CN215495063U

CN215495063U - Seed cotton hair sorting device based on FPGA and deep learning

Info

Publication number: CN215495063U
Application number: CN202023201583.4U
Authority: CN
Inventors: 李元哲; 倪超; 朱婷婷
Original assignee: Nanjing Forestry University
Current assignee: Nanjing Forestry University
Priority date: 2020-12-28
Filing date: 2020-12-28
Publication date: 2022-01-11
Anticipated expiration: 2030-12-28

Abstract

The utility model discloses a seed cotton hair sorting device based on FPGA and deep learning, which comprises a feeding box, a linear array camera, a camera bracket, a conveyor belt, an FPGA board card, a valve circuit board and a fan, wherein the feeding box is arranged on the front end of the feeding box; the feeding box is located above the left end of the conveying belt, the linear array camera is placed above the right section of the conveying belt, the linear array camera is fixed through a camera support and perpendicular to the conveying belt, the linear array camera is connected with the FPGA board card, the FPGA is connected with the valve circuit board, and the valve circuit board is connected with the fan. According to the method, the hair impurities in the seed cotton are correctly identified through a deep learning algorithm, and meanwhile, the deep learning process is accelerated by utilizing the strong computing power of the FPGA, so that the seed cotton containing the hair is accurately and quickly sorted.

Description

Seed cotton hair sorting device based on FPGA and deep learning

Technical Field

The utility model relates to the field of seed cotton hair identification and sorting, in particular to a seed cotton hair sorting device based on FPGA and deep learning.

Background

Cotton is one of the most important crops in China, and the yield and consumption of cotton in China always stay the top of the world. During the cotton picking and transporting process, the hair is inevitably mixed, and the problems of color difference, breakage and the like of textiles are easily caused in the subsequent processing process. However, the hair strands remaining therein are very close to the cotton fibers and cannot be effectively identified by means of conventional machine learning algorithms. Accordingly, there is a need for a more efficient sorting apparatus for identifying and sorting hair impurities in seed cotton.

SUMMERY OF THE UTILITY MODEL

The purpose of the utility model is as follows: aiming at the defects in the prior art, the utility model provides the seed cotton hair sorting device based on the FPGA and the deep learning, which can quickly and accurately identify the seed cotton containing hair and accurately eject the seed cotton to obtain the clean seed cotton without hair.

The technical scheme is as follows: in order to achieve the purpose of the utility model, the utility model adopts the technical scheme that:

a seed cotton hair sorting device based on FPGA and deep learning comprises a feeding box, a linear array camera, a camera support, a conveyor belt, an FPGA board card, a valve circuit board and a fan; the feeding box is located above the left end of the conveying belt, the linear array camera is placed above the right section of the conveying belt, the linear array camera is fixed through a camera support, the linear array camera is kept perpendicular to the conveying belt, the linear array camera is connected with the FPGA board card, the FPGA is connected with the valve circuit board, and the valve circuit board is connected with the fan.

Preferably, the feeding box is provided with a cotton opening device, so that large-cluster seed cotton can be loosened into single seed cotton, the single seed cotton is uniformly spread on the conveyer belt, hair impurities in the seed cotton are exposed, and the impurity images can be conveniently collected.

Preferably, the linear array camera adopts a linear array CCD camera to acquire RAW format images of the seed cotton.

Preferably, the camera support is a rigid support for fixing the line camera perpendicular to the conveyor belt.

Preferably, the conveyer belt is black rubber conveyer belt, and difficult reflection of light guarantees that the image of shooing can not receive the interference of background light.

Preferably, the configuration parameters of each convolution kernel of the deep learning model are embedded in the FPGA board card, and the convolution kernels and the board card have a corresponding relation, so that data calculation can be accelerated.

Preferably, the valve circuit board is a 24-way valve circuit, and the fan is driven to work by generating a control valve signal.

Preferably, the fan is driven by the valve circuit board, the working pressure is 5 kilograms, the fan is started and stopped 50 times per second, and the seed cotton containing the hair can be rapidly and accurately ejected.

Has the advantages that: compared with the prior art, the utility model has the following advantages:

1. the method has the advantages that the deep learning algorithm is utilized to identify the hair similar to the cotton fiber, so that the problem that the hair in the cotton is difficult to identify is solved;

2. the convolution kernel parameters of the convolution neural network are embedded into the FPGA board card, so that the calculation process is accelerated while hair is effectively identified, and the part containing the hair can be removed in time in the process of cotton flow high-speed movement.

