Masakazu Ejiri

A process for detecting defects in complicated patterns

A method of detecting defects in complicated patterns such as printed circuit boards is described. This method employs two-dimensional nonlinear logical filtering. In the first stage, small portions of the pattern are extracted as potential examples of the defects. These examples include small portions near the boundary of the pattern which are due to the unevenness caused by spatial digitizing. The subsequent process eliminates these small portions on the boundary, thus leaving only the real defects. This method operates without using any prememorized standard patterns and, instead, utilizes a pseudo-standard pattern generated from the input pattern itself. This eliminates the positioning problem between the input and the standard pattern, as well as the need for any memory for the pattern. The effectiveness of this method was confirmed by digital computer simulation, and afterward, a defect extraction device was developed. This device makes it possible to detect the defects in complicated patterns in a real-time mode. It therefore has promise in facilitating the automation of tedious visual inspection processes such as those for printed circuit boards.

An automatic wafer inspection system using pipelined image processing techniques

An automatic wafer pattern inspection system has been developed that can detect defective patterns 6 mu m or larger in multilayered wafer patterns at a speed 30 times faster than that of a human inspector. The false-alarm rate is less than 0.5 occurrences/chip. This performance is achieved mainly by the use of a special comparison method between two adjacent patterns obtained through a single optical setup, and also by the use of digital design pattern data (CAD data). The main functions of the design pattern data are to specify the inspection area, to designate optimum parameters for inspection, and to separate defective portions into different layers, thereby facilitating the classification of the defects. All image processing is performed in one pass by a high-speed pipeline-structured image processor that can analyze an input image signal at a 7 MHz video rate. >

A Transistor Wire-Bonding System Utilizing Multiple Local Pattern Matching Techniques

A fully automatic transistor wire-bonding system utilizing image-processing techniques is described. The system is composed of a minicomputer, image processors, and a maximum of 50 wire-bonding machines. Each machine has a TV camera installed to pick up, through a microscope, the image of transistor chips fed into the machine. In this system, one image processor services several machines in a group and analyzes the image signals in a time-sharing mode to determine the chip position at each machine. From the detected coordinates, an XY servomechanism in each machine is activated to bond and stretch wires between the electrodes on the chip and the outer leads. The recognition of electrode positions is accomplished by the multiple local pattern matching method in which redundancy is used in achieving a high recognition rate. An analysis of the system and basic features of the position recognition algorithm are also described. This method opens the possibility of economical automation of processes possible only by the human eye to date.

A Prototype Intelligent Robot that Assembles Objects from Plan Drawings

An intelligent robot that recognizes and assembles three-dimensional objects by means of vidicon cameras, an articulated mechanical hand, and a digital computer is described. Its problem-solving functions include three essential parts: the recognition of macro-instructions from a human master, the recognition of the objects to be handled, and the decision making for executing the necessary tasks. The instruction is in the form of a three-view plan of a simple polyhedral assemblage whose overall spatial configuration is recognized together with its component parts. In this process the set of planes of the assemblage is disassembled into open shells; these are then reconstructed into closed shells to find the parts by solving a linear equation where the shell vectors are taken into consideration. In the object recognition, all the geometric features of polygonal prisms in the field of vision are extracted to find the specific parts required for the assembly. Further computation is made to search for the assembly procedure and is based on the restraint vectors of each object and consequent disassembly free. Finally, the mechanical hand starts the autonomous manipulation of the parts so as to accomplish the assigned assembly.

A multiprocessor system utilizing enhanced DSPs for image processing

A general-purpose image processor (GPIP) consisting of 64 digital signal processors (DSPs) in a 0.31-m/sup 3/ box is proposed to perform a wide range of image processing tasks. A high-speed DSP called DSP-i has been especially developed for this purpose. It has a highly parallel architecture with a two-level instruction hierarchy, multibank cache, and multiprocessor interface. The DSP-i machine cycle is 50 ns. A novel ring shift register bus architecture offers a flexible structure and an efficient data-exchange method for the system. Along with four proposed operation modes, it cuts the multiprocessing overhead to as little as 20%. The performance of the GPIP is 1000 MOPS (million operations per second).<<ETX>>

Knowledge-directed inspection for complex multilayered patterns

Combining the knowledge-based processing with image processing is considered a key issue in the future of visual inspection of complex patterns such as multilayered semiconductor wafers. However, present technology restricts this combination, mainly because of the exhaustively long time usually required for each type of processing. To cope with this situation, a unique knowledge-directed image processing method is proposed, in which every image processing step is controlled in real time by parametric knowledge driven by design patterns. The resulting structure of the image processor is a pipeline, in which each piece of knowledge is embodied as a combination of a hardware processing unit and control unit. In this paper the types of knowledge and their implementation are explained, and an inspection machine for logic IC wafers based on this pipelined knowledgedirected image processing is introduced.

A recognition algorithm of dashed and chained lines for automatic inputting of drawings

Recognition of lines in drawings is a major task in automatic data acquisition for CAD (computer-aided design). This paper proposes an algorithm for the recognition of dashed and chained lines in graphical images. This algorithm consists of two major steps: the local recognition step which extracts local connectivity of line segments, and the global recognition step which performs route finding of connected segments based on syntax. The syntax is usually specific to each type of drawing, and is represented as rules. The performance of the proposed algorithm is demonstrated based on experiments using LSI cell drawings and topographical maps containing dashed and chained lines as graphical components.

Towards meaningful robotics for the future: Are we headed in the right direction?

Previous and current robotics research is overviewed, and future directions are discussed. Criticism and warning are given to the present approaches that promise little practical application. More promising directions are suggested, though the clear thinking of individual researchers about what is important is essential for opening a new era in robotics.

Neural network approach to path planning for two dimensional robot motion

A method for robot obstacle avoidance and path planning is proposed. The algorithm is based on a camera image feedback loop utilizing a neural network for image processing. The method can successfully generate collision-free paths in a 2D robot workspace containing randomly-placed polygonal obstacles. The control algorithm is simple and robust and has low computational requirements. Controller simulation implemented on a personal computer can generate collision-free robot paths in real time, requiring approximately 1 sec. per robot move.<<ETX>>

Knowledge-based approaches to practical image processing

Recent practical approaches to knowledge-based image processing are discussed. Three application areas are considered: document understanding, drawing understanding and industrial machine vision. The authors indicate that, in the field of document understanding and drawing understanding, the capability of real-time processing is not generally a critical problem. Therefore, a complicated systematized inference mechanism can be utilized for knowledge-based processing. However, in the industrial machine vision field, real-time processing is essential. As knowledge-based processing usually poses a great problem in its processing speed, it must be executed in a highly simplified manner.<<ETX>>