Research on The Acoustic-Vibration Detection Method for The Apple Moldy Core Disease Based on D-S Evidence Theory

doi:10.12133/j.smartag.SA202505032

Abstract

Abstract:

[Objective] The moldy core disease is a common internal disease of the apple and is highly infectious. In early storage, the mold symptoms are confined to the interior of the core. The apple tissue is in a sub-healthy state and still has commercial value, so early detection of moldy-core apples is critical. [Methods] In this study, a non-destructive acoustic vibration detection system was used to acquire acoustic vibration response signals. Symmetrized dot pattern (SDP), gramian angular field (GAF), and stockwell transform (ST) were applied to obtain multi-domain acoustic vibration spectra, including SDP images, gramian angular summation field (GASF) images, gramian angular difference field (GADF) images, and ST images. These images were uniformly converted to grayscale, transforming the time-domain signals into multi-domain visual spectra to facilitate subsequent feature analysis and recognition. Uniform local binary pattern (ULBP) and gray-level-gradient co-occurrence matrix (GLGCM) methods were used to extract handcrafted features from the multi-domain visualized images. Subsequently, the maximum relevance and minimum redundancy (mRMR) criterion was applied to select the dominant features from each analysis domain that are sensitive to early disease information. Principal component analysis (PCA) was employed to reduce the dimensionality of the multi-domain spectral ULBP texture features. From the statistical features extracted from one-dimensional acoustic-vibration signals in the time and frequency domains, and the GLGCM texture features extracted from two-dimensional images in each domain, 5 to 8 features sensitive to early moldy core detection were selected. From the high-dimensional sensitive ULBP texture features extracted from each domain, 3 to 7 principal components were obtained through dimensionality reduction using principal component analysis. This selection aimed to maximize the relevance between features and class labels while minimizing redundancy among features, thereby identifying the most informative features for early mold core apple detection in each domain. Meanwhile, a ResNet50 feature extractor improved with a convolutional attention mechanism module and the Adam optimizer (Adam-IResNet50) was designed to automatically extract deep features from visualized images in each domain. The optimal shallow features were used to train a multiple support vector machine (MSVM) classifier, while the optimal deep features were used to train an extreme learning machine (ELM) classifier. The IResNet50 network, improved with a convolutional attention mechanism and optimized using the Adam algorithm, was employed as a feature extractor. The deep features extracted from the time-domain and frequency-domain GADF images, as well as time–frequency images, resulted in higher sample matching scores (SC) and cluster compactness (CHS) values, along with lower inter-class overlap (DBI) values for the three apple categories. These results clearly indicate that the deep features extracted by the Adam-IResNet50 model from multi-domain images exhibit strong capability in identifying subhealth and moldy core apples. The preliminary outputs of the two classifiers were converted into basic probability assignments for independent evidence bodies. Dempster's combination rule and the associated decision criterion of Dempster–Shafer (D-S) theory were then applied to yield the final decision on early-stage moldy apples. Consequently, a decision-level fusion model was established for both shallow and deep features of the acoustic-vibration multi-domain spectra. [Results and Discussions] The constructed Adam-IResNet50-IPSO-ELM-DS model based on D-S evidence theory achieved a Kappa coefficient and Matthews Correlation Coefficient (MCC) slightly below 90% for multi-class classification of apples from known origins. The F₁-Score and Overall Accuracy (OA) reached 93.01% and 93.22%, respectively. The classification accuracy for sub-healthy apples was 87.37%, while the misclassification rate for diseased apples was 8.33%. These results indicate that the model maintains a balanced precision and recall while achieving high detection accuracy for three classes of apples from unknown origins. After decision fusion, the IPSO-MSVM-DS and Adam-IResNet50-IPSO-ELM-DS models demonstrated significant performance improvements. Among them, the Adam-IResNet50-IPSO-ELM-DS model achieved an accuracy of 93.22%, which is significantly higher than other methods. This demonstrates that decision-level fusion can effectively enhance the model's discriminative ability and further improve classification performance. [Conclusions] The proposed acoustic vibration detection method for mold core apples, based on Dempster-Shafer evidence theory, effectively accomplishes the detection task and provides technical support for future online batch detection of early mold core apples. Early screening of sub-healthy apples is of great significance for quality control during postharvest storage. In future work, the model will be further optimized to develop a rapid acoustic vibration-based prediction method for early detection of mold core, providing technical support for quality control during apple distribution.

