A vision algorithm that identifies inventory by the fork-entry side of a pallet

Abstract

The present invention relates to a visual algorithm for identifying inventory through the fork entry side of a pallet, comprising the following steps: various storage locations are displayed on a camera screen of an on-site warehouse, with the side of each storage location close to the camera being the side where a forklift enters the pallet; the image is rotated so that the bottom line of the pallet is horizontal, and area A and area B of each storage location are extracted, with the rectangular area at the forklift entry position being area A, and another rectangular area in front of the area being area B; for each area B, if there is an edge increase discovered by a CANNY algorithm, it is considered that a forklift pickup process has started; the kernel used by the CANNY algorithm is set to a long strip, with the pixel in the height direction being 1; for each area A, the edge increase discovered by the CANNY algorithm is extracted; if it is greater than a standard set for this storage location, it is considered that there is a pallet, that is, there is inventory, otherwise it is considered that there is no inventory; the present invention reduces the false alarm rate by eliminating the influence of edges in some directions, has a low missed alarm rate, and is less affected by ambient light.

Description

[Technical field]

[0001] The present invention relates to the technical field of inventory recognition, and in particular to a visual algorithm for identifying inventory through the fork entry side of a pallet. [Background Technology]

[0002] Currently, when identifying materials or boxes in warehouses, if the edge length of the pallet is used for identification without screening the direction of the pallet's edge, the influence of edges in some directions cannot be eliminated, resulting in a high false alarm rate, a high missed alarm rate, and a significant impact from ambient light.

[0003] If deep learning is used, the recognition stability will be poor. [Summary of the invention]

[0004] The purpose of the present invention is to solve the above-mentioned shortcomings and provide a visual algorithm for identifying inventory through the fork-entry side of the pallet. By eliminating the influence of edges in some directions, the false alarm rate can be reduced, the missed alarm rate is low, and the impact of ambient light is low.

[0005] To achieve the above objectives, a visual algorithm is designed to identify inventory by the fork-entry side of the pallet, including the following steps:

[0006] 1) Install a fixed, stationary network camera in the on-site warehouse. The camera screen displays each storage location. The side of each storage location closest to the camera is the side where forklifts enter and exit pallets, hereinafter referred to as the fork entry side.

[0007] 2) Rotate the video image so that the bottom line of the pallet fork entry side is horizontal, and extract area A and area B for each storage location. The rectangular area where the forklift enters each storage location is extracted as area A, and the other rectangular area in front of area A is area B. Area B is the area where the forklift arm passes. When the forklift arm enters the storage location, it passes through area B and area A in sequence.

[0008] 3) For each B region, if there is an increase in the edge discovered by the CANNY algorithm, it is considered that the forklift pickup process has started, and the system rests for a certain period of time to allow the forklift to complete the operation;

[0009] 4) Set the kernel used by the CANNY algorithm to a long strip with a pixel in the height direction of 1; for each A area, extract the increase in the number of pixels of the edge found by the CANNY algorithm; if it is greater than the standard set for this location, it is considered that there is a pallet, that is, there is inventory; otherwise, it is considered that there is no inventory.

[0010] Furthermore, N+1 results are obtained for each storage location and compared to see if they are the same as the previous N times. If they are the same, they are sent to the AGV forklift coordination system through communication to enable the AGV forklift to pick up the goods; if they are different, they are not sent.

[0011] Furthermore, the previous N times are defined as at least 3 times.

[0012] Furthermore, in step 3), the system rests for a certain period of time, preferably for example 20 seconds.

[0013] Compared with the existing technology, the present invention installs a network camera on the on-site warehouse, so that each storage location can be shown on the camera screen. The side of each storage location close to the camera is the side where the forklift forks the pallet, and visual processing is performed on the video frame. Test standards are set for the judgment involved in the visual processing step, and finally the inventory is recognized. By eliminating the influence of edges in some directions, the false alarm rate can be reduced, the missed alarm rate is low, and the influence of ambient light is low, which is worthy of promotion and application. [Brief Description of the Drawings]

[0014] Figure 1 It is a schematic diagram of the line side inventory after rotation of the present invention;

[0015] Figure 2 It is a schematic diagram of area B of a storage location of the present invention;

[0016] Figure 3 This is an example diagram of a warehouse scene next to the production line (line-side warehouse) of the present invention. [Specific implementation method]

[0017] The present invention will be further described below with reference to the accompanying drawings and specific embodiments:

[0018] The present invention provides a visual algorithm for identifying inventory through the fork-entry side of a pallet, comprising the following steps: 1. Installing a fixed, stationary webcam in an on-site warehouse.

