Griffith University
Plant Disease Detection Based on Computer Vision and Machine Learning

Are you eligible for Breaking Barriers Category:


My project aims to develop advanced methods for plant disease detection based on computer vision and machine learning. In agriculture, plant disease remains a major threat to the yield of agricultural production. They cause significant economic losses and reduce food security at national and global levels. To maintain the plant health and improve yields, farmers throughout the world struggle to prevent and eradicate various disease from their plants. However, farmers have substantial difficulty in managing plant diseases because they lack access to information and technology that could help them raise healthy crops. In recent years, computer-vision based models have been widely used for smart farming and precision agriculture. However, there are significant challenges posed to developing fast and accurate detection models. My work is to solve these challenging technical bottlenecks, and develop real-time, low-cost, non-destructive plant disease detection methods based on computer vision and machine learning.

BENEFIT – A description of the benefit of your work to Queensland (max 500 words)

Queensland’s horticulture sector contributes 22% (about $4.5 billion) gross value of production of the Queensland economy according to report of “Queensland AgTrends”, and more than 30,000 employees are working in this sector. However, the threat of plant pest and diseases often has devastating impact on agriculture. For example, the most recent and severe outbreak of stem rust in several barley crops in south-east Queensland from October 2018. Strawberries are also an important crop in Queensland, this industry is worth about $130 million to the Queensland economy. However, Strawberry angular leaf spot presents a significant risk to Queensland strawberry industry. The disease can cause stress to the plant and reduces fruit quality that downgrades markets value.

The most common plant disease diagnosis is based on visual detection. When there are symptoms on a plant, farmers first recognize the disease totally based on their growing experience. However, diseases are usually caused by a variety of factors, identifying diseases only by appearance of leaves by human eyes may cause misdiagnosis which can lead to misuse of chemicals leading to the emergence of resistant pathogen strains, increased input costs, and more outbreaks with significant economic loss and environmental impacts.

Biotechnology has also been widely used for assisting farmers in managing different diseases affecting plants. However, the laboratory testing for disease diagnosis may take days or even weeks to complete by professional experts and are, in some time, relatively insensitive. Delays in detecting disease may miss the best time for treating disease at its early stages, causing severe disease spreads within a few days. 

Motivated by the above reasons, I propose to use machine learning and computer vision-based technologies to assist the plant disease detection. The learned models will be easily installed on smartphones for farmers to use, and benefit the Queensland’s agriculture sector in the following aspects:

 1), Precise: these disease detection applications are developed by making full use of huge plant disease images databases, which provides comprehensive information from various plant disease. In machine learning, the more data provided for the model, the more accurate the model will be. 

2), Real-time: If farmers suspect that their plants may be diseased, they can simply take photos for plant leaves by the smartphone cameras and then identify the disease on these photos by using the installed applications. 

3), Low-cost: the farmers do not need any other equipment except smartphones to detect and analysis the disease on their crop, and also the proposed project is committed to designing user-friendly applications for the general public and farmers who usually don’t have professional knowledge in plant disease, thus the methods produced within my project will be a simple and economical way for plant disease detection among farmers.  

 4), Non-destructive: Compared with other destructive disease detection methods which requires the professional researchers to pick a lot of plant leaves to do experiments, the proposed method analyse the disease only by the leaf images.

The above analysis show that my work will enable farmers to make management decisions of plant disease in real time, regardless of their locations, bring huge contributions for preventing disease spread, minimizing economic loss, and protecting biosecurity of Queensland. 

ROLE MODEL – Why do you think you are a good role model for women and girls aspiring to work in STEM? (max 500 words)

I am a good role model for women and girls in the following aspects:

I am working in the forefront of international scientific research within the field of new hot spot, and have achieved the state-of-the-art results. My research achievements on computer vision and machine learning are highly recognised by international peers.

Research shows that interaction with STEM professionals/role models have positive impact on students’ STEM interest. Therefore, it is reasonable to expect that girls are more motivated to engage in subjects related with STEM fields, such as computer vision, after interacting with female role models in STEM than before doing so. Based on this, I provide opportunities for female students to enhance their hands-on skills, improving their competitiveness in job market. For those students who are interested in research and pursuing a research degree, research training plans will be made according to their interests. From the eyes of students, I am a supervisor and also a friend who can always provide supports when they need. 

My research not only can directly change people’s daily life, but also will facilitate the employment in the downstream agriculture sector. Male employees always dominate the agriculture field since traditional agriculture is driven by manual labor where men have their natural advantages. Then we will see a contradictory situation that on the one hand females cannot find o job around their villages, and on the other hand there is not enough workforce engaging in agriculture field. The computer-aided technologies can solve this problem. The females will be trained to control the computer to use the installed plant monitoring software to assistant agricultural monitoring and management. This working pattern can not only solve the females’ employment problems, but also can improve the production precision and efficiency.

To summarize, I not only get recognised in top tier research groups,  and also do mentors to help broaden the perspectives of who can work in the STEM field. As an indirect potential, my work also can improve the living and working status of females in agriculture area. Therefore, I think I am a good role model for females aspiring to work in STEM.

ENGAGEMENT – Describe any STEM promotion or engagement activities that you have undertaken, including both scientific and non-scientific audiences, particularly with women and girls (maximum 500 words)

Higher degree research training and experience can enhance students’ creativity, critical thinking, and self-learning skills which are very important abilities for those who want to develop themselves as researchers or team leaders. However, it is reported that only 16% of engineers in Australia are women, there are even less female PhD candidates and researchers in HDR component. For example, I major in computer vision and image processing (Engineering), for the past years previously, I am the only female researcher in my lab. 

What’s the reason behind this situation? For the past years, when someone asked me what’s your major, and I answered “computer vision”, they always had surprised expressions on their face.  And I asked them why they are so surprise, and what’s their idea about researcher who are working on computer vision. They said the computer vision are male-dominated area and they must be working on some high-tech technologies that are far away from their lives, like tracking, holographic projection, robots, etc. 
Another important reason behind this is that most undergraduate and graduate female students are focusing on learning courses and do not have much opportunities to get hands on skills, which is difficult for them to bridge the gap between theories and practical use. Over time, they lose their interests and confidence to work in STEM field.

To improve this situation  and change people’s mindset, I actively share my research experiences and stories with female undergraduates and graduates, and provide them some internship opportunities to let them know what’s computer vision,  what’s research projects, and what’s the benefits of these projects to our society.  For very long time, girls have the fixed mindset of “ability is inherited, criticism means failure, improvement is impossible” in STEM subjects. So growing girls’ mindset and teaching girls to understand “struggle is part of learning process” is important. To teach them to know not only service industries but also STEM can benefit people’s life, and STEM and our daily life will only be positively effected by engaging a diverse range of people.