Multi-tray automatic goods taking method compatible with field tray truck

Abstract

The invention relates to the technical field of tray goods taking, in particular to a multi-tray automatic goods taking method compatible with a field tray truck, which comprises the following steps: a vehicle arrives at a photographing point for photographing, tray types are identified, forks are adjusted and goods are taken, the vehicle arriving at the photographing point for photographing comprises a power compartment, and a camera is fixedly mounted above the power compartment; a front wheel is fixedly installed below the front end of the power compartment, a middle wheel is fixedly installed on the side, away from the front wheel, below the power compartment, a fork arm is installed on the side, close to the middle wheel, of the power compartment, and a rear wheel is fixedly installed below the end, away from the power compartment, of the fork arm. According to the method, different types of carriers are recognized according to a posture recognition camera or radar on the vehicle, the fork entering depth is adjusted, the pallet truck compatible with the field support can insert various types of pallets, a more robust pallet classification algorithm is adopted, it is ensured that recognition is more stable, and the problem of an existing matching algorithm in a pallet recognition scene is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of pallet picking, and specifically to a method for automatically picking multiple pallets compatible with a Tian pallet truck. Background Technique

[0002] With the rapid development of the logistics industry, AGVs are increasingly widely used in the industry. There are many types of pallets on the market. For Tian-character pallets, currently on the market, counterbalanced or wide-leg forklifts are mainly used for insertion and picking.

[0003] During the process of the vehicle picking multiple different specifications of Tian-character pallets, if the pallet type is not known, it is impossible to judge the fork insertion depth for picking, which may cause damage to the pallet during the picking process. Summary of the Invention

[0004] The purpose of the present invention is to provide a method for automatically picking multiple pallets compatible with a Tian pallet truck to solve the problem of the inability to automatically pick multiple carriers of the pallet truck mentioned in the above background technique.

[0005] To achieve the above purpose, the present invention provides the following technical solution: A method for automatically picking multiple pallets compatible with a Tian pallet truck, including:

[0006] The vehicle arrives at the photographing point to take a photo, identify the pallet type, adjust the fork insertion and pick up the goods

[0007] The vehicle in the vehicle arriving at the photographing point to take a photo includes a power compartment. A camera is fixedly installed above the power compartment. A front wheel is fixedly installed below the front end of the power compartment. An intermediate wheel is fixedly installed on one side of the power compartment below and away from the front wheel. A fork arm is installed on one side of the power compartment close to the intermediate wheel. A rear wheel is fixedly installed below the end of the fork arm away from the power compartment;

[0008] The identifying the pallet type includes pallet front surface point cloud segmentation and pallet type identification;

[0009] The pallet type identification includes establishing a model library, dimensionality reduction of the vehicle plane point cloud, gradient calculation, and similarity calculation.

[0010] Preferably, the shooting direction of the camera is on the side close to the fork arm.

[0011] Preferably, both the intermediate wheel and the rear wheel can be retracted.

[0012] Preferably, the pallet front surface point cloud segmentation uses the camera to take a photo to obtain an image, identifies the pallet according to the object detection algorithm yolov5, extracts the point cloud information on the pallet, and extracts the plane of the vehicle according to the point cloud information using the ransac algorithm.

[0013] Preferably, the method for establishing the model library is asFigure 1 。

[0014] Preferably, after reducing the dimensionality of the vehicle's planar point cloud, the depth information is ignored, and the extracted plane is as Figure 2 , and according to the extracted pallet plane, the angle angle of the plane is calculated. Through inverse transformation, the pallet point cloud with an angular offset can be converted into a pallet without an angle, as Figure 3 shown.

[0015] Preferably, the gradient is calculated using the Sobel operator, which can accurately extract the edge information of the pallet, and each model can obtain the best matching score when calculating the similarity.

[0016] Compared with the prior art, the beneficial effects of the present invention are:

[0017] 1. Identify different types of vehicles based on the attitude recognition camera or radar on the vehicle and adjust the fork insertion depth.

