Machine Vision

Machine vision plays an important role in modern agriculture to improve productivity and sustainability in crop and livestock management, from monitoring crop health, and pest control to tracking animal behavior and diagnosing diseases. Camera images and videos contain a wealth of information about the agricultural environment, allowing, for instance, to monitor crop and animal health and detect diseases. This information can be used by farmers, breeders and scientists to get insight in processes and make data-driven decisions. Moreover, machine vision is essential for the automatic control of machines and robots, enabling tasks like sorting, pest control, and selective harvesting.

In this course, you learn about important machine-vision methods and how to apply them to extract information from images and videos. We will go through the whole pipeline, from image acquisition and noise filtering to segmentation, detection, and the control of a machine. Apart from this classical machine-vision pipeline, you will also learn about the use of machine learning and specifically deep learning to process image data. Most of the course will be focused on 2D colour images, but you will also learn to deal with spectral images, and with 3D point clouds

The main activities used in this course are tutorials and projects. The tutorials are interactive lectures where new theories are presented in small chunks, alternated with programming and pen-and-paper exercises to process the information. In the projects, you will work on project assignments by combining the previously learned methods into an integrated solution for a given agricultural problem.

• Discuss important aspects of data acquisition

• Explain key theories and methods in traditional machine vision

• Discuss advantages and disadvantages of using machine learning for computer vision

• Apply these methods to real-world agricultural problems using a programming language

• Integrate vision methods to physically control a machine

• Analyse an agricultural problem and conclude which sensors and processing methods to apply

• Create a machine-vision algorithm to solve an agricultural problem and evaluate its performance

• Assignment other (50%) The project grade based on a weighted average over a predefined set of project assignments, which will be graded based on a demonstration, explanation and discussion (in pairs). In case of a failed project students can hand in a revised version at the next re-sit period
• Written test with open and closed questions (50%)

INF22306 Programming in Python

• To be announced.