Shinichi Meguro

Gray‐level image recognition based on multiple cell‐features

This paper proposes a gray-level image recognition algorithm based on multiple cell-features for general-purpose model matching using the generalized Hough transform, which removes the instability of conventional edge-segment-based feature description and simplifies its feature matching. n n n nImage features are derived in terms of four basic representations: density; edge direction; edge density; and global edge configuration. These are integrated into a multidimensional feature vector for each “cell,” which is a sampled subregion of the image. This method provides powerful and stable description since it describes contour and area properties as well as local and global features. It also allows the effective use of generalized Hough transform matching. n n n nMultiple cell features are extracted systematically by combining three fundamental operations. These consist of two feature extraction operations, extended convolution and radial traverse probing, and a data compression operation which converts pixel features to cell features by calculating a histogram of extracted features for each cell. Feature dimensionality is decreased efficiently by selecting feature kinds and salient cell positions based on the F-ratio calculated from training samples. Using the gray-level images of prepaid telephone cards, a recognition experiment with 20 different card images and a positioning experiment with overlapping cards has been carried out.

Flexible gray-level vision system based on multiple cell-feature description and generalized Hough transform

This paper presents a flexible and highly-reliable gray-level vision system based on multiple cell-feature descriptions using only three basic operation modules: extended convolution radially traversing probing and histogram compression. The generalized Hough transform is introduced as a universal method for object model matching. Model learning is automatically performed by acquiring image samples while rotating each object. A prototype system demonstrates successful recognition of mechanical parts.

Automatic inspection system for full-color printed matter

We have developed a full-color inspection system that detects defects such as pin-holes and stains on surface of prepaid cards. The system consists ofa card conveyer unit with an automatic exchange mechanism and a custom designed image processing unit with a three-stage pipeline structure. Inspection performance ofO. 7 seconds/card with a defect detection resolution of 0. 1 5mm diameter is achieved with a new algorithm that reflects human visual criteria.