| Abstrakt: |
Agriculture is considered the backbone of any country’s economy as it contributes significantly towards the improvement in peoples’ life standards and economic growth. However, pests’ attack on crops is one of the serious issues in the agricultural sector that results in crop quality degradation. Specifically, pests, germs and weeds cause massive damage to the crops, thereby resulting in low market value for the final agricultural products. Thus, the detection and classification of pests in damaged crops become essential for making reactive decisions associated with the recovery of crops. This pest detection and classification in damaged crops can bring about a potential gain in the agricultural sector. At the same time, a traditional or manual method of pest detection and classification is identified to be time-consuming and completely dependent on the expertise of the investigator. At this juncture, artificial intelligent techniques confirmed better pest detection and classification with maximum accuracy and minimum time. In this paper, a comprehensive review of the different artificial intelligence-based pest detection and classification schemes is presented with their merits and limitations. This review particularly demonstrated the potentiality of the existing machine and deep learning techniques contributed to the literature over recent years. Analysis and comparison of these two methods reveal that current agricultural pest data resources make transfer learning the better option. It depicts the feasible future scope of work which could be carried out depending on the shortcomings identified in the literature review. It also portrayed the motivation of the review, challenges faced during the process of ML and DL-based pest detection and classification and frameworks also mentioned. |
Full text from SpringerLink