Scholarlink

<p>Crime scene investigation (CSI) image classification plays a vital role in forensic analysis and greatly enhances police investigative efficiency by providing timely and accurate visual information interpretations. Imbalanced class distributions and inefficient hyperparameter tuning pose challenges to existing classification models, and this leads to unsatisfactory performance. To overcome these challenges, this paper introduces the Imbalanced Maximizing-AUC Proximal Support Vector Machine (ImAUC-PSVM) that integrates AUC optimization into the process of classification itself to enhance resistance against class imbalances. The model also employs the Artificial Bee Colony (ABC) algorithm for effective hyperparameter tuning, removing the computational inefficiencies of traditional tuning methods. The effectiveness of ImAUC-PSVM is reflected by its application on three diverse datasets—CIIP-CSID 1, CIIP-CSID 2, and CIIP-CSID 3. The result shows that it outperforms other techniques, particularly in handling class imbalances and improving computational efficiency. These advancements propel forensic technology to the next level and establish the new standard for automated image analysis in law enforcement.</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

Crime scene investigation image classification using AUC and ABC optimization

  • 期刊论文

  • 发表日期：

    2025年9月9日

查看全文
  • Yang Lei

摘要

Crime scene investigation (CSI) image classification plays a vital role in forensic analysis and greatly enhances police investigative efficiency by providing timely and accurate visual information interpretations. Imbalanced class distributions and inefficient hyperparameter tuning pose challenges to existing classification models, and this leads to unsatisfactory performance. To overcome these challenges, this paper introduces the Imbalanced Maximizing-AUC Proximal Support Vector Machine (ImAUC-PSVM) that integrates AUC optimization into the process of classification itself to enhance resistance against class imbalances. The model also employs the Artificial Bee Colony (ABC) algorithm for effective hyperparameter tuning, removing the computational inefficiencies of traditional tuning methods. The effectiveness of ImAUC-PSVM is reflected by its application on three diverse datasets—CIIP-CSID 1, CIIP-CSID 2, and CIIP-CSID 3. The result shows that it outperforms other techniques, particularly in handling class imbalances and improving computational efficiency. These advancements propel forensic technology to the next level and establish the new standard for automated image analysis in law enforcement.