Drawings

FIG. 1 is a schematic view of the overall structure of the present invention;

fig. 2 is a flow chart of the present invention.

Detailed Description

The present invention will be described in further detail with reference to the accompanying drawings.

As shown in figure 1, the seed cotton hair sorting device based on FPGA and deep learning mainly structurally comprises a feeding box 1, a linear array camera 2, a camera support 3, a conveyor belt 4, an FPGA board card 6, a valve circuit board 7, a fan 8 and the like. The feeding box 1 is installed above the left end of the conveyor belt 4, the movement direction of the conveyor belt 4 is clockwise, the linear array camera 2 is located above the right section of the conveyor belt 4 and is kept perpendicular to the conveyor belt 4 through the camera support 3, the linear array camera 2 is connected with the FPGA board card 6, the FPGA board card 6 is connected with the valve circuit board 7, and the fan 8 is connected with the valve circuit board 7 and driven by the valve circuit board 7.

A mechanical opening device is arranged in the charging box 1, the agglomerated seed cotton can be combed and opened to be changed into single seed cotton, the single seed cotton falls onto the conveyor belt 4, and the falling seed cotton moves to the right along with the conveyor belt. The conveying belt 4 is a black rubber conveying belt, light reflection is not easy, the shot images are not interfered by background light, RAW format images of seed cotton are collected by the linear array camera 2, and the images are transmitted to the FPGA board card 6 to be processed. The FPGA board card 6 identifies whether the image acquired by the linear array camera 2 is a seed cotton image containing hair, the identified result is transmitted to the valve circuit board 7, the valve circuit board 7 controls the on-off state of the fan 8 through a relay, after a delay time, the fan 8 sprays seed cotton containing hair identified by the FPGA board card 6, and other seed cotton falls to a designated position due to inertia to realize the sorting of the seed cotton hair.

As shown in fig. 2, in the seed cotton hair sorting algorithm flowchart based on FPGA and deep learning of the present application, the convolutional neural network has excellent performance in terms of processing images in deep learning, so that seed cotton images containing hair are input into the convolutional neural network in advance for training until the obtained training model can correctly identify the seed cotton images containing hair, and the obtained model parameters are configured in the FPGA board 6. The FPGA detection module identifies the seed cotton image collected in real time, sends an instruction to the valve circuit board for the seed cotton containing hair, and controls the fan to spray out the corresponding seed cotton; and if no hair is detected, continuously detecting the image at the next moment, thereby separating the clean seed cotton from the seed cotton containing impurities.

The foregoing is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, various modifications and decorations can be made without departing from the principle of the present invention, and these modifications and decorations should also be regarded as the protection scope of the present invention.

Claims

1. The utility model provides a seed cotton hair sorting unit based on FPGA and degree of depth study which characterized in that: the automatic feeding device comprises a feeding box (1), a linear array camera (2), a camera support (3), a conveyor belt (4), an FPGA board card (6), a valve circuit board (7) and a fan (8); the feeding box (1) is located above the left end of the conveying belt (4), the linear array camera (2) is placed above the right section of the conveying belt (4), the linear array camera (2) is fixed through the camera support (3), the linear array camera (2) is perpendicular to the conveying belt (4), the linear array camera (2) is connected with the FPGA board card (6), the FPGA board card (6) is connected with the valve circuit board (7), and the valve circuit board (7) is connected with the fan (8).

2. The seed cotton hair sorting device based on the FPGA and the deep learning of claim 1, characterized in that: the cotton opening device is arranged in the feeding box (1).

3. The seed cotton hair sorting device based on the FPGA and the deep learning of claim 1, characterized in that: the linear array camera (2) adopts a linear array CCD camera.

4. The seed cotton hair sorting device based on the FPGA and the deep learning of claim 1, characterized in that: the camera support (3) is a rigid support.

5. The seed cotton hair sorting device based on the FPGA and the deep learning of claim 1, characterized in that: the conveying belt (4) is a black rubber conveying belt.

6. The seed cotton hair sorting device based on the FPGA and the deep learning of claim 1, characterized in that: the configuration parameters of each convolution kernel of the deep learning model are embedded in the FPGA board card (6).

7. The seed cotton hair sorting device based on the FPGA and the deep learning of claim 1, characterized in that: the valve circuit board (7) is a 24-way valve circuit.

8. The seed cotton hair sorting device based on the FPGA and the deep learning of claim 1, characterized in that: the fan (8) is driven by a valve circuit board, the working pressure is 5 kilograms, and the fan is started and stopped for 50 times per second.