Inspection of Cereal Grains
Inspection of cereal grains is an exceptionally mundane and repetitive task, and to some extent contrasts with many of the examples given in the previous chapter in that the emphasis is on detecting contaminants rather than finding manufacturing faults. This chapter presents three case studies that air and solve important theoretical questions.
• the limitations of the immediately obvious technique—global thresholding.
• the value of morphological filtering.
• the unusual need for median filtering as a morphological operation.
• the use of bar (linear feature) detectors for locating insects.
• how the vectorial bar detector operator is optimized.
• how sampling can be used to speed up object location.
• how the outputs of oblique template masks can optimally be combined.
This chapter focuses on the problems of achieving real-time operation and largely solves these problems for less demanding situations (flow rates up to ∼300 items per second). For more demanding situations, dedicated hardware accelerators are needed, as discussed in Chapter 28. At a more detailed level, the balance between false positives (false alarms) and false negatives needs to be optimized, as discussed in Section 24.7.