a validation dataset. Metrics such as loss function and accuracy were monitored throughout the training process. The CNN achieved an accuracy rate of 99.90% on the training dataset. This study highlights that the CNN model is an effective method for non-destructive detection of wood deformations. The proposed CNN model has potential applications in wood deformation detection and quality control processes.

Research on X-ray nondestructive defect detection method of tire based on dynamic Snake Convolution YOLO model

Tire X-ray nondestructive testing before leaving the factory is crucial for driving safety. Given the complexity of tire structures and the diversity of defect types, traditional manual visual inspections and machine learning methods face significant challenges in terms of accuracy and efficiency. This study proposes an innovative tire X-ray image nondestructive testing technique based on the YOLOv5 model, incorporating several advanced technologies to enhance detection performance. Specifically, we introduce Dynamic Snake Convolution (DSConv), which adaptively focuses on slender and curved features within tires. Additionally, we have designed a C3 module based on DSConv, specifically targeting slender defects such as cord-overlap and cord-cracking. To improve the detection accuracy of small defects, we redesigned the neck network structure and introduced the Scale sequence feature fusion module (SSFF) and the Triple feature encoding module (TFE) to integrate multi-scale information from different network layers. Furthermore, we developed the Convolution Block Attention Module, integrated into the SSFF, which effectively reduces the interference of complex backgrounds and focuses on defect recognition. In the post-processing stage, we employed the Soft-NMS algorithm to optimize the confidence of candidate detection boxes, enhancing the accuracy of box selection. The experimental results show that compared to the YOLOv5 benchmark model, the algorithm proposed in this study achieved a 5.9 percentage point increase in mAP0.5 and a 5.7 percentage point increase in mAP0.5:0.95, demonstrating superior detection accuracy compared to current mainstream object detection algorithms and effectively completing the nondestructive testing task of tire defects.

Enhancing Regional Topsoil Total Nitrogen Mapping Through Differentiated Fusion of Ground Hyperspectral Data and Satellite Images Under Low Vegetation Cover

Total nitrogen in soil (STN) serves as a crucial indicator of soil nutrient content and provides an essential nitrogen source necessary for crop growth. Precisely inversion of STN content is crucial for the sustainable management of soil resources and the advancement of agricultural development, particularly to achieve efficient fertilization—reduction in fertilizer usage without compromising yield or increase in yield while maintaining the total fertilization amount. Spectroscopy technology is regarded as an ideal non-destructive method for nutrient detection. However, due to the weak spectral signals of STN and its spatial heterogeneity, hyperspectral imaging technology presents significant potential for high-resolution measurements and precise characterization of STN heterogeneity. In this paper, the STN content was selected as the study subject, and three aspects of soil spectral feature enhancement, multi-source remote sensing data differentiated fusion, and STN content inversion model construction were studied. Therefore, a differentiated fusion of enhanced multispectral image bands (DFE_MSIBs) method combined with Random Forest (RF) algorithms was developed for spectral inversion of STN content. The findings demonstrate the following: 1. The enhanced spectral characteristics and differentiated fusion method not only strengthen the relationship between STN and Sentinel-2A MSI data but also enhance the precision of regional STN inversion models. 2. For the differentiated fusion of enhanced multispectral image bands (DFE_MSIBs) method combined with Random Forest (RF) algorithms, the R2 was 0.95, RMSE was 0.10 g/kg, and LCCC was 0.89. 3. Compared to the unfused model, the average R2 value was increased by 0.02, the average RMSE was decreased by 0.01 g/kg, and the average LCCC was increased by 0.03. These findings hold practical significance for utilizing multi-source remote sensing data in STN mapping and precision fertilization in agricultural fields.

Non-Destructive Pest Detection: Innovations and Challenges in Sensing Airborne Semiochemicals.

Pests, especially invasive ones, pose significant threats to the global ecosystem, crop security, and agriculture economy. Sensing airborne semiochemicals as a nondestructive detection method has been recognized as a promising strategy to detect the presence of these living pests on site. However, sensing airborne semiochemicals in fields is challenging, as they are transmitted in concentrations as low as several nanograms per cubic meter in chemically diverse environments. This low vapor pressure together with similarity in functional groups of pheromones among different species have curtailed the practical deployment of corresponding sensors for real world applications. This review describes the advances in semiochemical detection methods and technologies including traditional analytical instruments, trained animals, and electroantennography with a focus on electronic noses (e-noses). Several key types of volatile organic compound (VOC) sensors used in e-noses are summarized, including their transduction methods, sensing materials, and sensing performance for semiochemical and simulants detection. Notably, it was found that many commercial VOC sensors failed to respond to airborne semiochemicals effectively, leading to a reduced efficiency of e-noses. Future work may focus on developing stable and robust sensing materials with higher sensitivity and selectivity to pheromones and understanding the feasibility of the deployment of the sensors under field conditions.

