Image analysis and interpretation by computers has many potential applications for guidingor controlling agricultural processes. However, research from industrial applications cannot be applied directly to agriculture because of problems such as the biological variability of objects or workpieces, and difficulty in interpreting unstructured environments. This paper describes the process of image analysis, its application to agriculture, and thegeneric problems particularly relevant to agricultural processes. Priority areas for further research are suggested. Ways of including knowledge in algorithms should be studied. Specific applications should also be studied including the practical problems of integrating image processing into larger systems. © 1991 Silsoe Research Institute.