CN106845481A - A kind of goods and materials shape recognition process based on binocular image vision - Google Patents

Abstract

The invention relates to recycling technology of waste bottle-shaped materials, in particular to a material shape recognition method based on binocular image vision. The invention solves the problems of limited application range, high implementation cost, high technical difficulty, low identification efficiency and low identification accuracy of the existing waste bottle-shaped materials identification technology. A material shape recognition method based on binocular image vision, which is realized by the following steps: Step S1: Simultaneously take top-down and head-up shooting of bottle-shaped materials placed upright; Step S2: Grayscale image A and grayscale image B for correction; step S3: perform median filtering on the corrected grayscale image A and grayscale image B respectively; step S4: identify the circular feature area and rectangular feature area of the bottle-shaped material ; Step S5: Identify the edges of the circular feature area and the edge of the rectangular feature area of the bottle-shaped material; Step S6: Reconstruct the image of the bottle-shaped material. The invention is suitable for recycling waste bottle-shaped materials.

Description

A Material Shape Recognition Method Based on Binocular Image Vision

technical field

The invention relates to recycling technology of waste bottle-shaped materials, in particular to a material shape recognition method based on binocular image vision.

Background technique

The identification of waste bottle-like materials is the primary link in the recycling process of waste bottle-like materials. Under the existing technical conditions, the following identification technologies are usually used to identify waste bottle materials: 1. Radio frequency identification: use the technology of scanning barcodes to identify the types of waste bottle materials. 2. Material identification: Use sensors to identify the material of waste bottle materials. 3. Ultrasonic ranging + weighing identification: Use ultrasonic ranging technology to roughly calculate the length and diameter of waste bottle-shaped materials, and use weighing technology to identify the weight of waste bottle-shaped materials. 4. Conventional image recognition: Use the background difference method to identify waste bottle-like materials. However, practice has shown that the above-mentioned waste bottle-shaped materials identification technology generally has the following problems due to its own principle limitations: 1. The radio frequency identification technology requires waste and used bottle-shaped materials to have a barcode. Once the barcode of the waste and used bottle-shaped materials falls off, it cannot Identification of waste bottle materials, resulting in a limited scope of application. Second, the material recognition technology has the problems of high implementation cost, high technical difficulty, and low recognition efficiency. 3. Ultrasonic ranging + weighing identification technology has the problem of low identification accuracy. 4. Conventional image recognition technology requires the background of the image to be uniform and unchanged, which leads to the problems of high technical difficulty and low recognition efficiency. Based on this, it is necessary to invent a brand-new waste bottle-shaped material identification technology to solve the above-mentioned problems existing in the waste and used bottle-shaped material identification technology.

Contents of the invention

In order to solve the problems of limited application range, high implementation cost, high technical difficulty, low recognition efficiency and low recognition accuracy of existing waste bottle-shaped material recognition technology, the present invention provides a material shape recognition method based on binocular image vision .

The present invention is realized by adopting the following technical solutions:

A material shape recognition method based on binocular image vision, which is realized by the following steps:

Step S1: Use an orthogonal binocular camera to simultaneously take top-down and head-up shots of the bottle-shaped materials placed upright, thereby obtaining grayscale image A and grayscale image B respectively; the grayscale image A obtained by top-down shooting The grayscale image B obtained from the head-up shooting contains both bottle-shaped materials and the background;

Step S2: using a perspective transformation algorithm to correct the grayscaled image A and the grayscaled image B respectively, thereby eliminating distortion from the grayscaled image A and the grayscaled image B respectively;

Step S3: performing median filtering on the corrected grayscaled image A and grayscaled image B respectively, thereby eliminating noise from grayscaled image A and grayscaled image B respectively;

Step S4: Using the maximum cross-correlation algorithm, on the one hand, identify the circular feature area of the bottle-shaped material from the grayscale image A. The circular feature area is composed of the bottle bottom and the bottle mouth; In B, the rectangular characteristic area of the bottle-shaped material is identified, and the rectangular characteristic area is composed of a bottleneck and a bottle body;

Step S5: Using the Sobel operator, on the one hand, identify the edge of the circular feature area of the bottle-shaped material from the grayscale image A, and on the other hand, identify the edge of the rectangular feature area of the bottle-shaped material from the grayscale image B edge;

Step S6: Calculate the size of the circular feature area of the bottle-shaped material from the grayscale image A, calculate the size of the rectangular feature area of the bottle-shaped material from the grayscale image B, and reconstruct the bottle-shaped material Image.

