Textile fuzzing and pilling level detection method, device and equipment and readable medium

By filtering and segmenting the three-dimensional point cloud data of textiles, the textile wool pill detection model is trained, which solves the detection accuracy problems and waste of computing resources under the influence of light and perspective changes, and achieves more stable and efficient detection results .

CN118941523AActive Publication Date: 2024-11-12HANGZHOU INNOVATION RES INST OF BEIJING UNIV OF AERONAUTICS & ASTRONAUTICS
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202410981616.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-22
Publication Date
2024-11-12
Estimated Expiration
2044-07-22

AI Technical Summary

Technical Problem

The existing textile wool pill detection methods have poor accuracy under light and perspective changes, and the calculation complexity is high during training of initial item identification networks, resulting in waste of resources.

Method used

Point cloud filtering is performed by obtaining the initial point cloud training sample set, the target point cloud training sample set is generated, the initial item recognition network is trained, and the textile point cloud training sample set is obtained, the textile wool pill detection model is trained, and the three-dimensional point cloud is collected. The data is segmented to generate local lint-pulling level information.

Benefits of technology

It improves the accuracy and stability of textile wool pill detection, reduces the waste of computing resources, and enhances the model's processing ability and adaptability to point cloud data.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118941523A_ABST
    Figure CN118941523A_ABST
Patent Text Reader

Abstract

The embodiment of the invention discloses a textile fuzzing and pilling level detection method, device and equipment and a readable medium. A specific embodiment of the method comprises the steps of obtaining an initial point cloud training sample set; performing point cloud filtering processing on the initial point cloud training sample set; performing first training on the initial article identification network; determining the trained initial article identification model as a to-be-trained textile fuzzing and pilling level detection model; obtaining a textile point cloud training sample set; carrying out second training on the to-be-trained textile fuzzing and pilling level detection model; determining the trained to-be-trained textile fuzzing and pilling level detection model as a textile fuzzing and pilling level detection model; collecting point cloud data of a to-be-detected textile; segmenting the to-be-detected textile point cloud data; inputting the point cloud block information into a textile fuzzing and pilling level detection model; and generating textile fuzzing and pilling level detection information. According to the embodiment, the accuracy of the fluffing and pilling level detection information of the textile is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] Embodiments of the present disclosure relate to the field of computer technology, and in particular to a method, device, equipment, and readable medium for detecting the pilling level of textiles. Background Art

[0002] Pilling level detection of textiles is of great significance to textile production. Pilling level detection of textiles is a technology for detecting the pilling degree of textiles. At present, when testing the pilling degree of textiles, the method usually adopted is: through the two-dimensional image of the textile, the pilling level of the textile is detected.

[0003] However, when using the above method to test the pilling level of textiles, the following technical problems often occur:

[0004] First, the pilling level of textiles is tested through two-dimensional images of textiles. Two-dimensional images are greatly affected by environmental changes and changes in viewing angles during imaging. For example, in dim light conditions, pilling may appear more difficult to distinguish, resulting in poor accuracy of the pilling level test information for textiles. At the same time, changes in viewing angles may cause deformation of the textile in the image or changes in the surface texture, which may make the appearance of pilling different at different angles, and may cause unstable pilling level test results for textiles.

[0005] Continuing, in the process of adopting technical solutions to solve the above-mentioned technical problem 1, the following technical problems often occur: Before the pilling level detection of textiles is performed, the initial object recognition network needs to be trained. The solution for training the initial object recognition network is generally to directly use the unprocessed point cloud data including the corresponding recognition object to train the initial object recognition network. However, the above-mentioned conventional solution directly uses the unprocessed point cloud data including the corresponding recognition object to train the initial object recognition network, and there are still the following problems:

[0006] Second, the initial object recognition network is trained directly using the unprocessed point cloud data of the corresponding identified objects. The point cloud data of the identified objects usually includes a large amount of data point information, which increases the computational complexity and computing resource requirements of the model training, resulting in a waste of computer computing resources.

[0007] The above information disclosed in this Background section is only for enhancement of understanding of the background of the inventive concept and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the invention

[0008] The content of this disclosure is used to introduce concepts in a brief form, which will be described in detail in the detailed implementation section below. The content of this disclosure is not intended to identify the key features or essential features of the technical solution claimed for protection, nor is it intended to limit the scope of the technical solution claimed for protection.

[0009] Some embodiments of the present disclosure propose a method, device, electronic device and computer-readable medium for detecting the pilling level of textiles to solve one or more of the technical problems mentioned in the above background technology section.

[0010] In a first aspect, some embodiments of the present disclosure provide a method for detecting the pilling level of textiles, the method comprising: obtaining an initial point cloud training sample set; performing point cloud filtering processing on the initial point cloud training sample set to obtain a target point cloud training sample set; performing a first training on an initial object recognition network according to the target point cloud training sample set to obtain a trained initial object recognition model; determining the trained initial object recognition model as a model for detecting the pilling level of textiles to be trained; obtaining a textile point cloud training sample set; performing a second training on the model for detecting the pilling level of textiles to be trained according to the textile point cloud training sample set to obtain a trained model for detecting the pilling level of textiles to be trained; determining the trained model for detecting the pilling level of textiles to be trained as a model for detecting the pilling level of textiles to be trained Pilling level detection model; collecting three-dimensional point cloud data of the textile to be detected as the point cloud data of the textile to be detected, wherein the point cloud data of the textile to be detected includes information of each data point, and each data point information in the above data point information includes coordinate position information and color information; segmenting the point cloud data of the textile to be detected to obtain a point cloud block information set, wherein each point cloud block information in the point cloud block information set includes information of each data point; for each point cloud block information in the point cloud block information set, inputting the point cloud block information into the textile pilling level detection model to obtain the local pilling level information of the textile corresponding to the point cloud block information; based on the obtained local pilling level information of each textile, generating the textile pilling level detection information corresponding to the textile to be detected.

[0011] In a second aspect, some embodiments of the present disclosure provide a device for detecting the pilling level of a textile, the device comprising: a first acquisition unit, configured to acquire an initial point cloud training sample set; a point cloud filtering processing unit, configured to perform point cloud filtering processing on the initial point cloud training sample set to obtain a target point cloud training sample set; a first training unit, configured to perform a first training on an initial object recognition network according to the target point cloud training sample set to obtain a trained initial object recognition model; a first determination unit, configured to determine the trained initial object recognition model as a textile pilling level detection model to be trained; a second acquisition unit, configured to acquire a textile point cloud training sample set; a second training unit, configured to perform a second training on the textile pilling level detection model to be trained according to the textile point cloud training sample set to obtain a trained textile pilling level detection model to be trained; and a second determination unit, configured to determine the trained textile pilling level detection model to be trained as a textile pilling level detection model to be trained. The product pilling level detection model is determined as the textile pilling level detection model; the acquisition unit is configured to acquire three-dimensional point cloud data of the textile to be detected as the textile point cloud data to be detected, wherein the above-mentioned textile point cloud data to be detected includes information of each data point, and each data point information in the above-mentioned data point information includes coordinate position information and color information; the segmentation processing unit is configured to segment the above-mentioned textile point cloud data to be detected to obtain a point cloud block information set, wherein each point cloud block information in the above-mentioned point cloud block information set includes information of each data point; the input unit is configured to input the above-mentioned point cloud block information into the above-mentioned textile pilling level detection model for each point cloud block information in the above-mentioned point cloud block information set, and obtain the textile local pilling level information corresponding to the above-mentioned point cloud block information; the generation unit is configured to generate the textile pilling level detection information corresponding to the above-mentioned textile to be detected based on the obtained local pilling level information of each textile.