[0018] 2. The pallet truck compatible with the Tian pallet can insert and pick up various types of pallets.

[0019] 3. Adopt a more robust pallet classification algorithm to ensure more stable recognition and solve the problems of existing matching algorithms in the pallet recognition scenario. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 is a schematic diagram of the model library ratio of the present invention;

[0021] Figure 2 is a schematic diagram of the dimensionality reduction plane of the present invention;

[0022] Figure 3 is a schematic top view of the pallet correction of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0023] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0024] Please refer to Figures 1 - 3 , an embodiment provided by the present invention: A method for automatically picking up various pallets by a pallet truck compatible with the Tian pallet.

[0025] Including: the vehicle arrives at the photographing point to take a photo, identify the pallet type, adjust the fork insertion and pick up the goods

[0026] The vehicle arriving at the photographing point. The vehicle in the photographing process includes a power compartment. A camera is fixedly installed above the power compartment. A front wheel is fixedly installed below the front end of the power compartment. An intermediate wheel is fixedly installed on one side of the power compartment below and away from the front wheel. A fork arm is installed on one side of the power compartment close to the intermediate wheel. A rear wheel is fixedly installed below one end of the fork arm away from the power compartment.

[0027] The identification of the tray type includes the point cloud segmentation of the front surface of the tray and the identification of the tray type. The point cloud segmentation of the front surface of the tray is to take a photo to obtain an image, identify the tray according to the target detection algorithm yolov5, then extract the point cloud information on the tray, and use the ransac algorithm to extract the plane of the vehicle according to the point cloud information.

[0028] The identification of the tray type includes establishing a model library, dimensionality reduction of the vehicle plane point cloud, gradient calculation, and calculation of similarity. Gradient calculation is to convert the xy information of the point cloud obtained by dimensionality reduction into grayscale image data img1, and then the gradient of the image can be calculated. Calculating similarity is to calculate the similarity according to the grayscale image img1 and the edge image img2. By traversing the grayscale image and the edge image at the same time, two scores score1 and score2 can be obtained, and finally the final similarity score is calculated.

[0029] Further, the shooting direction of the camera is the side close to the fork arm. When the fork arm needs to be inserted under the field tray and moves towards the field tray, the camera can directly take a photo of the field tray.

[0030] Further, both the intermediate wheel and the rear wheel can be retracted. The intermediate wheel and the rear wheel are used alternately to complete the picking up of goods.

[0031] Further, the point cloud segmentation of the front surface of the tray uses the camera to take a photo to obtain an image, identify the tray according to the target detection algorithm yolov5, extract the point cloud information on the tray, and use the ransac algorithm to extract the plane of the vehicle according to the point cloud information. yolov5 is a popular computer vision algorithm mainly used for object detection. It is the fifth version of the YOLO series of algorithms and inherits the main features of the YOLO series of algorithms: fast speed and high accuracy.

[0032] Further, the method of establishing the model library is as Figure 1 , and the established model library can facilitate the subsequent identification of a certain type of tray and enable it to make a correct response.

[0033] Further, after dimensionality reduction of the vehicle plane point cloud, the depth information is ignored, and the extracted plane is as Figure 2 , and the angle angle of the plane is calculated according to the extracted tray plane. Through inverse transformation, the tray point cloud with an angular offset can be converted into a tray without an angle, as Figure 3 shown.

[0034] Furthermore, the gradient is calculated using the Sobel operator, which can accurately extract the edge information of the tray. Through the above method, the edge information img2 of the image can be extracted. When calculating the similarity, each model can obtain the best matching score, and the model corresponding to the best score is the type of the target tray.