Key words: acoustic vibration technology, sub-healthy apple, decision fusion, non-destructive detection

CLC Number:

TP274.5

LIU Jie, ZHAO Kang, ZHAO Qinjun, SONG Ye. Research on The Acoustic-Vibration Detection Method for The Apple Moldy Core Disease Based on D-S Evidence Theory[J]. Smart Agriculture, doi: 10.12133/j.smartag.SA202505032.

Figures/Tables 18

Fig. 1

Fig. 2

Fig.3

Table 1

Textural feature parameters of vibro-acoustic multi-domain images and the corresponding formulas

序号	特征参数	计算公式	序号	特征参数	计算公式
1	小梯度优势	$G 1 = ∑ i = 0 l f - 1 ∑ j = 0 l g - 1 H (i, j) / (j + 1) 2 ∑ i = 0 l f - 1 ∑ i = 0 l g - 1 H (i, j)$ （1）	9	梯度均方差	$G 9 = ∑ j = 0 l g - 1 (j - G 7) 2 ∑ i = 0 l f - 1 P (i, j) 12$ （9）
2	大梯度优势	$G 2 = ∑ i = 0 l f - 1 ∑ j = 0 l g - 1 H (i, j) j 2 ∑ i = 0 l f - 1 ∑ j = 0 l g - 1 H (i, j)$ （2）	10	相关性	$G 10 = ∑ i = 0 l f - 1 ∑ j = 0 l g - 1 (i - G 6) (j - G 7) P (i, j)$ （10）
3	灰度分布不均匀性	$G 3 = ∑ i = 0 l f - 1 ∑ j = 0 l g - 1 H (i, j) 2 ∑ i = 0 l f - 1 ∑ j = 0 l g - 1 H (i, j)$ （3）	11	灰度熵	$G 11 = - ∑ i = 0 l f - 1 ∑ j = 0 l g - 1 P (i, j) l o g ∑ j = 0 l g - 1 P (i, j)$ （11）
4	梯度分布不均匀性	$G 4 = ∑ j = 0 l g - 1 ∑ i = 0 l f - 1 H (i, j) 2 ∑ i = 0 l f - 1 ∑ j = 0 l g - 1 H (i, j)$ （4）	12	梯度熵	$G 12 = - ∑ j = 0 l g - 1 ∑ i = 0 l f - 1 P (i, j) l o g ∑ i = 0 l - f 1 P (i, j)$ （12）
5	角二阶距	$G 5 = ∑ i = 0 l f - 1 ∑ j = 0 l g - 1 [P (i, j)] 2$ （5）	13	混合熵	$G 13 = - ∑ i = 0 l f - 1 ∑ j = 0 l g - 1 P (i, j) l o g P (i, j)$ （13）
6	灰度均值	$G 6 = ∑ i = 0 l f - 1 i ∑ j = 0 l g - 1 P (i, j)$ （6）	14	惯性矩	$G 14 = ∑ i = 0 l f - 1 ∑ j = 0 l g - 1 (i - j) 2 H ∧ (i, j)$ （14）
7	梯度均值	$G 7 = ∑ j = 0 l g - 1 j ∑ i = 0 l f - 1 P (i, j)$ （7）	15	逆差矩	$G 15 = ∑ i = 0 l f - 1 ∑ j = 0 l g - 1 P (i, j) 1 + (i - j) 2$ （15）
8	灰度均方差	$G 8 = ∑ i = 0 l f - 1 (i - G 6) 2 ∑ j = 0 l g - 1 P (i, j) 12$ （8）