[0012] In a third aspect, some embodiments of the present disclosure provide an electronic device comprising: one or more processors; a storage device on which one or more programs are stored, and when the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any implementation manner of the above-mentioned first aspect.

[0013] In a fourth aspect, some embodiments of the present disclosure provide a computer-readable medium having a computer program stored thereon, wherein when the program is executed by a processor, the method described in any implementation manner of the above-mentioned first aspect is implemented.

[0014] The above-mentioned embodiments of the present disclosure have the following beneficial effects: through the textile pilling level detection method of some embodiments of the present disclosure, the accuracy of the textile pilling level detection information and the stability of the detection results are improved. Specifically, the reason for the accuracy of the textile pilling level detection information and the stability of the detection results is that the textile pilling level detection is performed on the textile through the two-dimensional image of the textile, and the two-dimensional image is greatly affected by environmental changes and changes in viewing angles during imaging. For example, in the case of darker light, pilling may appear more difficult to distinguish, resulting in poor accuracy of the textile pilling level detection information. At the same time, changes in viewing angle may cause deformation of the textile in the image or changes in the surface texture, which may make the appearance of pilling different at different angles, and may cause the textile pilling level detection result to be unstable. Based on this, the textile pilling level detection method of some embodiments of the present disclosure, first, obtains an initial point cloud training sample set. Thus, an initial point cloud training sample set for generating a target point cloud training sample set can be obtained. Then, the initial point cloud training sample set is subjected to point cloud filtering processing to obtain a target point cloud training sample set. Thus, a target point cloud training sample set can be obtained. Afterwards, the initial object recognition network is first trained according to the target point cloud training sample set to obtain the trained initial object recognition model. Thus, the initial object recognition network can be trained by the enhanced model through the target point cloud training sample set, thereby improving the processing capability and adaptability of the initial object recognition network to the input point cloud data. The target point cloud training sample set in the first training process helps the model to form a relatively smooth decision boundary to reduce the risk of overfitting of the subsequent textile pilling level detection model to be trained. Then, the trained initial object recognition model is determined as the textile pilling level detection model to be trained. Afterwards, the textile point cloud training sample set is obtained. Thus, a textile point cloud training sample set for training the textile pilling level detection model to be trained can be obtained. Then, according to the textile point cloud training sample set, the textile pilling level detection model to be trained is second trained to obtain the trained textile pilling level detection model to be trained. The trained textile pilling level detection model to be trained is determined as the textile pilling level detection model. Thus, a textile pilling level detection model for detecting textiles can be obtained. Then, the three-dimensional point cloud data of the textile to be detected is collected as the point cloud data of the textile to be detected, wherein the point cloud data of the textile to be detected includes information of each data point, and each data point information in the above data point information includes coordinate position information and color information. Thus, the three-dimensional point cloud data of the textile to be detected can be obtained. Afterwards, the point cloud data of the textile to be detected is segmented to obtain a point cloud block information set, wherein each point cloud block information in the above point cloud block information set includes information of each data point.Thus, a point cloud block information set for generating local pilling level information of each textile can be obtained. Then, for each point cloud block information in the above point cloud block information set, the above point cloud block information is input into the above textile pilling level detection model to obtain the local pilling level information of the textile corresponding to the above point cloud block information. Thus, the point cloud block information set representing the three-dimensional point cloud data of the textile to be detected can be input into the textile pilling level detection model to perform local pilling detection and obtain the local pilling level information of each textile. Finally, based on the obtained local pilling level information of each textile, the textile pilling level detection information corresponding to the above textile to be detected is generated. Also because the local pilling level information of each textile corresponding to the three-dimensional point cloud data of the textile to be detected is generated through the textile pilling level detection model. The three-dimensional point cloud data of the textile to be tested can not only capture the planar features of the textile surface, but also reflect the slight changes in its three-dimensional structure and surface. In addition, the detection based on the three-dimensional point cloud data of the textile to be tested is less affected by changes in lighting and viewing angle, thereby improving the accuracy of the textile pilling level detection information and the stability of the detection results. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the accompanying drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and that components and elements are not necessarily drawn to scale.

[0016] Figure 1 is a flow chart of some embodiments of a method for detecting pilling levels of textile products according to the present disclosure;

[0017] Figure 2 It is a schematic structural diagram of some embodiments of a textile pilling level detection device according to the present disclosure;

[0018] Figure 3 It is a schematic diagram of the structure of an electronic device suitable for implementing some embodiments of the present disclosure. DETAILED DESCRIPTION

[0019] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments set forth herein. On the contrary, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0020] It should also be noted that, for ease of description, only the parts related to the invention are shown in the drawings. In the absence of conflict, the embodiments and features in the embodiments of the present disclosure can be combined with each other.

[0021] It should be noted that the concepts such as "first" and "second" mentioned in the present disclosure are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0022] It should be noted that the modifications of "one" and "plurality" mentioned in the present disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0023] The names of the messages or information exchanged between multiple devices in the embodiments of the present disclosure are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0024] The present disclosure will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0025] Figure 1 The process 100 of some embodiments of the method for detecting the pilling level of textiles according to the present disclosure is shown. The method for detecting the pilling level of textiles comprises the following steps:

[0026] Step 101: Obtain an initial point cloud training sample set.

[0027] In some embodiments, the execution subject (e.g., computing device) of the method for detecting the pilling level of textiles may obtain an initial point cloud training sample set through a wired connection or a wireless connection. The initial point cloud training sample set may be a training data set for training a neural network. Each initial point cloud training sample in the initial point cloud training sample set may include open source non-textile three-dimensional point cloud data and an item identifier corresponding to the non-textile three-dimensional point cloud data.

[0028] It should be noted that the above-mentioned wireless connection methods may include but are not limited to 3G / 4G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection, and other wireless connection methods currently known or to be developed in the future.

[0029] Step 102: Perform point cloud filtering processing on the initial point cloud training sample set to obtain a target point cloud training sample set.

[0030] In some embodiments, the execution entity may perform point cloud filtering processing on the initial point cloud training sample set to obtain a target point cloud training sample set.