[0019] a) There are various storage locations on the camera screen;

[0020] b) The side of each storage location closest to the camera is the side where forklifts enter and exit pallets (hereinafter referred to as the fork entry side).

[0021] 2. Set test criteria for the judgments involved in the following steps.

[0022] 3. Visual processing of video frames. Step 1: Extract area A and area B of each storage location

[0023] a) Rotate the video image so that the bottom line of the pallet fork entry side is horizontal;

[0024] b) Extract the rectangular area where the forklift enters each storage location as area A, and another rectangular area in front of it as area B. Area B is the area where the forklift arm passes. When the forklift arm enters the storage location, it passes through area B and area A in sequence.

[0025] 4. Step 2:

[0026] a) For each B region, if there is an increase in the edge discovered by the CANNY algorithm, it is considered that the forklift pickup process has started, and the system rests for a certain period of time (for example, 20 seconds) to allow the forklift to complete the operation.

[0027] 5. Step 3:

[0028] a) Set the kernel used by the CANNY algorithm to be a long strip, and its height direction pixel is 1;

[0029] b) for each A region, extract the increase in the number of pixels of the edge found by the CANNY algorithm;

[0030] c) If it is greater than the standard set for this storage location, it is considered that there is a pallet (inventory); otherwise, it is considered that there is no pallet.

[0031] 6. Step 4:

[0032] a) Perform statistical determination on the results of each storage location. For example, obtain N+1 results to see if they are the same as the previous N times (preferably at least 3 times). If they are the same, send the result to the AGV forklift coordination system via communication, and let the AGV forklift pick up the goods;

[0033] b) If different, do not send.

[0034] As attached Figure 1 As shown in the figure, it is a picture obtained by installing a network camera in the on-site warehouse, and the line inventory map after rotation; as shown in the attached Figure 2 As shown in the figure, it is a schematic diagram of area B of a storage location; Figure 3 The figure below illustrates an example scenario for a warehouse near a production line (lineside warehouse). By combining the above steps, inventory can be identified by the fork-entry side of the pallet. Therefore, by eliminating the influence of edges in certain directions, the present invention can reduce the false alarm rate, achieve a low missed alarm rate, and be less affected by ambient lighting.

[0035] The contents not described in detail in this specification belong to the prior art known to professional and technical personnel in this field and will not be described in detail here.

[0036] The present invention is not limited to the above-mentioned embodiments. Any other changes, modifications, substitutions, combinations, and simplifications that do not deviate from the spirit and principles of the present invention should be considered as equivalent replacement methods and are included in the scope of protection of the present invention.

Claims

1. A vision algorithm for identifying inventory by the fork-entry side of a pallet, characterized by: The following steps are involved: 1) Install a fixed, stationary network camera in the on-site warehouse. The camera screen displays each storage location. The side of each storage location closest to the camera is the side where forklifts enter and exit pallets, hereinafter referred to as the fork entry side. 2) Rotate the video image so that the bottom line of the pallet fork entry side is horizontal, and extract area A and area B for each storage location. The rectangular area where the forklift enters each storage location is extracted as area A, and the other rectangular area in front of area A is area B. Area B is the area where the forklift arm passes. When the forklift arm enters the storage location, it passes through area B and area A in sequence. 3) For each B region, if there is an increase in the edge discovered by the CANNY algorithm, it is considered that the forklift pickup process has started, and the system rests for a certain period of time to allow the forklift to complete the operation; 4) Set the kernel used by the CANNY algorithm to be a long strip, with pixels in the height direction equal to 1. For each region A, extract the number of pixels along the edge found by the CANNY algorithm. If the number is greater than the standard set for this location, it is considered that there is a pallet, that is, inventory; otherwise, it is considered that there is no inventory. Obtain N+1 results for each storage location and compare them to see if they are the same as the previous N times. If they are the same, they are sent to the AGV forklift coordination system through communication to enable the AGV forklift to pick up the goods; if they are different, they are not sent.

2. The visual algorithm for identifying inventory by the fork-entry side of a pallet as claimed in claim 1, characterized in that: The previous N times are defined as at least 3 times.

Images