Image processing method, three-dimensional position measuring method and image processing apparatus

• US 20060221417A1
• Filed: 03/13/2006
• Published: 10/05/2006
• Est. Priority Date: 03/15/2005
• Status: Active Grant

• Current Assignee: Omron Corporation
• Original Assignee: Omron Corporation
• Inventors: Ikeda, Yasuyuki, Fujieda, Shiro
• Granted Patent: US 7,742,634 B2
• Time in Patent Office: Days
• US Class Current: 358/538
• CPC Class Codes: G06T 2207/10004 Still image; Photographic image
• CPC Class Codes: G06T 2207/30108 Industrial image inspection
• CPC Class Codes: G06T 7/12 Edge-based segmentation
• CPC Class Codes: G06T 7/73 using feature-based methods
• CPC Class Codes: G06V 10/23 based on positionally close patterns or neighbourhood relationships
• CPC Class Codes: G06V 10/44 Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components

Abstract

An object of this invention is to provide an image processing method and image processing apparatus capable of extracting the position of an object in the unit of sub-pixel. An edge pixel, density gradient direction and edge position in the unit of sub-pixel are extracted from a model image obtained by photographing a high-quality model of the object and registered in a hard disk. After a processing object image is inputted, CPU obtains a position of an image region corresponding to the model image and after that, extracts an edge pixel whose edge position in the unit of sub-pixel is related to a corresponding pixel on the model image side in that image region. The edge position in the unit of sub-pixel is extracted for these edge pixels and the quantity of deviation between the extracting position and the edge position on the model image side is calculated. The position in the unit of pixel is corrected with an average value of the quantity of deviation obtained for each edge pixel so as to obtain a real position of the object.

Claims

1. A method for, with a gray image of a predetermined size as a processing object, extracting the position of an object corresponding to a preliminarily specified model from the processing object image, wherein

after a step A of extracting an edge pixel from an edge of a model image obtained by photographing the model;

a step B of retrieving a density gradient direction of each extracted edge pixel based on each edge pixel so as to extract an edge position in the unit of sub-pixel; and

a step C of registering the extraction position and the density gradient direction used for extraction of an edge pixel whose edge position is extracted in the step B with correspondence to the edge pixel are carried out preliminarily, first, second and third steps are carried out to the processing object image,

the first step calculating similarity between the image of a retrieval region and the model image of each scanning position by scanning the retrieval region of the same size as the model image in the unit of pixel as a processing object image,

the second step resetting up a retrieval region at a position extracted based on the similarity calculated in the first step, extracting an edge pixel corresponding to an edge pixel in a model image whose edge position in the sub-pixel unit is registered in the retrieval region and executing a processing of calculating the quantity of deviation between the edge position in the sub-pixel unit obtained in step B and the registered edge position for each extracted edge pixel,

the third step correcting a setting position of the retrieval region according to the quantity of deviation calculated in the second step and specifying a setting position after correction as the position of the object.

2. The image processing method according to claim 1 wherein the first step calculates similarity between a distribution pattern in the density gradient direction of each edge pixel of the model image and a distribution pattern in the density gradient direction of each edge pixel within the retrieval region at each scanning position.

3. The image processing method according to claim 1 wherein the second step resets up a retrieval region at a position in which the highest similarity is calculated in the first step.

4. A three-dimensional position measuring method for calculating an edge position of each object of an image generated with plural cameras each having overlapping vision field in accordance with the image processing method of claim 1 and measuring an object three-dimensionally by specifying a corresponding position based on the edge position of each obtained image.

5. An image processing apparatus comprising:

an image inputting device that inputs a density image;

an edge extracting device that extracts an edge pixel and the density gradient direction of the edge pixel from an edge in a density image inputted from the image inputting device;

an edge position extracting device that retrieves the density gradient direction with reference to an edge pixel on the density image based on an extraction result of the edge extracting device and extracts an edge position in the unit of sub-pixel;

a memory that stores a model image including a model of an object and data extracted from the model;

a registering device that creates data that when an image including the model image is inputted from the image inputting device, relates a processing result of the edge extracting device and the edge position extracting device to the inputted image to each edge pixel extracted by the edge extracting device by an edge extracting device and registers in the memory;

a similarity calculating device that when a processing object image including an object corresponding to the model is inputted from the image inputting device, scans a retrieval region of the same size as the model image in the processing object image in the unit of pixel so as to calculate a similarity between an image in the retrieval region and the model image for each scanning position;

a corresponding edge extracting device that resets up the retrieval region at a position extracted based on the similarity calculated by the similarity calculating device and extracts an edge pixel corresponding to an edge pixel in the model image whose edge position is registered in the memory based on a processing result of the edge extracting device in the retrieval region;

a deviation quantity calculating device that makes the edge position extracting device process an edge pixel extracted by the corresponding edge pixel extracting device and pixels in the neighborhood so as to calculate the quantity of deviation between an edge position in the unit of sub-pixel extracted from the processing and an edge position registered in the memory; and

a position specifying device that corrects the setting position of the retrieval region with the quantity of deviation calculated by the deviation quantity calculating device so as to specify the setting position after a correction as a position of the object.