[0031] In some optional implementations of some embodiments, the execution subject may perform point cloud filtering processing on the initial point cloud training sample set to obtain a target point cloud training sample set through the following steps:

[0032] In the first step, for each initial point cloud training sample in the above initial point cloud training sample set, perform the following update steps:

[0033] In the first sub-step, the sample initial point cloud data included in the initial point cloud training sample is determined as the point cloud data to be updated.

[0034] The second sub-step is to determine the various coordinate position information included in the various data point information in the above-mentioned point cloud data to be updated as a coordinate position information set. Among them, each coordinate position information in the above-mentioned various coordinate position information includes a horizontal coordinate, a vertical coordinate, and a vertical coordinate. The above-mentioned coordinate position information can represent the spatial position of the data point. It should be noted that none of the coordinate values ​​in the above-mentioned coordinate position information set is a negative number. In practice, the above-mentioned execution entity can determine the horizontal coordinate with the largest value in the above-mentioned coordinate position information set as the first target horizontal coordinate. The horizontal coordinate with the smallest value in the above-mentioned coordinate position information set is determined as the second target horizontal coordinate. The vertical coordinate with the largest value in the above-mentioned coordinate position information set is determined as the first target vertical coordinate. The vertical coordinate with the smallest value in the above-mentioned coordinate position information set is determined as the second target vertical coordinate. The vertical coordinate with the largest value in the above-mentioned coordinate position information set is determined as the first target vertical coordinate. The vertical coordinate with the smallest value in the above-mentioned coordinate position information set is determined as the second target vertical coordinate.

[0035] The third sub-step is to generate the first target horizontal coordinate, the second target horizontal coordinate, the first target vertical coordinate, the second target vertical coordinate, the first target vertical coordinate, and the second target vertical coordinate based on the above coordinate position information set.

[0036] The fourth sub-step is to generate the space information to be divided according to the abscissa of the first target, the abscissa of the second target, the ordinate of the first target, the ordinate of the second target, the vertical coordinate of the first target, and the vertical coordinate of the second target. In practice, the execution subject may generate the space information to be divided according to the abscissa of the first target, the abscissa of the second target, the ordinate of the first target, the ordinate of the second target, the vertical coordinate of the first target, and the vertical coordinate of the second target by a bounding box algorithm. The space information to be divided may be the coordinates of each vertex of a space prism. As an example, the abscissa of the first target may be represented by xmax. The abscissa of the second target may be represented by xmin. The vertical coordinate of the first target may be represented by ymax. The vertical coordinate of the second target may be represented by ymin. The vertical coordinate of the first target may be represented by zmax. The vertical coordinate of the second target may be represented by zmin. The above-mentioned spatial information to be divided can be “(xmin, ymin, zmin), (xmax, ymin, zmin), (xmin, ymax, zmin), (xmax, ymax, zmin), (xmin, ymin, zmax), (xmax, ymin, zmax), (xmin, ymax, zmax), (xmax, ymax, zmax)”.

[0037] The fifth sub-step is to determine the three-dimensional space corresponding to the above-mentioned space information to be divided as the surrounding space where the space points corresponding to each data point information are located. Wherein, each data point information in the above-mentioned data point information includes coordinate position information and color information. The above-mentioned coordinate position information can be a three-dimensional space coordinate. The above-mentioned color information can be a pixel value corresponding to the data point.

[0038] The sixth sub-step is to divide the enclosed space into three-dimensional grids. The lengths of the corresponding sides of every two three-dimensional grids in the three-dimensional grids are equal. In practice, the execution subject may divide the enclosed space into three-dimensional grids by voxelization technology.

[0039] The seventh sub-step is to perform the following filtering steps for each three-dimensional grid in the three-dimensional grid:

[0040] Sub-step 1: determining each data point information of the spatial points corresponding to the point cloud data to be updated in the above three-dimensional grid as each data point information to be filtered.

[0041] Sub-step two, in response to determining that the number of the to-be-filtered data point information in the above-mentioned each to-be-filtered data point information is greater than a preset value, at least one outlier data point information is determined from the above-mentioned each to-be-filtered data point information. In practice, the above-mentioned execution subject can detect each data point corresponding to each to-be-filtered data point information through the LOF outlier factor detection algorithm to obtain at least one outlier data point information. One outlier data point information in the above-mentioned at least one outlier data point information can represent a data point within a preset spatial range where the data point corresponding to the above-mentioned outlier data point information is located, and the data point density is less than or equal to a preset value. The above-mentioned preset spatial range can represent a spherical space with the data point corresponding to the above-mentioned outlier data point information as the sphere center and the preset radius as the sphere radius.

[0042] Sub-step three: deleting the at least one outlier data point information from the above-mentioned each data point information to be filtered, so as to update each data point information to be filtered.

[0043] Sub-step 4: determining the updated information of each to-be-filtered data point as sample updated point cloud data.

[0044] Sub-step five: updating the sample initial point cloud data included in the above-mentioned initial point cloud training sample to the above-mentioned sample update point cloud data, so as to update the initial point cloud training sample.

[0045] In the second step, the initial point cloud training sample set after each initial point cloud training sample is updated is determined as the target point cloud training sample set.