Investigating the Surface Damage to Fuzhou’s Ancient Houses (Gu-Cuo) Using a Non-Destructive Testing Method Constructed via Machine Learning

As an important part of traditional Chinese architecture, Fuzhou’s ancient houses have unique cultural and historical value. However, over time, environmental factors such as efflorescence and plant growth have caused surface damage to their gray brick walls, leading to a decline in the quality of the buildings’ structure and even posing a threat to the buildings’ safety. Traditional damage detection methods mainly rely on manual labor, which is inefficient and consumes a lot of human resources. In addition, traditional non-destructive detection methods, such as infrared imaging and laser scanning, often face difficulty in accurately identifying specific types of damage, such as efflorescence and plant growth, on the surface of gray bricks and are easily hampered by diverse surface features. This study uses the YOLOv8 machine learning model for the automated detection of two common types of damage to the gray brick walls of Fuzhou’s ancient houses: efflorescence and plant growth. We establish an efficient gray brick surface damage detection model through dataset collection and annotation, experimental parameter optimization, model evaluation, and analysis. The research results reveal the following. (1) Reasonable hyperparameter settings and model-assisted annotation significantly improve the detection accuracy and stability. (2) The model’s average precision (AP) is improved from 0.30 to 0.90, demonstrating good robustness in detecting complex backgrounds and high-resolution real-life images. The F1 value of the model’s gray brick detection efficiency is improved (classification model performance index) from 0.22 to 0.77. (3) The model’s ability to recognize the damage details of gray bricks under high-resolution conditions is significantly enhanced, demonstrating its ability to cope with complex environments. (4) The simplified data enhancement strategy effectively reduces the feature extraction interference and enhances the model’s adaptability in different environments.

Experimental study on internal damage mechanisms and performance state perception of tunnel linings under different loading paths

The state of tunnel lining structures plays a crucial role in determining safety and service quality throughout their serving life. Therefore, establishing a quantitative method for evaluating the bearing capacity of linings is crucial. Understanding the process and status of crack propagation is key to ensuring structural safety. The three-dimensional development of lining damage cracks is closely related to structural safety. Unveiling the spatial growth of these cracks and the evolution of structural performance is vital for an accurate assessment of lining safety. Key factors in crack propagation, along with non-destructive detection methods, serve as primary approaches for evaluating the safety condition. This paper explores the evolution law and evaluation model of the bearing capacity of plain concrete linings under different loading paths. The distribution of structural stiffness changes significantly under different boundary settings. Therefore, full-scale loading tests of lining components were conducted, designing two types of linings: one intact and the other with initial defects, to simulate the failure of plain concrete linings under diverse loading conditions. The results show that: (1) The damage development of linings exhibits clear staged evolution characteristics; (2) The spatial development of lining cracks can be divided into three stages; (3) There are significant differences in ultrasonic parameters at different stages. Consequently, a method for evaluating the local bearing capacity state of linings based on the coupling of crack characteristics and ultrasonic parameters is proposed; (4) This method is also applicable to linings with initial damage, where damage evolution follows similar stages. The findings provide theoretical guidance for predicting and preventing crack propagation in operational tunnel linings.

Detection of tea seed oil adulteration based on near-infrared and Raman spectra information fusion

The adulteration of tea seed oil severely harms consumer interests, such as affecting the taste experience and economic losses, creating an urgent need for a rapid, efficient, and non-destructive detection method. In this study, near-infrared and Raman spectroscopy were used to detect the adulteration of tea seed oil with four common edible vegetable oils. After preprocessing the raw spectral data, three variable selection methods namely uninformative variable elimination, competitive adaptive reweighted sampling, and successive projections algorithm were used individually and consecutively to select key spectral variables. And the NIR and Raman characteristic variables were then fused to develop a discrimination model. The results indicate that normalization is the superior preprocessing method and consecutive variable selection methods are generally superior to single ones. The best NIR and Raman models, based on consecutive variable selection methods, achieve accuracy and F1 score of the prediction set of 95.652%, 97.479% and 92.754%, 95.868%, respectively. Spectral information fusion can fully utilize the spectral information, the accuracy and F1 score of the optimal binary adulteration discrimination model based on information fusion for calibration and prediction sets are all 100%. Meanwhile, the accuracy and F1 score of the multiple adulteration discrimination model are 98.225% and 98.932%, respectively. Therefore, the fusion of NIR and Raman spectral information can enable highly accurate discrimination of adulterated oil. The proposed method in this study will aid in developing portable detection devices for tea seed oil adulteration.

A Nondestructive Detection Method for the Muti-Quality Attributes of Oats Using Near-Infrared Spectroscopy.