Table 1

Fig. 4

Fig. 5

Fig. 6

Table 2

Fig. 7

Table3

Fig. 8

Table 4

Table 5

Table 6

Fig. 9

Fig. 10

Table 7

Table 8

References 51

[1]	LIANG X F, ZHANG R, GLEASON M L, et al. Sustainable apple disease management in China: Challenges and future directions for a transforming industry[J]. Plant disease, 2022, 106(3): 786-799.
[2]	TEMPELAERE A, VAN DOORSELAER L, HE J Q, et al. BraeNet: Internal disorder detection in 'Braeburn' apple using X-ray imaging data[J]. Food control, 2024, 155: ID 110092.
[3]	SUCHANEK M, KORDULSKA M, OLEJNICZAK Z, et al. Application of low-field MRI for quality assessment of 'Conference' pears stored under controlled atmosphere conditions[J]. Postharvest biology and technology, 2017, 124: 100-106.
[4]	LI X L, WEI Y Z, XU J, et al. SSC and pH for sweet assessment and maturity classification of harvested cherry fruit based on NIR hyperspectral imaging technology[J]. Postharvest biology and technology, 2018, 143: 112-118.
[5]	GENANGELI A, ALLASIA G, BINDI M, et al. A novel hyperspectral method to detect moldy core in apple fruits[J]. Sensors, 2022, 22(12): ID 4479.
[6]	ZHAO K, ZHA Z H, LI H, et al. Early detection of moldy apple core based on time-frequency images of vibro-acoustic signals[J]. Postharvest biology and technology, 2021, 179: ID 111589.
[7]	ZHANG H, ZHA Z H, KULASIRI D, et al. Detection of early core browning in pears based on statistical features in vibro-acoustic signals[J]. Food and bioprocess technology, 2021, 14(5): 887-897.
[8]	ZHAO K, ZHAO J, YANG Y, et al. Detection of moldy pear core based on the time-frequency analysis of acoustic vibration signals and multi-domain features fusion[J]. Postharvest biology and technology, 2025, 225: ID 113495.
[9]	YIN K, SHEN Y W, CHEN Y F, et al. Multi-source information fusion diagnosis method for aero engine[J]. Applied sciences, 2025, 15(9): ID 5083.
[10]	ZHOU K, TANG J. Harnessing fuzzy neural network for gear fault diagnosis with limited data labels[J]. The international journal of advanced manufacturing technology, 2021, 115(4): 1005-1019.
[11]	CERRADA M, LI C, SÁNCHEZ R V, et al. A fuzzy transition based approach for fault severity prediction in helical gearboxes[J]. Fuzzy sets and systems, 2018, 337: 52-73.
[12]	JAFARI H, POSHTAN J, SADEGHI H. Application of fuzzy data fusion theory in fault diagnosis of rotating machinery[J]. Proceedings of the institution of mechanical engineers, Part I: journal of systems and control engineering, 2018, 232(8): 1015-1024.
[13]	JI W T, PANG Y, JIA X Y, et al. Fuzzy rough sets and fuzzy rough neural networks for feature selection: A review[J]. WIREs data mining and knowledge discovery, 2021, 11(3): ID e1402.
[14]	WANG C Z, WANG C Y, QIAN Y H, et al. Feature selection based on weighted fuzzy rough sets[J]. IEEE transactions on fuzzy systems, 2024, 32(7): 4027-4037.
[15]	宋玉琴, 朱紫娟, 姬引飞. 基于粗糙集优化的信息融合故障诊断系统[J]. 电子技术应用, 2015, 41(8): 143-145, 148.
	SONG Y Q, ZHU Z J, JI Y F. Information fusion fault diagnosis system based on rough set[J]. Application of electronic technique, 2015, 41(8): 143-145, 148.
[16]	PACHECO F, CERRADA M, SÁNCHEZ R V, et al. Attribute clustering using rough set theory for feature selection in fault severity classification of rotating machinery[J]. Expert systems with applications, 2017, 71: 69-86.
[17]	CHEN G Z, WU Y F, FU L, et al. Fault diagnosis of full-hydraulic drilling rig based on RS–SVM data fusion method[J]. Journal of the Brazilian society of mechanical sciences and engineering, 2018, 40(3): ID 140.
[18]	ZHANG P F, LI T R, WANG G Q, et al. Multi-source information fusion based on rough set theory: A review[J]. Information fusion, 2021, 68: 85-117.
[19]	WANG Y X, LIU F, ZHU A H. Bearing fault diagnosis based on a hybrid classifier ensemble approach and the improved dempster-shafer theory[J]. Sensors, 2019, 19(9): ID 2097.
[20]	SHI J C, YI J Y, REN Y, et al. Fault diagnosis in a hydraulic directional valve using a two-stage multi-sensor information fusion[J]. Measurement, 2021, 179: ID 109460.
[21]	ZHANG J Z, WU Y B, XU Z, et al. Fault diagnosis of high voltage circuit breaker based on multi-sensor information fusion with training weights[J]. Measurement, 2022, 192: ID 110894.
[22]	WANG J B, YANG B, ZENG C Y, et al. Recent advances and summarization of fault diagnosis techniques for proton exchange membrane fuel cell systems: A critical overview[J]. Journal of power sources, 2021, 500: ID 229932.
[23]	LIU Y W, CHENG Y Q, ZHANG Z Z, et al. Multi-information fusion fault diagnosis based on KNN and improved evidence theory[J]. Journal of vibration engineering & technologies, 2022, 10(3): 841-852.
[24]	蔡浩, 郭宏亮. 基于多分类器DS证据理论融合的水果识别研究[J]. 中国农机化学报, 2021, 42(2): 184-189.
	CAI H, GUO H L. Research on fruit recognition based on multi-classifier DS evidence theory fusion[J]. Journal of Chinese agricultural mechanization, 2021, 42(2): 184-189.
[25]	毕淑慧, 李雪, 申涛, 等. 基于多模型证据融合的苹果分类方法[J]. 农业工程学报, 2022, 38(13): 141-149.
	BI S H, LI X, SHEN T, et al. Apple classification based on evidence theory and multiple models[J]. Transactions of the Chinese society of agricultural engineering, 2022, 38(13): 141-149.
[26]	YAN X W, MA L Y, BI S H, et al. Application of DS evidence theory in apple's internal quality classification[M]//The 10th International Conference on Computer Engineering and Networks. Singapore: Springer Singapore, 2020: 563-571.