[0046] The above technical solution and its related contents, as an inventive point of an embodiment of the present disclosure, solve the second technical problem mentioned in the background technology, "directly using the unprocessed point cloud data corresponding to the identified object to train the initial object recognition network. The point cloud data of the identified object usually includes a large amount of data point information, which increases the computational complexity and computing resource requirements of model training, resulting in a waste of computer computing resources". The factors that lead to the waste of computer computing resources are often as follows: directly using the unprocessed point cloud data corresponding to the identified object to train the initial object recognition network. The point cloud data of the identified object usually includes a large amount of data point information, which increases the computational complexity and computing resource requirements of model training, resulting in a waste of computer computing resources. If the above factors are solved, the effect of reducing the waste of computer computing resources can be achieved. In order to achieve this effect, first, for each initial point cloud training sample in the above initial point cloud training sample set, the following update steps are performed: In the first sub-step, the sample initial point cloud data included in the above initial point cloud training sample is determined as the point cloud data to be updated. In the second sub-step, each coordinate position information included in each data point information in the above-mentioned point cloud data to be updated is determined as a coordinate position information set, wherein each coordinate position information in the above-mentioned coordinate position information includes a horizontal coordinate, a vertical coordinate, and a vertical coordinate. In the third sub-step, based on the above-mentioned coordinate position information set, the first target horizontal coordinate, the second target horizontal coordinate, the first target vertical coordinate, the second target vertical coordinate, the first target vertical coordinate, and the second target vertical coordinate are generated. Thus, the first target horizontal coordinate, the second target horizontal coordinate, the first target vertical coordinate, the second target vertical coordinate, the first target vertical coordinate, and the second target vertical coordinate used to generate the space information to be divided can be obtained. In the fourth sub-step, the space information to be divided is generated according to the first target horizontal coordinate, the second target horizontal coordinate, the first target vertical coordinate, the second target vertical coordinate, the first target vertical coordinate, and the second target vertical coordinate. Thus, the space information to be divided representing the three-dimensional space to be divided can be obtained. In the fifth sub-step, the three-dimensional space corresponding to the above-mentioned space information to be divided is determined as the enclosing space where the space points corresponding to each data point information are located. Thus, the enclosing space used to generate the three-dimensional grid can be obtained. The sixth sub-step is to divide the above-mentioned enclosed space into three-dimensional grids, wherein the lengths of the corresponding sides of every two three-dimensional grids in the above-mentioned three-dimensional grids are equal. The seventh sub-step is to perform the following filtering steps for each three-dimensional grid in the above-mentioned three-dimensional grid: Sub-step one is to determine the individual data point information of the spatial points corresponding to the point cloud data to be updated in the above-mentioned three-dimensional grid as the individual data point information to be filtered. In this way, the individual data point information to be filtered can be obtained. Sub-step two is to determine at least one outlier data point information from the above-mentioned individual data point information to be filtered in response to determining that the number of the data point information to be filtered in the above-mentioned individual data point information to be filtered is greater than a preset value.Thus, at least one outlier data point information for updating each data point information to be filtered can be obtained. Sub-step three, delete the above-mentioned at least one outlier data point information from the above-mentioned each data point information to be filtered to update each data point information to be filtered. Thus, at least one outlier data point information can be deleted from the above-mentioned each data point information to be filtered, and at least one outlier data point information is deleted on the basis of retaining the original shape features of the point cloud corresponding to each data point information to be filtered to reduce the amount of data. Sub-step four, determine the updated each data point information to be filtered as sample update point cloud data. Thus, sample update point cloud data representing the removal of at least one outlier data point information can be obtained. Sub-step five, update the sample initial point cloud data included in the above-mentioned initial point cloud training sample to the above-mentioned sample update point cloud data to update the initial point cloud training sample. Thus, the initial point cloud training sample can be updated to the sample update point cloud data with the outlier data point information removed. Finally, the initial point cloud training sample set after each initial point cloud training sample is updated is determined as the target point cloud training sample set. Thus, the target point cloud training sample set can be obtained after the outlier data point information is removed and the data volume is reduced. Also, because the sample initial point cloud data included in each initial point cloud training sample in the initial point cloud training sample set is divided, the information of each data point to be filtered corresponding to each three-dimensional grid is obtained. Then, the outlier data point information is deleted from each data point information to be filtered corresponding to each three-dimensional grid, and the data volume is reduced on the basis of retaining the original shape features of the point cloud corresponding to each data point information to be filtered. The data volume of the training data in the target point cloud training sample set is reduced, the computational complexity and computing resource requirements of the model training are reduced, and the waste of computer computing resources is reduced.

[0047] Step 103: Perform a first training on the initial object recognition network according to the target point cloud training sample set to obtain a trained initial object recognition model.

[0048] In some embodiments, the execution entity may perform a first training on the initial object recognition network according to the target point cloud training sample set to obtain a trained initial object recognition model.

[0049] In some optional implementations of some embodiments, the execution subject may perform a first training on the initial object recognition network according to the target point cloud training sample set to obtain a trained initial object recognition model through the following steps:

[0050] In the first step, the following first training step is performed based on the target point cloud training sample set:

[0051] In the first sub-step, the sample point cloud data of at least one target point cloud training sample in the target point cloud training sample set is input into the initial object recognition network to obtain an object recognition identifier corresponding to each target point cloud training sample in the at least one target point cloud training sample. The object recognition identifier may be an object name.

[0052] In the second sub-step, the sample target item identification corresponding to each target point cloud training sample in at least one target point cloud training sample is compared with the item identification identification corresponding to the target point cloud training sample. In practice, the above-mentioned execution subject can compare through a cross entropy loss function to determine the difference between the sample target item identification and the corresponding item identification identification.

[0053] The third sub-step is to determine whether the initial object recognition network has achieved a preset optimization target based on the comparison result. The above optimization target may be that the cross entropy loss function value is less than or equal to a preset value.

[0054] In a fourth sub-step, in response to determining that the initial object recognition network has achieved the optimization goal, the initial object recognition network is used as a trained initial object recognition model.

[0055] In a fifth sub-step, in response to determining that the initial object recognition network does not reach the optimization target, the network parameters of the initial object recognition network are adjusted, and the target point cloud training sample set is formed using the unused target point cloud training samples, and the adjusted initial object recognition network is used as the initial object recognition network to perform the first training step again. As an example, the network parameters of the initial object recognition network can be adjusted using a back propagation algorithm (BP algorithm).

[0056] Step 104: determine the trained initial object recognition model as the textile pilling level detection model to be trained.

[0057] In some embodiments, the execution entity may determine the trained initial object recognition model as the textile pilling level detection model to be trained.

[0058] Step 105: Obtain a textile point cloud training sample set.

[0059] In some embodiments, the execution entity may obtain a textile point cloud training sample set. In practice, the execution entity may obtain the textile point cloud training sample set via a wired connection or a wireless connection. The textile point cloud training sample set may be a training data set for performing a second training on the initial object recognition model that has completed training. Each textile point cloud training sample in the textile point cloud training sample set includes sample textile point cloud data and sample target pilling level information corresponding to the sample textile point cloud data. The sample textile point cloud data may be point cloud data of a textile. The sample target pilling level information may indicate the degree of pilling of the textile. For example, the sample target pilling level information may be level "1".

[0060] Step 106: Perform a second training on the to-be-trained textile pilling level detection model according to the textile point cloud training sample set to obtain a trained textile pilling level detection model.

[0061] In some embodiments, the execution entity may perform a second training on the to-be-trained textile pilling level detection model based on the textile point cloud training sample set to obtain a trained textile pilling level detection model.

[0062] In some optional implementations of some embodiments, the execution subject may perform a second training on the to-be-trained textile pilling level detection model according to the textile point cloud training sample set to obtain a trained textile pilling level detection model:

[0063] In the first step, the following second training step is performed based on the textile point cloud training sample set:

[0064] In the first sub-step, for each textile point cloud training sample in the textile point cloud training sample set, the following steps are performed:

[0065] Sub-step 1: Input the sample textile point cloud data included in the textile point cloud training sample into the above-mentioned point cloud data input layer to obtain each divided point cloud data. The above-mentioned point cloud data input layer may be an input layer that takes the sample textile point cloud data as input information and each divided point cloud data as output information. The above-mentioned each divided point cloud data may be each point cloud patch corresponding to each non-overlapping area. The above-mentioned each non-overlapping area may be each spatial area obtained by spatially dividing the sample textile point cloud data. One of the above-mentioned each point cloud patch may be a subset of the sample textile point cloud data.

