Detection of Tomatoes Using Artificial Intelligence Implementing Haar Cascade Technique
- 29 Downloads
The twenty-first century is consumed with the automation of the world around us. Almost everything from things as easy as walking to driving a car has been automated. But according to our study, one of the least researched areas for automation has been agriculture. There are endless possibilities to the possibilities of automating the agricultural field yet many chose not to walk down this path. Machine learning is a sub-branch of artificial intelligence, it has been applied to image processing and this intelligence is demonstrated by machines in contrast to the natural intelligence displayed by humans. The artificial intelligence can be used in many sectors like transportation, finance, health care and banking and also it can be used in image processing and it helps us to implement object detection to detect and recognize the objects from our given input images and video. This is why we have dedicated this paper to help make the life of many farmers far easier. We decided on constructing low-cost agricultural robot architecture. This robotic architecture reduces the work upon farmers and helps them survey their farm area in a matter of minutes. The robotic architecture can survey the fields and assess every tomato to identify the perfect and ripe tomatoes that can be harvested. Hence, the farmers won’t have to waste their time searching and can go only to the designated areas of their farm for the harvest. We were able to accomplish this with the help of Haar Cascade. We developed a model that can be run upon various servers such as Windows, Unix or even a Linux Server. We were able to train our model by identifying the images with a ‘positive’ and ‘negative’ label. All those images which contained a background without a tomato were labelled as ‘negative’ images, while those that contained a tomato were labelled as ‘positive’ images. The goal of our model implementing Haar Cascade was to create a negative image far larger than the positive images. We were able to make this possible by implementing large data sets that consisted of nearly 200 images of tomatoes. These large data sets are then further clearly segregated to identify the state of every tomato. Haar Cascade classifier provides high accuracy even the images are highly affected by the illumination. The Haar Cascade classifier has shown superior performance with simple background images.