A fully automated waste fabric identification and sorting control system

Abstract

The application discloses a kind of full automatic waste fabric identification sorting control system, it is related to resource reusing technical field, including identification analysis module, the type, color, component and fibre fineness data of waste fabric on the surface of automatic conveying module are acquired and analyzed by identification analysis module, obtain the component composition and its weight distribution of each area of waste fabric, generate data form and send to data module, and sorting action module is controlled by sorting strategy control module to carry out sorting screening;Using machine vision technology and fabric visual identification model determines the type of waste fabric and divides its area, so that different parts of fabric can be detected and identified, improve the component identification precision of fabric;By extracting the color value of different areas and correcting the color value of spectral data in the identification analysis module, the influence of the color on the surface of the fabric on light, especially visible light, is eliminated, making the spectral data closer to the truth, and the analysis result is more accurate.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of resource recycling, and particularly relates to a full-automatic waste fabric identification and sorting control system. BACKGROUND

[0002] Waste fabric refers to textile materials and products abandoned in the production and use process. The sources of waste fabric mainly include three aspects: first, waste silk, residual materials, offcuts and the like generated in the process of manufacturing textiles; second, clothes discarded after use, abandoned bedding, curtains, carpets and the like; and third, waste plastics such as polyester bottles having utilization value. The generation of waste fabric is mainly some expired resources in production and life.

[0003] With the improvement of the living standards of residents and the advancement of textile technology, the amount of discarded waste fabric also increases; and in the face of such a large amount of waste fabric, the main treatment methods include landfill or incineration, and only a small amount is recycled. Landfill treatment occupies a large amount of land, and is accompanied by hidden troubles such as residual bacteria, viruses, methane, heavy metals and leakage. Incineration treatment will produce harmful gases. These traditional waste fabric treatment methods will pollute the environment; therefore, it is necessary to identify and sort the waste fabric;

[0004] Traditional identification and sorting of waste clothes and textiles mainly rely on the subjective judgment of skilled workers. It includes hand feeling, luster, state and smell after burning and the like, but the identification and sorting efficiency and quality of this method cannot be guaranteed, and safety cannot be guaranteed. In addition, traditional chemical analysis methods can accurately identify the composition of waste textiles, but the above method is time-consuming, destroys the measured sample and has high cost, and therefore cannot be widely used in the waste textile recycling industry, and is only suitable for third-party institutions to conduct sampling inspection on the composition of clothes materials.

[0005] The current so-called non-destructive rapid identification technology mainly adopts near-infrared spectroscopy technology, but the technology currently mainly uses traditional laboratory near-infrared spectroscopic instruments for offline identification. The above identification technology generally needs a long identification time, and is not suitable for identification and sorting of a large number of clothes and textile compositions in large industrial production. Therefore, we provide a full-automatic waste fabric identification and sorting control system. SUMMARY

[0006] The present application relates to the technical field of resource recycling, and particularly relates to a full-automatic waste fabric identification and sorting control system.

[0007] The present application relates to the technical field of resource recycling, and particularly relates to a full-automatic waste fabric identification and sorting control system.

[0008] How to determine the type of waste fabric and divide the area by combining machine vision technology and near-infrared spectrum technology, solve the problem of inaccurate content composition of fabric composition in the online detection and sorting of the prior art;

[0009] How to solve the problem that the near-infrared spectrum technology is affected by the color of the fabric itself by extracting the color value of different areas and correcting the color value of the spectrum data in the identification analysis module;

[0010] How to get the total weight ratio of the main component of the fabric by area weight ratio calculation of the corresponding type of fabric, solve the problem of single sorting strategy and weak data support in the subsequent sorting process;

[0011] How to simulate the composition of the corresponding waste fabric through the big data support and data prediction module, directly develop a strategy to complete the sorting, solve the problem that the system cannot effectively complete the sorting under the pressure of a large number of fabric backlog and data processing in the prior art.