[0066] Sub-step 2: Input each divided point cloud data into the above-mentioned point cloud embedding layer to obtain each divided point cloud data vector corresponding to each divided point cloud data. The above-mentioned point cloud embedding layer can be a patch Embedding layer that takes each divided point cloud data as input information and each divided point cloud data vector as output information. One of the above-mentioned divided point cloud data vectors can represent a point cloud patch.

[0067] Sub-step three, input each partitioned point cloud data vector into the above encoding layer to obtain each extracted point cloud feature information corresponding to each partitioned point data vector. The above encoding layer may be a Transformer Encoder layer that takes each partitioned point cloud data vector as input information and each extracted point cloud feature information as output information. Each of the extracted point cloud feature information may be a feature vector obtained by extracting features from the partitioned point data vector.

[0068] Sub-step 4: Input each extracted point cloud feature information into the first feature fusion layer to obtain fused feature extraction information. The first feature fusion layer may be a feature fusion layer (FFL) that takes each extracted point cloud feature information as input information and takes fused feature extraction information as output information. The fused feature extraction information may be a feature vector obtained by concatenating each extracted point cloud feature information.

[0069] Sub-step 5: Input each segmentation point cloud data vector into the position coding layer to obtain position coding information. The position coding information may be a position coding vector including absolute position information and relative position information. The absolute position information may be the coordinates of the data points in the segmentation point cloud data in space. The relative position information may represent the relative position relationship between each data point in the segmentation point cloud data.

[0070] Sub-step six, inputting the fused extracted feature information and the position coding information into the second feature fusion layer to obtain the target fused feature extraction information. The second feature fusion layer may be a feature fusion layer (Feature Fusion Layer, FFL) that takes the fused extracted feature information and the position coding information as input information and takes the target fused feature extraction information as output information. The target fused feature extraction information may be a vector obtained by fusing the fused extracted feature information with the position coding information.

[0071] Sub-step 7, input the target fusion feature extraction information into the above-mentioned pilling level detection layer to obtain the pilling level information of the corresponding textile point cloud training sample. The pilling level detection layer can be a fully connected layer that takes the target fusion feature extraction information as input information and takes the pilling level information as output information. The pilling level information can be the predicted pilling level information of the corresponding textile point cloud training sample. The pilling level information can indicate the pilling degree of the textile.

[0072] In the second sub-step, the sample target pilling level information corresponding to each textile point cloud training sample in the textile point cloud training sample set is compared with the detected pilling level information corresponding to the textile point cloud training sample. In practice, the above-mentioned execution subject can compare through a cross entropy loss function to determine the difference between the sample target pilling level information and the corresponding detected pilling level information.

[0073] In a third sub-step, in response to determining that the to-be-trained textile pilling level detection model reaches the optimization target, the to-be-trained textile pilling level detection model is used as the trained textile pilling level detection model.

[0074] In the fourth sub-step, in response to determining that the to-be-trained textile pilling level detection model has not reached the optimization target, the network parameters of the to-be-trained textile pilling level detection model are adjusted, and the unused textile point cloud training samples are used to form a textile point cloud training sample set, and the adjusted to-be-trained textile pilling level detection model is used as the to-be-trained textile pilling level detection model, and the second training step is performed again. As an example, the network parameters of the to-be-trained textile pilling level detection model can be adjusted by using a back propagation algorithm (BP algorithm).

[0075] Step 107: determine the trained textile pilling level detection model to be trained as the textile pilling level detection model.

[0076] In some embodiments, the execution entity may determine the trained textile pilling level detection model to be trained as the textile pilling level detection model.

[0077] Step 108 : collecting three-dimensional point cloud data of the textile to be inspected as point cloud data of the textile to be inspected.

[0078] In some embodiments, the execution entity may collect three-dimensional point cloud data of the textile to be detected as the point cloud data of the textile to be detected, wherein the point cloud data of the textile to be detected includes information of each data point, and each data point information in the above data point information includes coordinate position information and color information. In practice, the execution entity may collect three-dimensional point cloud data of the textile to be detected as the point cloud data of the textile to be detected by a linear array laser 3D camera. The coordinate position information may represent the coordinates of a data point in the point cloud data of the textile to be detected in three-dimensional space.

[0079] Step 109, segmenting the point cloud data of the textile to be inspected to obtain a point cloud block information set.

[0080] In some embodiments, the execution subject may segment the point cloud data of the textile to be detected to obtain a point cloud block information set, wherein each point cloud block information in the point cloud block information set includes information of each data point.

[0081] In some optional implementations of some embodiments, the execution subject may segment the point cloud data of the textile to be detected through the following steps to obtain a point cloud block information set:

[0082] In the first step, each data point information included in the above-mentioned textile point cloud data to be detected is determined as a data point information set to be divided.

[0083] In the second step, based on the information set of the data points to be divided, perform the following segmentation steps:

[0084] The first sub-step is to select a data point information to be divided from the data point information set to be divided as the central data point information. In practice, the execution subject may randomly select a data point information to be divided from the data point information set to be divided as the central data point information.

[0085] The second sub-step is to determine the distance between the spatial position corresponding to each of the data point information to be divided in the data point information set to be divided except the central data point information and the spatial position corresponding to the central data point information. The above distance may be the Euclidean distance. In practice, the execution subject may determine the Euclidean distance between the spatial position corresponding to the coordinate position information included in the data point information to be divided and the spatial position corresponding to the coordinate position information included in the central data point information.

[0086] The third sub-step is to determine the determined distances as a distance set to be screened.

[0087] The fourth sub-step is to determine each distance in the to-be-screened distance set that satisfies a preset screening condition as each target distance. The preset screening condition may be that the screening distance is less than or equal to the preset distance.

[0088] In the fifth sub-step, each piece of data point information to be divided corresponding to each target distance in the data point information set to be divided is determined as point cloud block information.

[0089] The sixth sub-step is to delete each piece of information of the data points to be divided from the information set of the data points to be divided, so as to update the information set of the data points to be divided.

[0090] In the third step, in response to determining that the updated information set of the data points to be divided is not empty, the segmentation step is performed again according to the updated information set of the data points to be divided.

[0091] In the fourth step, in response to determining that the updated information set of the data points to be divided is empty, the determined point cloud block information is determined as a point cloud block information set.