[27]	MOOSAVIAN A, KHAZAEE M, NAJAFI G, et al. Spark plug fault recognition based on sensor fusion and classifier combination using Dempster-Shafer evidence theory[J]. Applied acoustics, 2015, 93: 120-129.
[28]	WANG Z, WANG R X, GAO J M, et al. Fault recognition using an ensemble classifier based on Dempster–Shafer Theory[J]. Pattern recognition, 2020, 99: ID 107079.
[29]	WEN Y, TAN J W, ZHAN H, et al. Fault diagnosis based on multi-sensor data fusion for numerical control machine[J]. International journal of online and biomedical engineering (iJOE), 2016, 12(2): 29-34.
[30]	LI L, PENG Y K, LI Y Y, et al. Rapid and low-cost detection of moldy apple core based on an optical sensor system[J]. Postharvest biology and technology, 2020, 168: ID 111276.
[31]	王志鹏. 香梨声振特性分析及内部品质的无损检测研究[D]. 石河子: 石河子大学, 2016.
	WANG Z P. Study on Acoustic Vibration Characteristics and Non-destruction Detection of Internal Quality of Korla Pear[D]. Shihezi: Shihezi University, 2016.
[32]	MOHARANA S, GOTTUMUKKALA S. Scour monitoring at the downstream of hydraulic structure using piezoelectric transducer[J]. Experimental techniques, 2024, 48(1): 141-157.
[33]	ZHANG H, JI S, WANG K, et al. Detection of early browning in pears based on time–frequency images of vibro-acoustic signals and improved MobileNetV3[J]. Food and bioprocess technology, 2025, 18(3): 2721-2736.
[34]	DING C Q, FENG Z, WANG D C, et al. Acoustic vibration technology: Toward a promising fruit quality detection method[J]. Comprehensive reviews in food science and food safety, 2021, 20(2): 1655-1680.
[35]	GUO Z M, WANG M M, AGYEKUM A A, et al. Quantitative detection of apple watercore and soluble solids content by near infrared transmittance spectroscopy[J]. Journal of food engineering, 2020, 279: ID 109955.
[36]	KHALEEFAH S H, MOSTAFA S A, MUSTAPHA A, et al. The ideal effect of Gabor filters and Uniform Local Binary Pattern combinations on deformed scanned paper images[J]. Journal of king Saud university - computer and information sciences, 2021, 33(10): 1219-1230.
[37]	TU B, KUANG W L, ZHAO G Z, et al. Hyperspectral image classification by combining local binary pattern and joint sparse representation[J]. International journal of remote sensing, 2019, 40(24): 9484-9500.
[38]	LAKSHMI SOWJANYA U, KRITHIGA R. Decoding student emotions: An advanced CNN approach for behavior analysis application using uniform local binary pattern[J]. IEEE access, 2024, 12: 106273-106284.
[39]	PAN Z B, HU S Q, WU X Q, et al. Adaptive center pixel selection strategy in Local Binary Pattern for texture classification[J]. Expert systems with applications, 2021, 180: ID 115123.
[40]	KHAN K A, SHANIR P P, KHAN Y U, et al. A hybrid Local Binary Pattern and wavelets based approach for EEG classification for diagnosing epilepsy[J]. Expert systems with applications, 2020, 140: ID 112895.
[41]	ZHAO X, LI K Y, LI Y X, et al. Identification method of vegetable diseases based on transfer learning and attention mechanism[J]. Computers and electronics in agriculture, 2022, 193: ID 106703.
[42]	WOO S, PARK J, LEE J Y, et al. CBAM: Convolutional Block attention module[C]// Computer Vision-ECCV 2018. Cham, Germany: Springer, 2018: 3-19.
[43]	赵康, 查志华, 李贺, 等. 基于声振信号对称极坐标图像的苹果霉心病早期检测[J]. 农业工程学报, 2021, 37(18): 290-298.
	ZHAO K, ZHA Z H, LI H, et al. Early detection of moldy apple core using symmetrized dot pattern images of vibro-acoustic signals[J]. Transactions of the Chinese society of agricultural engineering, 2021, 37(18): 290-298.
[44]	ODHIAMBO OMUYA E, ONYANGO OKEYO G, WAEMA KIMWELE M. Feature selection for classification using principal component analysis and information gain[J]. Expert systems with applications, 2021, 174: ID 114765.
[45]	SHAO R P, HU W T, WANG Y Y, et al. The fault feature extraction and classification of gear using principal component analysis and kernel principal component analysis based on the wavelet packet transform[J]. Measurement, 2014, 54: 118-132.
[46]	XU Y P, LI Z, HUANG G L, et al. Application of multi-source information fusion based on D-S evidence theory in insulation defect identification of DC XLPE cable[J]. High voltage, 2021, 6(4): 599-607.
[47]	吕中亮. 基于变分模态分解与优化多核支持向量机的旋转机械早期故障诊断方法研究[D]. 重庆: 重庆大学, 2016.
	LÜ/LV/LU/LYU) Z L. Research on incipient fault diagnosis methods for rotating machinery based on VMD and optimized MSVM[D]. Chongqing: Chongqing University, 2016.
[48]	杜伟, 房立清, 齐子元, 等. 基于LFOA优化多核支持向量机的液压泵故障诊断[J]. 机械设计与研究, 2018, 34(1): 137-140, 144.
	DU W, FANG L Q, QI Z Y, et al. Fault diagnosis of hydraulic pump based on LFOA optimized multi-kernel SVM[J]. Machine design & research, 2018, 34(1): 137-140, 144.
[49]	张帅. 基于优化型混合核函数支持向量机的个人信用评估[D]. 太原: 太原理工大学, 2020.
	ZHANG S.Personal Credit Evaluation Based on Optimized Mixed Kernel Function Support Vector Machine. Taiyuan: Taiyuan University of Technology, 2020.
[50]	KHAN M, HOODA B K, GAUR A, et al. Ensemble and optimization algorithm in support vector machines for classification of wheat genotypes[J]. Scientific reports, 2024, 14(1): ID 22728.
[51]	LIU G Y, HAN X, TIAN L, et al. ECG quality assessment based on hand-crafted statistics and deep-learned S-transform spectrogram features[J]. Computer methods and programs in biomedicine, 2021, 208: ID 106269.

