Team members: Amreen Khan, Moitrish Majumdar, Nidhi, Vandana Prasad
Agriculture is the primary source of livelihood for about 58 per cent of India‟s population. An activity which employs a majority of the country‟s workforce is plagued with losses and inefficiency, a large chunk of which arise from diseases and pest infestation of crops, fruits and vegetables. We explored the different dimensions of the issue and constructed a potential solution which can enable the reduction of losses throughout the entire sector.
Ultrasound imaging (sonography) is a diagnostic medical procedure that uses high-frequency sound waves to produce dynamic visual images of organs, tissues or blood flow inside the body. The sound waves are transmitted to the area to be examined and the returning echoes are captured to provide the physician with a „live‟ image of the area. Ultrasound does not require the use of ionizing radiation, nor the injection of toxic contrast agents. These advantages enable its use in the field of agricultural science and crop disease diagnosis.
Image processing is a method to perform some operations on an image, in order to get an enhanced image or to extract some useful information from it. It is a type of signal processing in which input is an image and output may be image or characteristics/features associated with that image. Machine learning is a powerful tool to make predictions about an image. Given a large set of images, machine learning algorithms are able to classify an image based on certain characteristics. Thus, it can be employed in remote applications which do not require any technical knowledge and can provide a swift solution to the user.
The advantages of these tools are combined and presented in a mobile application, which can be deployed with relative ease on several platforms. Portable ultrasound machines and a simple smartphone can be enough to provide an efficient solution to the farmer. Inputs obtained in the form of an ultrasound scan or a superficial image from a usual smartphone camera and processing techniques using machine learning algorithms can accurately detect and diagnose a variety of diseases affecting crops, fruits and vegetables.
Our initial brainstorming and mind map consisted of literature reviews of existing techniques and a study of the most common crop diseases. Insights from the field introduced a new aspect to the problem, i.e the detection of internal damage and worms in fruits. Re-brainstorming addressed these issues and led to the consideration of ultrasound imaging for efficient detection. We also ruled out other possible solutions such as biospeckle techniques. The prior art search revealed the multiple existing solutions and the issues encountered in each. We studied the various stakeholders in the problem and considered insights from field experts, targeted users and academic experts. Finally, the prototyping and proof of concept is presented in three parts, which consist of the input stage (ultrasound scanned image), processing techniques (using image processing and machine learning) and the output is modeled in an app design. Finally, the way ahead was explored, and potential developments were described.