The food security of any country may be jeopardized due to improper management of agricultural insect pests. Accurate pest detection and efficient pest control strategies must be employed in time to grow healthy crops for achieving food security in a country and worldwide. Hence, developing efficient and robust techniques to detect agricultural insect pests using computer vision approaches is one of the essential steps for timely managing insect pests. This chapter presents a short survey of deep learning-based object detection techniques focusing on insect pest detection and associated insect pest image datasets. Subsequently, a transfer learning-based custom You Only Look Once (YOLOv5) model is developed using the publicly available dataset IP102 for detecting agricultural insect pests with the help of computer vision approaches. The hyperparameters of the proposed insect pest detector are optimized using the genetic algorithm-based hyperparameter evolution method. The performance metrics of the proposed insect pest detector are found to be promising.
TopIntroduction
It has been observed that from the post-green revolution decades to the early twentieth century, crop losses due to agricultural insect pests have decreased from 13.6% to 10.8% (Dhaliwal, Jindal, & Mohindru, 2015). India is mostly an agriculture-based economy. However, the farmers and residents in India are dealing with agriculture pests (such as rice leaf roller, grub, aphids, peach borer, red spider, and so on) in their fields as well as house gardens, respectively (Wu, Zhan, Lai, Cheng, & Yang, 2019). Food security in the country is jeopardized due to agricultural insect pests, and agriculture in India suffers a total annual crop loss of approximately US$36 billion (Dhaliwal, Jindal, & Mohindru, 2015). Since the post-green revolution decades to the early twentieth century, crop losses due to agricultural insect pests in India have decreased from 23.3% to 15.7% (Dhaliwal, Jindal, & Mohindru, 2015). Accurate pest detection and efficient pest control strategies must be employed in time to grow healthy crops for achieving food security in the country and worldwide. Hence, developing efficient and robust techniques to detect agricultural insect pests using computer vision approaches is one of the essential steps for timely managing insect pests.
Motivation
Due to the country's rapidly expanding population, the Indian government must ensure food security for its people. However, insect pests play a critical role in destroying healthy crops, making it challenging to manage agriculture-based product production. Farmers are spraying pesticides (chemicals) in an unregulated quantity to combat the problem of insect pests and associated diseases. It harms the crop, soil, water, and the health of the consumers. The majority of the time, pesticides (chemicals) are chosen using a labor-intensive human-centric approach. Farmers identify the insect pest in the crop using manual inspection, which mostly leads to wrong classification of the insect pests due to the high resemblances among the various insect pests. Subsequently, incorrect selection of pesticide (chemical) happens, resulting in severe crop and soil damage as well as a significant financial loss for the farmer. So, the use of the required quantity of pesticides (chemicals) in the crop is dependent on the desired information about the insect pest in the crop, such as where the insect pest is widely distributed (location) and what categories they belong to (classification). Further, agricultural pests are of various sizes ranging from tiny, small, and large. Usually, the insect pests reside over the body or leaf of the crop as overlapping clusters. Usually, there is a lack of proper visibility in the crop field due to the heavily populated plants in the fields. Hence, key research challenges of insect pest detection include high variation in the object sizes, complex background, occlusion, and low illumination (Ahmad, et al., 2022). Hence, these research challenges make object detection even more challenging for insect pest detection applications. Due to the availability of massive computational facilities and large datasets, a data-driven technique like deep learning-based computer vision approaches can efficiently classify and detect agricultural insect pests (Domingues, Brandão, & Ferreira, 2022; Saleem, Potgieter, & Arif, 2021). These are the major motivating factors for carrying out the research on this topic.
Contributions
The key contributions of the research work are briefly outlined as follows.
- •
A short survey of the Deep Learning (DL)-based object detection techniques focusing on insect pest detection is presented.
- •
A brief discussion about the publicly available image datasets for detecting agricultural insect pests is covered.
- •
A transfer learning-based custom You Only Look Once (YOLOv5) model is developed using the publicly available dataset IP102 for detecting agricultural insect pests with the help of computer vision approaches.
- •
The proposed insect detector is improved by optimizing its hyperparameters using the Genetic Algorithm-based hyperparameter evolution method.
- •
A brief comparative analysis of the proposed insect pest detector with state-of-the-art techniques is presented to validate its effectiveness.