[0092] The above technical solution and its related contents, as an inventive point of the embodiment of the present disclosure, solve the problem that "point cloud data is segmented and processed by the octree algorithm to obtain each point cloud block. In the segmentation process, additional storage space is usually required to store the nodes and pointer information of the tree. The tree structure itself may need to consume a large amount of memory space, resulting in a waste of computer storage resources". The factors that lead to the waste of computer storage resources are often as follows: point cloud data is segmented and processed by the octree algorithm to obtain each point cloud block. In the segmentation process, additional storage space is usually required to store the nodes and pointer information of the tree. The tree structure itself may need to consume a large amount of memory space, resulting in a waste of computer storage resources. If the above factors are solved, the waste of computer storage resources can be reduced. In order to achieve this effect, first, the information of each data point included in the above-mentioned textile point cloud data to be detected is determined as a data point information set to be divided. Then, based on the data point information set to be divided, the following segmentation steps are performed: In the first substep, a data point information to be divided is selected from the data point information set to be divided as the central data point information. In this way, the central data point information corresponding to the central position in a point cloud data block can be obtained. The second sub-step is to determine the distance between the spatial position corresponding to each of the data point information to be divided in the data point information set to be divided except the central data point information and the spatial position corresponding to the central data point information. The third sub-step is to determine the determined distances as the distance set to be screened. Thus, a distance set to be screened for determining each target distance can be obtained. The fourth sub-step is to determine each distance that meets the preset screening conditions in the distance set to be screened as each target distance. Thus, each target distance for generating point cloud block information can be obtained. The fifth sub-step is to determine each of the data point information to be divided corresponding to each target distance in the data point information set to be divided as point cloud block information. Thus, point cloud block information representing the point cloud block after division can be obtained. The sixth sub-step is to delete each of the data point information to be divided from the data point information set to be divided to update the data point information set to be divided. Thus, each of the data point information to be divided that has been divided can be deleted from the data point information set to be divided. Afterwards, in response to determining that the updated data point information set to be divided is not empty, the segmentation step is performed again according to the updated data point information set to be divided. Next, in response to determining that the updated data point information set to be divided is empty, the determined individual point cloud block information is determined as a point cloud block information set. Thus, a divided point cloud block information set can be obtained. Also, in the process of dividing the individual data point information included in the point cloud data to be detected, no additional storage space for storing nodes and pointers is created, and the point cloud data to be detected does not need to be divided into blocks by nodes and pointers during the division process, thereby reducing the waste of computer storage resources.

[0093] Step 110: for each point cloud block information in the point cloud block information set, the point cloud block information is input into a textile pilling level detection model to obtain textile local pilling level information corresponding to the point cloud block information.

[0094] In some embodiments, the execution subject may input each point cloud block information in the point cloud block information set into the textile pilling level detection model to obtain the textile local pilling level information corresponding to the point cloud block information. The textile local pilling level information may be the pilling level information detected corresponding to the point cloud block information.

[0095] Step 111: Generate textile pilling level detection information corresponding to the textile to be detected based on the obtained local pilling level information of each textile.

[0096] In some embodiments, the execution body may generate the textile pilling level detection information corresponding to the textile to be detected based on the obtained local pilling level information of each textile.

[0097] In some optional implementations of some embodiments, the execution subject may generate the textile pilling level detection information corresponding to the textile to be detected based on the obtained local pilling level information of each textile through the following steps:

[0098] In the first step, each pilling level value included in the local pilling level information of each textile is determined as each target pilling level value. For example, the pilling level value may be 1.

[0099] The second step is to generate a textile pilling level value corresponding to the textile to be tested based on the above target pilling level values. In practice, the execution subject may determine the average of the above target pilling level values ​​as the textile pilling level value corresponding to the textile to be tested.

[0100] The third step is to determine the pilling level value of the textile as the pilling level detection information of the textile.

[0101] The above-mentioned embodiments of the present disclosure have the following beneficial effects: through the textile pilling level detection method of some embodiments of the present disclosure, the accuracy of the textile pilling level detection information and the stability of the detection results are improved. Specifically, the reason for the accuracy of the textile pilling level detection information and the stability of the detection results is that the textile pilling level detection is performed on the textile through the two-dimensional image of the textile, and the two-dimensional image is greatly affected by environmental changes and changes in viewing angles during imaging. For example, in the case of darker light, pilling may appear more difficult to distinguish, resulting in poor accuracy of the textile pilling level detection information. At the same time, changes in viewing angle may cause deformation of the textile in the image or changes in the surface texture, which may make the appearance of pilling different at different angles, and may cause the textile pilling level detection result to be unstable. Based on this, the textile pilling level detection method of some embodiments of the present disclosure, first, obtains an initial point cloud training sample set. Thus, an initial point cloud training sample set for generating a target point cloud training sample set can be obtained. Then, the initial point cloud training sample set is subjected to point cloud filtering processing to obtain a target point cloud training sample set. Thus, a target point cloud training sample set can be obtained. Afterwards, the initial object recognition network is first trained according to the target point cloud training sample set to obtain the trained initial object recognition model. Thus, the initial object recognition network can be trained by the enhanced model through the target point cloud training sample set, thereby improving the processing capability and adaptability of the initial object recognition network to the input point cloud data. The target point cloud training sample set in the first training process helps the model to form a relatively smooth decision boundary to reduce the risk of overfitting of the subsequent textile pilling level detection model to be trained. Then, the trained initial object recognition model is determined as the textile pilling level detection model to be trained. Afterwards, the textile point cloud training sample set is obtained. Thus, a textile point cloud training sample set for training the textile pilling level detection model to be trained can be obtained. Then, according to the textile point cloud training sample set, the textile pilling level detection model to be trained is second trained to obtain the trained textile pilling level detection model to be trained. The trained textile pilling level detection model to be trained is determined as the textile pilling level detection model. Thus, a textile pilling level detection model for detecting textiles can be obtained. Then, the three-dimensional point cloud data of the textile to be detected is collected as the point cloud data of the textile to be detected, wherein the point cloud data of the textile to be detected includes information of each data point, and each data point information in the above data point information includes coordinate position information and color information. Thus, the three-dimensional point cloud data of the textile to be detected can be obtained. Afterwards, the point cloud data of the textile to be detected is segmented to obtain a point cloud block information set, wherein each point cloud block information in the above point cloud block information set includes information of each data point.Thus, a point cloud block information set for generating local pilling level information of each textile can be obtained. Then, for each point cloud block information in the above point cloud block information set, the above point cloud block information is input into the above textile pilling level detection model to obtain the local pilling level information of the textile corresponding to the above point cloud block information. Thus, the point cloud block information set representing the three-dimensional point cloud data of the textile to be detected can be input into the textile pilling level detection model to perform local pilling detection and obtain the local pilling level information of each textile. Finally, based on the obtained local pilling level information of each textile, the textile pilling level detection information corresponding to the above textile to be detected is generated. Also because the local pilling level information of each textile corresponding to the three-dimensional point cloud data of the textile to be detected is generated through the textile pilling level detection model. The three-dimensional point cloud data of the textile to be tested can not only capture the planar features of the textile surface, but also reflect the slight changes in its three-dimensional structure and surface. In addition, the detection based on the three-dimensional point cloud data of the textile to be tested is less affected by changes in lighting and viewing angle, thereby improving the accuracy of the textile pilling level detection information and the stability of the detection results.