[0035] Working principle: When in use, first establish a model library for the existing trays. Move the vehicle to the photographing point, take a photo to obtain an image, identify the tray according to the target detection algorithm yolov5, then extract the point cloud information on the tray, and use the ransac algorithm to extract the plane of the vehicle based on the point cloud information. For the vehicle plane point cloud extracted in the previous stage, first perform dimensionality reduction processing, ignore the depth information, and extract the plane. Calculate the angle angle of the plane according to the extracted tray plane. Finally, through inverse transformation, the tray point cloud with an angular offset can be converted into a tray without an angle. For the xy information of the point cloud obtained through dimensionality reduction, it is converted into grayscale image data img1, and then the gradient of the image can be calculated. The Sobel operator is used to calculate the gradient, which can accurately extract the edge information of the tray. Through the above method, the edge information img2 of the image can be extracted. Calculate the similarity according to the grayscale image img1 and the edge image img2. Suppose the number of overlapping points between the template and the target tray is n1, and the number of valid points of the template tray is n2. Then the current total score traversed is score = n1 / n2 (the score is 1 for a perfect match and 0 for a complete mismatch). At the same time, traverse the grayscale image and the edge image to obtain two scores score1 and score2. Finally, calculate the final similarity score through total_score = 1 / x*(score1)+(x - 1) / x*(score2) (the x coefficient is the weighting coefficient, which is adjusted according to the type of tray in the actual scenario and is generally defaulted to 2). When there are multiple types of trays on-site, the model library established by the user traverses the target tray in turn, and each model can obtain the best matching score. Then the model corresponding to the best score is the type of the target tray. When the vehicle needs to pick up a field tray, tighten the rear wheels and open the middle wheels to pick up the goods. When the goods are picked up, tighten the middle wheels and open the rear wheels to switch to the standard state of the pallet truck. The above is the entire working principle of the present invention.

[0036] It is obvious to those skilled in the art that the present invention is not limited to the details of the above-described exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention. Therefore, in any respect, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Accordingly, all changes that fall within the meaning and scope of the equivalent elements of the claims are intended to be embraced within the present invention. Any reference signs in the claims should not be construed as limiting the claims involved.

Claims

1. A method for automatically picking up multiple pallets compatible with a pallet truck, characterized in that: include: The vehicle arrives at the photo point to take photos, identify the pallet type, adjust the fork and pick up the goods The vehicle arrives at the photo-taking point. The vehicle being photographed includes a power compartment, a camera is fixedly mounted above the power compartment, a front wheel is fixedly mounted below the front end of the power compartment, an intermediate wheel is fixedly mounted below the power compartment on a side away from the front wheel, a fork arm is mounted on a side of the power compartment close to the intermediate wheel, and a rear wheel is fixedly mounted below one end of the fork arm away from the power compartment; The identifying of the pallet type includes pallet front surface point cloud segmentation and pallet type identification; The pallet type identification includes establishing a model library, reducing the dimension of the carrier plane point cloud, gradient calculation and similarity calculation.

2. The method for automatically picking up multiple pallets compatible with a pallet truck according to claim 1, characterized in that: The shooting direction of the camera is the side close to the fork arm.

3. The method for automatically picking up multiple pallets compatible with a pallet truck according to claim 1, characterized in that: The middle wheels and the rear wheels can be retracted.

4. The method for automatically picking up multiple pallets compatible with a pallet truck according to claim 1, characterized in that: The point cloud segmentation of the front surface of the pallet uses a camera to obtain an image, identifies the pallet according to the target detection algorithm YOLO V5, extracts the point cloud information on the pallet, and extracts the plane of the carrier according to the point cloud information using the RANSAC algorithm.

5. The method for automatically picking up multiple pallets compatible with a pallet truck according to claim 2, characterized in that: The method for establishing the model library is shown in FIG1 .

6. The method for automatically picking up multiple pallets compatible with a pallet truck according to claim 2, characterized in that: After reducing the dimension of the carrier plane point cloud, the depth information is ignored, and the extracted plane is shown in Figure 2. The angle of the plane is calculated based on the extracted pallet plane. The original pallet point cloud with angle offset can be converted into a pallet without angle through inverse transformation, as shown in Figure 3.

7. The method for automatically picking up multiple pallets compatible with a pallet truck according to claim 2, characterized in that: The gradient calculation uses the Sobel operator to calculate the gradient, which can accurately extract the edge information of the tray. When calculating the similarity, each model can obtain the best matching score.

Images