数据集类型		筛选的敏感特征参数	数据集类型		筛选的敏感特征参数
时域信号统计特征		TD ₂、TD ₇、TD ₁₈、TD ₅ 、 TD ₉、TD ₁₁、TD ₁₀、TD ₆	频域信号统计特征		FD ₇、FD ₂、FD ₃、FD ₈
时域 SDP图	GLGCM	TD_G ₉、TD_G ₁₃、TD_G ₆ 、 TD_G ₁₅、TD_G ₁₀、TD_G ₇	频域GASF图	GLGCM	FD_G ₆、FD_G ₉、FD_G ₁₃ 、 FD_G ₁、FD_G ₁₁、FD_G ₁₄
时域 SDP图	ULBP	TD_U ₆、TD_U ₂₈、TD_U ₁₃、TD_U ₉、TD_U ₃₀、TD_U ₅₂、TD_U ₂₄、TD_U ₁₆、TD_U ₁₂、TD_U ₃₆、TD_U ₇、TD_U ₄₆、TD_U ₄₂、TD_U ₂₅、TD_U ₁₀	频域GASF图	ULBP	FD_U ₁₀、FD_U ₃₃、FD_U ₂₀、FD_U ₄₉、FD_U ₃₅、FD_U ₁₈、FD_U ₂₄、FD_U ₂₈、FD_U ₄₅、FD_U ₁₁、FD_U ₅₄、FD_U ₃₂、FD_U ₆、FD_U ₁₅ 、FD_U ₃₉、FD_U ₁₃、FD_U ₂
时域GASF图	GLGCM	TD_G ₈、TD_G ₁₄、TD_G ₁₁ 、 TD_G ₉、TD_G ₁₀	频域GADF图	GLGCM	FD_G ₆、FD_G ₉、FD_G ₁₃ 、 FD_G ₁₁、FD_G ₁₄、FD_G ₁
时域GASF图	ULBP	TD_U ₁₂、TD_U ₃、TD_U ₂₁、TD_U ₆、TD_U ₃₅、TD_U ₂₂、TD_U ₁₅、TD_U ₃₈、TD_U ₂₆、TD_U ₁₀、TD_U ₄₂、TD_U ₁₄、TD_U ₃₃、TD_U ₅₀	频域GADF图	ULBP	FD_U ₁₆、FD_U ₂₅、FD_U ₈、FD_U ₃₂、FD_U ₂₀、FD_U ₇、FD_U ₄₃、FD_U ₅₀、FD_U ₂₁、FD_U ₅、FD_U ₁₂、FD_U ₃₈、FD_U ₂₃、FD_U ₄₄、FD_U ₁₅、FD_U ₃₁
时域GADF图	GLGCM	TD_G ₈、TD_G ₁₄、TD_G ₉ 、 TD_G ₁、TD_G ₁₃	时频域ST图	GLGCM	TFD_G ₁₀、TFD_G ₁₃、TFD_G ₅、TFD_G ₁₅、TFD_G ₆、TFD_G ₄、TFD_G ₁₄
时域GADF图	ULBP	TD_U ₃、TD_U ₂₀、TD_U ₁₅、TD_U ₂₃、TD_U ₈、TD_U ₂₇、TD_U ₅₃、TD_U ₁₉、TD_U ₂₄、TD_U ₄₅、TD_U ₇、TD_U ₁₃、TD_U ₃₆、TD_U ₄₂	时频域ST图	ULBP	TFD_U ₉、TFD_U ₂₃、TFD_U ₁₆、TFD_U ₃₂、TFD_U ₁₄、TFD_U ₄₃、TFD_U ₃₈、TFD_U ₁₁、TFD_U ₂₅、TFD_U ₅₈、TFD_U ₁₂、TFD_U ₃₀、TFD_U ₄₄、TFD_U ₅₂、TFD_U ₂₇、TFD_U ₈、TFD_U ₂、TFD_U ₁₅、TFD_U ₄₆、TFD_U ₆

