Scholarlink

<p>Implementing sound-based pest detection systems in large fields comes with challenges, including the complexity of acoustic data, interference from environmental noise, and the need for scalable, real-time solutions capable of processing data over vast areas. To address these challenges, we propose an innovative pest detection system that integrates advanced sound analytics with deep learning techniques. This study proposes a feature selection technique that includes the real-valued discrete Gabor transform algorithm and linear frequency Cepstrum coefficients (LFCC) algorithm, which helps analyze the frequency spectrum and capture critical features. The approach operates in four stages: pre-processing, feature selection, feature matrix generation and sound classification. This study proposes primarily, that data is collected and pre-processed, followed by the feature selection technique including the Real-Valued Discrete Gabor Transform algorithm and Linear Frequency Cepstrum Coefficients (LFCC) algorithm, which help analyze the frequency spectrum and capture critical features. These selected features are then generated as a feature matrix. After that, the pest sound is classified using Modified EfficientNet. To optimize the model’s parameters and enhance its performance and accuracy, the Hippopotamus Optimization Algorithm (HOA) is integrated with EfficientNet. The dataset was utilized to validate the suggested approach, with accuracy levels of over 98% and specificity is 97%, respectively. These results demonstrate superior classification efficiency compared to existing pest detection systems.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Advanced sound-based pest detection in agriculture using deep learning and adaptive optimization

  • 期刊论文

  • 发表日期：

    2025年6月4日

查看全文
  • Sahana Lokesh R,
  • K. SailajaKumar,
  • R. S.Soundariya,
  • Kavitha K S,
  • S. Reshma

摘要

Implementing sound-based pest detection systems in large fields comes with challenges, including the complexity of acoustic data, interference from environmental noise, and the need for scalable, real-time solutions capable of processing data over vast areas. To address these challenges, we propose an innovative pest detection system that integrates advanced sound analytics with deep learning techniques. This study proposes a feature selection technique that includes the real-valued discrete Gabor transform algorithm and linear frequency Cepstrum coefficients (LFCC) algorithm, which helps analyze the frequency spectrum and capture critical features. The approach operates in four stages: pre-processing, feature selection, feature matrix generation and sound classification. This study proposes primarily, that data is collected and pre-processed, followed by the feature selection technique including the Real-Valued Discrete Gabor Transform algorithm and Linear Frequency Cepstrum Coefficients (LFCC) algorithm, which help analyze the frequency spectrum and capture critical features. These selected features are then generated as a feature matrix. After that, the pest sound is classified using Modified EfficientNet. To optimize the model’s parameters and enhance its performance and accuracy, the Hippopotamus Optimization Algorithm (HOA) is integrated with EfficientNet. The dataset was utilized to validate the suggested approach, with accuracy levels of over 98% and specificity is 97%, respectively. These results demonstrate superior classification efficiency compared to existing pest detection systems.