[0102] Further references Figure 2 As an implementation of the methods shown in the figures, the present disclosure provides some embodiments of a device for detecting the pilling level of textiles. These device embodiments are similar to Figure 1 Corresponding to the method embodiments shown, the device can be specifically applied to various electronic devices.

[0103] like Figure 2As shown, a textile pilling level detection device 200 in some embodiments includes: a first acquisition unit 201, a point cloud filtering processing unit 202, a first training unit 203, a first determination unit 204, a second acquisition unit 205, a second training unit 206, a second determination unit 207, a collection unit 208, a segmentation processing unit 209, an input unit 210 and a generation unit 211. Among them, the first acquisition unit 201 is configured to acquire an initial point cloud training sample set; the point cloud filtering processing unit 202 is configured to perform point cloud filtering processing on the above-mentioned initial point cloud training sample set to obtain a target point cloud training sample set; the first training unit 203 is configured to perform a first training on the initial object recognition network according to the above-mentioned target point cloud training sample set to obtain a trained initial object recognition model; the first determination unit 204 is configured to determine the trained initial object recognition model as the to-be-trained textile pilling level detection model; the second acquisition unit 205 is configured to acquire a textile point cloud training sample set; the second training unit 206 is configured to perform a second training on the to-be-trained textile pilling level detection model according to the above-mentioned textile point cloud training sample set to obtain a trained textile pilling level detection model; the second determination unit 207 is configured to determine the trained textile pilling level detection model as the to-be-trained textile pilling level detection model. A fabric pilling level detection model; a collection unit 208 is configured to collect three-dimensional point cloud data of the textile to be detected as the textile point cloud data to be detected, wherein the above-mentioned textile point cloud data to be detected includes information of each data point, and each data point information in the above-mentioned data point information includes coordinate position information and color information; a segmentation processing unit 209 is configured to segment the above-mentioned textile point cloud data to be detected to obtain a point cloud block information set, wherein each point cloud block information in the above-mentioned point cloud block information set includes information of each data point; an input unit 210 is configured to input the above-mentioned point cloud block information into the above-mentioned textile pilling level detection model for each point cloud block information in the above-mentioned point cloud block information set, and obtain the textile local pilling level information corresponding to the above-mentioned point cloud block information; a generation unit 211 is configured to generate textile pilling level detection information corresponding to the above-mentioned textile to be detected based on the obtained local pilling level information of each textile.

[0104] It is understood that the units described in the device 200 are similar to those described in the reference Figure 1 Therefore, the operations, features and beneficial effects described above for the method are also applicable to the device 200 and the units contained therein, and will not be described in detail here.

[0105] Reference below Figure 3 , which shows a structural schematic diagram of an electronic device 300 suitable for implementing some embodiments of the present disclosure. Figure 3The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0106] like Figure 3 As shown, the electronic device 300 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 301, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a storage device 308 into a random access memory (RAM) 303. In the RAM 303, various programs and data required for the operation of the electronic device 300 are also stored. The processing device 301, the ROM 302, and the RAM 303 are connected to each other via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0107] Typically, the following devices may be connected to the I / O interface 305: input devices 306 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; output devices 307 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices 308 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 309. The communication devices 309 may allow the electronic device 300 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 3 The electronic device 300 is shown with various devices, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead. Figure 3 Each block shown in the figure may represent one device, or may represent multiple devices as required.

[0108] In particular, according to some embodiments of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, some embodiments of the present disclosure include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In some such embodiments, the computer program can be downloaded and installed from the network through the communication device 309, or installed from the storage device 308, or installed from the ROM 302. When the computer program is executed by the processing device 301, the functions defined in the method of some embodiments of the present disclosure are executed.

[0109] It should be noted that the computer-readable medium recorded in some embodiments of the present disclosure may be a computer-readable signal medium or a computer-readable storage medium or any combination of the two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In some embodiments of the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, device or device. In some embodiments of the present disclosure, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which a computer-readable program code is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The computer readable signal medium may also be any computer readable medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination thereof.

[0110] In some embodiments, the client and the server may communicate using any currently known or future developed network protocol such as HTTP (HyperText Transfer Protocol), and may be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), an internet (e.g., the Internet), and a peer-to-peer network (e.g., an ad hoc peer-to-peer network), as well as any currently known or future developed network.

[0111] The computer-readable medium may be included in the electronic device; or it may exist independently without being installed in the electronic device. The computer-readable medium carries one or more programs. When the one or more programs are executed by the electronic device, the electronic device: obtains an initial point cloud training sample set; performs point cloud filtering processing on the initial point cloud training sample set to obtain a target point cloud training sample set; performs a first training on the initial object recognition network based on the target point cloud training sample set to obtain a trained initial object recognition model; determines the trained initial object recognition model as the pilling level detection model for the textile to be trained; obtains a textile point cloud training sample set; performs a second training on the textile pilling level detection model for the textile to be trained based on the textile point cloud training sample set to obtain a trained pilling level detection model for the textile to be trained; determines the trained pilling level detection model for the textile to be trained as the textile A pilling level detection model; collecting three-dimensional point cloud data of a to-be-detected textile as the to-be-detected textile point cloud data, wherein the to-be-detected textile point cloud data includes information of each data point, and each of the data point information includes coordinate position information and color information; segmenting the to-be-detected textile point cloud data to obtain a point cloud block information set, wherein each point cloud block information in the point cloud block information set includes information of each data point; for each point cloud block information in the point cloud block information set, inputting the point cloud block information into the textile pilling level detection model to obtain textile local pilling level information corresponding to the point cloud block information; and generating textile pilling level detection information corresponding to the to-be-detected textile based on the obtained local pilling level information of each textile.

[0112] Computer program code for performing the operations of some embodiments of the present disclosure may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0113] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0114] The units described in some embodiments of the present disclosure may be implemented by software or by hardware. The units described may also be set in a processor, for example, may be described as: a processor includes a first acquisition unit, a point cloud filtering processing unit, a first training unit, a first determination unit, a second acquisition unit, a second training unit, a second determination unit, a collection unit, a segmentation processing unit, an input unit, and a generation unit. Among them, the names of these units do not constitute a limitation on the unit itself under certain circumstances. For example, the generation unit may also be described as "a unit for generating textile pilling level detection information corresponding to the above-mentioned textile to be detected based on the obtained local pilling level information of each textile".

[0115] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.

[0116] The above descriptions are only some preferred embodiments of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present disclosure is not limited to the technical solutions formed by a specific combination of technical features, but should also cover other technical solutions formed by any combination of technical features or their equivalent features without departing from the inventive concept. For example, a technical solution formed by replacing the features with (but not limited to) the technical features with similar functions disclosed in the embodiments of the present disclosure.