Compared with the existing waste bottle-shaped materials recognition technology, the material shape recognition method based on binocular image vision in the present invention realizes the recognition of waste and old bottle-shaped materials by adopting a new recognition principle, thus possessing The advantages are as follows: 1. Compared with the radio frequency identification technology, the present invention can identify waste bottle materials without barcodes, so its scope of application is no longer limited. 2. Compared with the material recognition technology, the present invention has lower implementation cost, less technical difficulty and higher recognition efficiency. 3. Compared with the ultrasonic ranging + weighing identification technology, the identification accuracy of the present invention is higher. 4. Compared with the conventional image recognition technology, the present invention no longer requires the background of the image to be uniform and unchanged, so its technical difficulty is smaller and the recognition efficiency is higher.

The invention effectively solves the problems of limited application range, high implementation cost, high technical difficulty, low identification efficiency and low identification accuracy of the existing waste bottle-shaped material identification technology, and is suitable for recycling waste bottle-shaped materials.

Description of drawings

Fig. 1 is a schematic flow chart of the present invention.

detailed description

A material shape recognition method based on binocular image vision, which is realized by the following steps:

Step S1: Use an orthogonal binocular camera to simultaneously take top-down and head-up shots of the bottle-shaped materials placed upright, thereby obtaining grayscale image A and grayscale image B respectively; the grayscale image A obtained by top-down shooting The grayscale image B obtained from the head-up shooting contains both bottle-shaped materials and the background;

Step S2: using a perspective transformation algorithm to correct the grayscaled image A and the grayscaled image B respectively, thereby eliminating distortion from the grayscaled image A and the grayscaled image B respectively;

Step S3: performing median filtering on the corrected grayscaled image A and grayscaled image B respectively, thereby eliminating noise from grayscaled image A and grayscaled image B respectively;

Step S4: Using the maximum cross-correlation algorithm, on the one hand, identify the circular feature area of the bottle-shaped material from the grayscale image A. The circular feature area is composed of the bottle bottom and the bottle mouth; In B, the rectangular characteristic area of the bottle-shaped material is identified, and the rectangular characteristic area is composed of a bottleneck and a bottle body;

Step S5: Using the Sobel operator, on the one hand, identify the edge of the circular feature area of the bottle-shaped material from the grayscale image A, and on the other hand, identify the edge of the rectangular feature area of the bottle-shaped material from the grayscale image B edge;

Step S6: Calculate the size of the circular feature area of the bottle-shaped material from the grayscale image A, calculate the size of the rectangular feature area of the bottle-shaped material from the grayscale image B, and reconstruct the bottle-shaped material Image.

In the step S5, the template used by the Sobel operator is .

Claims (2)

1. A material shape recognition method based on binocular image vision, characterized in that: the method is realized by following steps: Step S1: Use an orthogonal binocular camera to simultaneously take top-down and head-up shots of the bottle-shaped materials placed upright, thereby obtaining grayscale image A and grayscale image B respectively; the grayscale image A obtained by top-down shooting The grayscale image B obtained from the head-up shooting contains both bottle-shaped materials and the background; Step S2: using a perspective transformation algorithm to correct the grayscaled image A and the grayscaled image B respectively, thereby eliminating distortion from the grayscaled image A and the grayscaled image B respectively; Step S3: performing median filtering on the corrected grayscaled image A and grayscaled image B respectively, thereby eliminating noise from grayscaled image A and grayscaled image B respectively; Step S4: Using the maximum cross-correlation algorithm, on the one hand, identify the circular feature area of the bottle-shaped material from the grayscale image A. The circular feature area is composed of the bottle bottom and the bottle mouth; In B, the rectangular characteristic area of the bottle-shaped material is identified, and the rectangular characteristic area is composed of a bottleneck and a bottle body; Step S5: Using the Sobel operator, on the one hand, identify the edge of the circular feature area of the bottle-shaped material from the grayscale image A, and on the other hand, identify the edge of the rectangular feature area of the bottle-shaped material from the grayscale image B edge; Step S6: Calculate the size of the circular feature area of the bottle-shaped material from the grayscale image A, calculate the size of the rectangular feature area of the bottle-shaped material from the grayscale image B, and reconstruct the bottle-shaped material Image. 2. a kind of material shape recognition method based on binocular image vision according to claim 1, is characterized in that: in described step S5, the template that Sobel operator adopts is .