特征		分类准确率/%
时域特征	统计特征	75.75
	TD-SDP-GLGCM	79.42
	TD-SDP-ULBP-PC	76.83
	TD-GASF-GLGCM	80.03
	TD-GASF-ULBP-PC	78.34
	TD-GADF-GLGCM	81.62
	TD-GADF-ULBP-PC	85.20
频域特征	统计特征	78.31
	FD-GASF-GLGCM	75.27
	FD-GASF-ULBP-PC	80.59
	FD-GADF-GLGCM	83.20
	FD-GADF-ULBP-PC	88.94
时频域特征	TFD-ST-GLGCM	93.53
时频域特征	TFD-ST-ULBP-PC	91.62

分类器	各单一域深层特征（V_i ）	总体分类准确率（A_i ）/%
ELM	TD-GADF	89.05
	FD-GADF	91.27
	TFD-ST	95.49

特征类型	特征名称	分类器	证据体	各证据体的基本概率分配值m_i （L_j ）
特征类型	特征名称	分类器	证据体	L ₁	L ₂	L ₃	Θ
浅层特征	TD-GADF-ULBP-PC	MSVM	E ₁₁	0.295	0.273	0.306	0.126
	FD-GADF-ULBP-PC		E ₁₂	0.300	0.292	0.323	0.085
	TFD-ST-GLGCM		E ₁₃	0.312	0.306	0.351	0.031
深层特征	TD-GADF	ELM	E ₂₁	0.298	0.286	0.314	0.102
	FD-GADF		E ₂₂	0.312	0.294	0.366	0.028
	TFD-ST		E ₂₃	0.319	0.298	0.382	0.001

分类模型	模型参数							分类准确率/%
分类模型	α	c	σ	d	l	ω_j	b_j	分类准确率/%
IPSO-MSVM-DS	0.638	68.325	9.651	2	—	—	—	98.52
Adam-IResNet50-IPSO-ELM-DS	—	—	—	—	10	2.542	0.121	99.36