Claims

1. A method for detecting the pilling level of textiles. include: Obtain an initial point cloud training sample set; Performing point cloud filtering processing on the initial point cloud training sample set to obtain a target point cloud training sample set; Performing a first training on the initial object recognition network according to the target point cloud training sample set to obtain a trained initial object recognition model; The trained initial object recognition model is determined as a textile pilling level detection model to be trained; Obtain a textile point cloud training sample set; According to the textile point cloud training sample set, a second training is performed on the textile pilling level detection model to be trained to obtain a trained textile pilling level detection model to be trained; Determine the trained textile pilling level detection model to be trained as the textile pilling level detection model; Collecting three-dimensional point cloud data of the textile to be detected as point cloud data of the textile to be detected, wherein the point cloud data of the textile to be detected includes information of each data point, and each data point information in the information of each data point includes coordinate position information and color information; Segmenting the point cloud data of the textile to be detected to obtain a point cloud block information set, wherein each point cloud block information in the point cloud block information set includes information of each data point; For each point cloud block information in the point cloud block information set, inputting the point cloud block information into the textile pilling level detection model to obtain the textile local pilling level information corresponding to the point cloud block information; Based on the obtained local pilling level information of each textile, textile pilling level detection information corresponding to the textile to be detected is generated.

2. The method according to claim 1, in, The target point cloud training samples in the target point cloud training sample set include sample point cloud data and sample target object identifiers corresponding to the sample point cloud data; And the first training of the initial object recognition network is performed according to the target point cloud training sample set to obtain a trained initial object recognition model, including: The following first training step is performed based on the target point cloud training sample set: Inputting sample point cloud data of at least one target point cloud training sample in the target point cloud training sample set into the initial object recognition network to obtain an object recognition identifier corresponding to each target point cloud training sample in the at least one target point cloud training sample; Comparing the sample target object identification corresponding to each target point cloud training sample in at least one target point cloud training sample with the object identification identification corresponding to the target point cloud training sample; Determine whether the initial object recognition network reaches a preset optimization goal based on the comparison result; In response to determining that the initial object recognition network achieves the optimization goal, using the initial object recognition network as a trained initial object recognition model; In response to determining that the initial object recognition network does not achieve the optimization goal, the network parameters of the initial object recognition network are adjusted, and unused target point cloud training samples are used to form a target point cloud training sample set, and the adjusted initial object recognition network is used as the initial object recognition network to perform the first training step again.

3. The method according to claim 1, in, The to-be-trained textile pilling level detection model comprises a point cloud data input layer, a point cloud embedding layer, an encoding layer, a first feature fusion layer, a position encoding layer, a second feature fusion layer, and a pilling level detection layer, and each textile point cloud training sample in the textile point cloud training sample set comprises sample textile point cloud data and sample target pilling level information corresponding to the sample textile point cloud data; And the second training of the to-be-trained textile pilling level detection model is performed according to the textile point cloud training sample set to obtain a trained textile pilling level detection model, including: The following second training step is performed based on the textile point cloud training sample set: For each textile point cloud training sample in the textile point cloud training sample set, perform the following steps: Inputting the sample textile point cloud data included in the textile point cloud training sample into the point cloud data input layer to obtain each divided point cloud data; Input each divided point cloud data into the point cloud embedding layer to obtain each divided point cloud data vector corresponding to each divided point cloud data; Input each divided point cloud data vector into the encoding layer to obtain each extracted point cloud feature information corresponding to each divided point data vector; Inputting each extracted point cloud feature information into the first feature fusion layer to obtain fused feature extraction information; Inputting each divided point cloud data vector into the position encoding layer to obtain position encoding information; Inputting the fused extracted feature information and the position encoding information into the second feature fusion layer to obtain target fused feature extraction information; Inputting the target fusion feature extraction information into the pilling level detection layer to obtain the pilling level detection information of the corresponding textile point cloud training sample; Compare the sample target pilling level information corresponding to each textile point cloud training sample in the textile point cloud training sample set with the detected pilling level information corresponding to the textile point cloud training sample; In response to determining that the to-be-trained textile pilling level detection model reaches the optimization target, the to-be-trained textile pilling level detection model is used as the trained textile pilling level detection model to-be-trained; In response to determining that the textile pilling level detection model to be trained does not reach the optimization target, the network parameters of the textile pilling level detection model to be trained are adjusted, and unused textile point cloud training samples are used to form a textile point cloud training sample set, and the adjusted textile pilling level detection model to be trained is used as the textile pilling level detection model to be trained, and the second training step is performed again.

4. The method according to claim 1, in, The step of generating textile pilling level detection information corresponding to the textile to be detected based on the obtained local pilling level information of each textile comprises: Determine each pilling level value included in each local pilling level information of the textile as each target pilling level value; Based on the target pilling level values, generating a textile pilling level value corresponding to the textile to be tested; The textile pilling level value is determined as textile pilling level detection information.

5. The method according to claim 1, in, The method further comprises: Controlling the associated printing device to print the obtained textile pilling level detection information onto label paper; The associated mechanical arm is controlled to stick the label paper onto the textile to be inspected.

6. A device for detecting the pilling level of textiles, include: A first acquisition unit is configured to acquire an initial point cloud training sample set; A point cloud filtering processing unit is configured to perform point cloud filtering processing on the initial point cloud training sample set to obtain a target point cloud training sample set; A first training unit is configured to perform a first training on the initial object recognition network according to the target point cloud training sample set to obtain a trained initial object recognition model; A first determination unit is configured to determine the trained initial object recognition model as a textile pilling level detection model to be trained; A second acquisition unit is configured to acquire a textile point cloud training sample set; A second training unit is configured to perform a second training on the to-be-trained textile pilling level detection model according to the textile point cloud training sample set, to obtain a trained textile pilling level detection model to be trained; A second determination unit is configured to determine the trained textile pilling level detection model to be trained as the textile pilling level detection model; A collection unit is configured to collect three-dimensional point cloud data of the textile to be detected as point cloud data of the textile to be detected, wherein the point cloud data of the textile to be detected includes information of each data point, and each data point information in the information of each data point includes coordinate position information and color information; A segmentation processing unit is configured to perform segmentation processing on the point cloud data of the textile to be detected to obtain a point cloud block information set, wherein each point cloud block information in the point cloud block information set includes information of each data point; An input unit is configured to input each point cloud block information in the point cloud block information set into the textile pilling level detection model to obtain textile local pilling level information corresponding to the point cloud block information; The generating unit is configured to generate the textile pilling level detection information corresponding to the textile to be detected based on the obtained local pilling level information of each textile.

7. An electronic device, include: one or more processors; a storage device having one or more programs stored thereon; When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 5.

8. A computer readable medium having a computer program stored thereon, in, When the program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

Citation Information

Patent Citations

  • Fabric fuzzing and pilling rating method based on deep learning

    CN113205491A

  • Point cloud detection model training method and device, electronic equipment and storage medium

    CN114820465A

  • Fabric fuzzing and pilling grade automatic evaluation method based on three-dimensional scanning

    CN116091465A

  • Point cloud detection and segmentation method and apparatus, and electronic device

    WO2023193400A1