IEEE Robotics & Automation Magazine - December 2023 - 39

TABLE 8. Performance matrix of USMA-BOF-based SVM classifier. (Continued)
CATEGORIES
Peach leaves
PARAMETERS
Precision
Recall
Accuracy
F1-Score
Average
Precision
Recall
Accuracy
F1-Score
WOA-BOF
0.6979
0.5723
0.6807
0.6288
0.7195
0.6842
0.6919
0.7014
APSO-BOF
0.5606
0.6393
0.5494
0.5973
0.4905
0.5539
0.6464
0.5202
TABLE 9. Performance matrix of USMA-BOF-based MLP classifier.
CATEGORIES
Apple leaves
PARAMETERS
Precision
Recall
Accuracy
F1-Score
Cherry leaves
Precision
Recall
Accuracy
F1-Score
Grape leaves
Precision
Recall
Accuracy
F1-Score
Peach leaves
Precision
Recall
Accuracy
F1-Score
Average
Precision
Recall
Accuracy
F1-Score
WOA-BOF
0.7262
0.6585
0.706
0.6906
0.5797
0.7797
0.606
0.6649
0.6585
0.7403
0.6363
0.6970
0.7797
0.706
0.5747
0.7410
0.7403
0.606
0.7262
0.6664
APSO-BOF
0.6989
0.6989
0.6818
0.6989
0.6787
0.7232
0.707
0.7002
0.6989
0.7403
0.6949
0.7190
0.7232
0.6818
0.6767
0.7018
0.7403
0.707
0.6989
0.7232
GWO-BOF
0.6343
0.6636
0.6696
0.6486
0.706
0.703
0.707
0.7044
0.6636
0.6535
0.6696
0.6585
0.703
0.6696
0.704
0.6858
0.6535
0.707
0.6343
0.6791
SMA-BOF
0.7262
0.6585
0.706
0.6906
0.5797
0.7797
0.606
0.6649
0.6585
0.7403
0.6363
0.6970
0.7797
0.706
0.5747
0.7410
0.7403
0.606
0.7262
0.6664
USMA-BOF
0.7552
0.7268
0.7342
0.7407
0.7342
0.8108
0.7358
0.7706
0.7256
0.7692
0.7226
0.7467
0.8108
0.7342
0.7322
0.7706
0.7699
0.7352
0.7552
0.7521
GWO-BOF
0.6474
0.6737
0.6454
0.6602
0.6087
0.6515
0.6767
0.6293
SMA-BOF
0.6979
0.5723
0.6807
0.62889
0.7195
0.6842
0.6919
0.7014
USMA-BOF
0.7258
0.7006
0.7079
0.7130
0.7482
0.7115
0.7195
0.7294
PERFORMANCE ANALYSIS OF
CLASSIFICATION METHOD
In this section, the performance analysis of the USMA-BOFbased
classification method is performed using 800 images
obtained from the PlantaeK repository [29], consisting of
infected and noninfected leaf images. Two classifiers, namely
SVM and MLP, are used to perform classification based on
the USMA-BOF algorithm. The dataset is divided into 80%
training and 20% testing datasets. The parameter settings of
the two classifiers during experimentation are shown in Table 7.
For the comparative analysis, several performance metrics are
used to estimate the algorithm efficiency, such as precision,
recall, accuracy, and F1-measure, and the results obtained are
shown in Table 8 and Table 9 for SVM and MLP, respectively.
The extensive experimental analysis shows the effectiveness
and efficiency of the proposed USMA-BOF-based classifiers
in comparison with other methods. The main limitation of the
proposed work is that currently it is evaluated on a small dataset
with only 800 leaf images. To check its scalability, it needs
to be tested on large datasets.
CONCLUSION AND FUTURE WORK
Manual identification of plant leaf disease is a labor-intensive
task and requires a lot of expertise. Automatic classification
DECEMBER 2023 IEEE ROBOTICS & AUTOMATION MAGAZINE
39
IEEE Robotics & Automation Magazine - December 2023

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - December 2023