This study aimed to achieve a precise and non-destructive quantification of the amounts of total starch, protein, β-glucan, and fat in oats using near-infrared technology in conjunction with chemometrics methods. Eight preprocessing methods (SNV, MSC, Nor, DE, FD, SD, BC, SS) were employed to process the original spectra. Subsequently, the optimal PLS model was obtained by integrating feature wavelength selection algorithms (CARS, SPA, UVE, LAR). After SD-SPA, total starch reached its optimal state (Rp2 = 0.768, RMSEP = 2.057). Protein achieved the best result after MSC-CARS (Rp2 = 0.853, RMSEP = 1.142). β-glucan reached the optimal value after BC-SPA (Rp2 = 0.759, RMSEP = 0.315). Fat achieved the optimal state after SS-SPA (Rp2 = 0.903, RMSEP = 0.692). The research has shown the performance of the portable FT-NIR for a rapid and non-destructive quantification of nutritional components in oats, holding significant importance for quality control and quality assessment within the oat industry.

Detection of mussels contaminated with cadmium by near-infrared reflectance spectroscopy based on RELS-TSVM.

Eating mussels contaminated with cadmium (Cd) can seriously harm health. In this study, a non-destructive and rapid detection method for Cd-contaminated mussels based on near-infrared reflectance spectroscopy was studied. The spectral data of Cd-contaminated and non-contaminated mussels were collected in the range of 950-1700nm. The model based on a robust energy-based least squares twin support vector machine (RELS-TSVM) was established to detect Cd-contaminated mussels. The influence of parameters on the RELS-TSVM model was analyzed, and the most suitable parameters were determined. The average accuracy of the proposed RELS-TSVM model in detecting Cd-contaminated mussels reached 99.92%, which was better than other twin support vector machine-derived models. For test datasets with different kinds of spectral noises (Gaussian noise, baseline shift, stray light, and wavelength shift), the RELS-TSVM model had a high robustness for noise disturbance. The results show that near-infrared spectroscopy combined with the RELS-TSVM model can realize the detection of Cd-contaminated mussels, which can provide technical support for the monitoring of heavy metals in shellfish. PRACTICAL APPLICATION: The method of detecting Cd-contaminated mussels by the NIRS has important practical significance for ensuring the safety of consumers. It provides a new way for the quality assessment and safety detection of shellfish and provides a technical basis for the marine environment assessment and management.

The Study on Nondestructive Detection Methods for Internal Quality of Korla Fragrant Pears Based on Near-Infrared Spectroscopy and Machine Learning.

Quality control and grading of Korla fragrant pears significantly impact their commercial value. Rapid and non-destructive detection of soluble solids content (SSC) and firmness is crucial to improving this. This study proposes a method combining near-infrared spectroscopy (NIRS) with machine learning for the rapid, non-destructive detection of SSC and firmness in Korla pears. By analyzing absorbance in the 900-1800 nm range, six preprocessing methods-Savitzky-Golay derivative (SGD), standard normal variate (SNV), multiplicative scatter correction (MSC), Savitzky-Golay smoothing (SGS), vector normalization (VN), and min-max normalization (MMN)-were applied to the raw spectral data. uninformative variable elimination (UVE) and successive projections algorithm (SPA) were then used to extract effective wavelengths. Partial least squares regression (PLSR) models were developed for SSC and firmness based on the extracted data. The results showed that all preprocessing and wavelength-extraction methods improved model accuracy. The optimal SSC prediction model was MSC-SPA-PLSR (R = 0.93, RMSE = 0.195), and the best hardness prediction model was MSC-UVE-PLSR (R = 0.83, RMSE = 0.249). This research aids in establishing a non-destructive testing system, offering producers a rapid and accurate quality assessment tool, and provides the food industry with better production control measures to enhance standardization and market competitiveness of Korla pears.

Application of ground-penetrating radar based on U-Net in the statistics on the bonded area of external wall insulation

Abstract The insulation of external walls of buildings often suffers from various diseases. Therefore, it is necessary to accurately detect the bonding area of the insulation layer. Ground-penetrating radar is a non-destructive detection method that can effectively determine the bonding status. In this paper, FDTD forward simulation of the exterior wall insulation layer by ground-penetrating radar is carried out, and the GPR images of the 4GHZ and 6GHZ antennas are simulated, respectively, and field experiments are conducted to collect the stepped frequency vector net data of 2.5-8GHZ and 6-8GHZ. The data obtained are processed by using the improved SP-ICA clutter suppression algorithm, followed by the improved loss function of the U-Net and training. The F1_score of the obtained model can reach 0.8399, and the error of bonding area statistics is 0.23%, which realizes more accurate semantic segmentation and area statistics.