Automatic inspection method

• US 5,012,524 A
• Filed: 02/27/1989
• Issued: 04/30/1991
• Est. Priority Date: 02/27/1989
• Status: Expired due to Fees

• Current Assignee: Motorola, Inc. (Motorola Solutions, Inc.)
• Original Assignee: Motorola, Inc. (Motorola Solutions, Inc.)
• Inventors: Le Beau, Christopher J.
• Primary Examiner(s): Moore, David K.
• Assistant Examiner(s): DINH, TAN X
• Application Number:
  US07/315,645
• Time in Patent Office: 792 Days
• Field of Search: 382/8, 382/54, 382/52, 382/22, 382/55, 364/513, 364/518
• US Class Current: 382/149
• CPC Class Codes: G06T 2207/30148 Semiconductor; IC; Wafer
• CPC Class Codes: G06T 7/73 using feature-based methods

Abstract

A method for automatically inspecting articles or work pieces uses a solid state television camera to obtain images of the article. An edge enhancement is done to obtain a first and a second threshold level which are dilated to eliminate noise edges. An edge correlation is then performed to determine orientation of the article. Gray scale or gray level imaging of the surface of the device is then performed and compared against a known device surface in order to determine acceptability of the device under inspection. The gray scale imaging techniques used are robust gray scale image processing that are tolerant to significant variations in image luminance or image degradation and is capable of identifying and classifying image features consistently while in the presence of such luminance variations. The data is then statistically characterized and the results used for real time statistical process control of the work piece process.

Claims

1. A method for automatically inspecting semiconductor die comprising:

obtaining an image of the semiconductor die;

performing a direction edge enhancement and selecting a first and a second threshold value of an edge enhanced image;

conditionally dilating the second threshold value with the first threshold value to conditionally grow prominent edge features; and

determining the semiconductor die location and rotation by correlating discrete die perimeter edge points, wherein the discrete die perimeter edge points are weighted, shifted and summed to perform the correlating.

2. The method of claim 1 further including dividing a top surface of the semiconductor die into a plurality of regions to cover the top surface;

determining an average gray level value for each of the plurality of regions;

against a predetermined average gray level for a known semiconductor die;

subtracting predetermined average gray level values of a known semiconductor die from the determined values to obtain difference values;

sorting the difference values from largest to smallest; and

normalizing the sorted difference values to determine acceptability of the semiconductor die.

3. The method of claim 1 wherein the correlating discrete die perimeter edge points is done by weighting the discrete edge points and shifting and summing discrete edge point values.

4. The method of claim 1 wherein rotation of the semiconductor die is done by shifting the discrete die perimeter edge points of two opposing edges toward center of the die by half of dimension between the opposing edges;

calculating a best fit line through the shifted points; and

determining rotation of the die from slope of the best fit line.

5. The method of claim 2 wherein sorting the difference values includes sorting the values into a matrix array to obtain calculated indices which are invariant to lighting luminance changes.

6. A method for automatically inspecting a semiconductor die, comprising:

determining average gray level value for each of a plurality of regions into which a surface of the semiconductor die has been divided;

subtracting the average gray level values against predetermined values for each of the plurality of regions to obtain difference values;

sorting the difference values from largest to smallest value to obtain normalized values;

calculating difference between a normalized center value and a smaller value to obtain a first index;

calculating difference between the normalized center value and a largest value to obtain a second index; and

determining acceptability of the semiconductor die from magnitude of the first and second indices.

7. The method of claim 6 further including:

performing a direction edge enhancement;

selecting a first and a second threshold of a resulting edge enhanced image to extract predominant edge features and wherein the second threshold is a higher value than the first threshold;

dilating the second threshold values with first threshold values to conditionally grow prominent edge features and avoiding growing low threshold noise edge points thereby defining meaningful semiconductor die edges.