[0012] The application can be realized by the following technical scheme: a full-automatic waste fabric identification and sorting control system, comprising an automatic conveying module for orderly conveying waste fabric and a sorting action module for performing grabbing and sorting actions on the waste fabric, the system further comprising an identification analysis module, which uses a combination of machine vision technology and near-infrared spectrum technology to obtain and analyze the type, color, composition and fiber fineness data of the waste fabric on the surface of the automatic conveying module, obtain the composition and weight distribution of each area of the waste fabric, generate a data table and send it to the data module, and the control module controls the sorting action module to perform sorting and screening through a sorting strategy.

[0013] Further technical improvements of the application are that the identification analysis module determines the type of waste fabric through a constructed fabric visual identification model, and divides the area according to the corresponding position of the human body on the image of the corresponding type of fabric, identifies and records the color value of each area.

[0014] Further technical improvements of the application are that the identification analysis module continuously irradiates the divided area with a continuous wavelength, generates original spectrum data and pre-processes and extracts features from the original spectrum data and corrects the color value.

[0015] Further technical improvements of the application are that the step of pre-processing the original spectrum data comprises:

[0016] Step one, using a data enhancement algorithm to enhance the original spectrum data, enhancing the difference of the spectrum data of different fabric fibers;

[0017] Step two, smoothing the data to eliminate random noise in the data acquisition process;

[0018] Step three, derivative processing of the above data to solve the baseline offset problem;

[0019] Step four, multivariate scattering correction processing of the spectral data to improve the signal-to-noise ratio of the spectrum.

[0020] Further technical improvements of the present application are that the feature extraction of the spectral data is based on a binary particle swarm algorithm, the particle velocity is mapped using a sigmoid function, the mapping result obtained is used as a probability value, and the probability value is compared with a probability setting threshold value. When and only when the probability value is not lower than the probability setting threshold value, it means that the wavelength point corresponding to the vector is selected, and the spectral data at the wavelength is the characteristic spectrum.

[0021] Further technical improvements of the present application are that the color value correction operation of the spectral data extracted by the feature extraction includes:

[0022] The spectral data extracted by the feature extraction in each set region is marked as , wherein i represents the number of the set region, and j represents the number of the data in the spectral data;

[0023] The color data of the i-th set region is marked as , and the corresponding fiber without dyeing is taken as a reference standard, and the color value correction coefficient about the current color data is obtained by comparison calculation, and the formula is: ;

[0024] , wherein is the color value correction coefficient; C represents the color data of the undyed fiber; represents the wavelength data of the visible light corresponding to the color of the undyed fiber; represents the wavelength data of the visible light corresponding to the color of the undyed fiber; represents the correlation degree of the color value correction coefficient and the corresponding wavelength data;

[0025] The spectral data and the color value correction coefficient are substituted into the calculation formula to obtain the corrected spectral data .

[0026] Further technical improvements of the present application are that the NIR spectrum is generated online according to the corrected spectral data, and the segmented regression model established based on principal component analysis and least square method is used to classify and identify the fibers in the waste fabric, to obtain the component composition of the fibers in the set region and transmit to the data prediction module.

[0027] Further technical improvements of the present application are that the data prediction module calculates the composition proportion of all fiber components and the weight of corresponding components in the entire waste fabric according to the weight, area and component composition of each region of the entire waste fabric, and generates a data table and sends it to the control module.

[0028] Further technical improvements of the present application are that under the support of big data, the data prediction module also predicts and sorts according to the type of fabric. In the case of determining the type of waste fabric, the relevant component composition and the corresponding component proportion interval can be directly predicted without the need for identification and analysis module to directly sort.

[0029] Further technical improvements of the present application are that the control module is provided with a sorting strategy, which sorts according to the type of waste fabric and the component of corresponding component weight proportion, or in the case of the same type and / or component weight proportion, further sorts according to the fiber fineness.

[0030] Compared with the prior art, the present application has the following beneficial effects:

[0031] 1. By combining machine vision technology and near-infrared spectroscopy technology, online detection of waste fabric is realized. The type of waste fabric is determined by machine vision technology and fabric visual recognition model, and the waste fabric is divided into regions, so that different parts of the fabric can be detected and recognized, greatly improving the component recognition accuracy of the fabric and being beneficial to subsequent sorting.

[0032] 2. By extracting the color values of different regions and correcting the color values of the spectral data in the identification and analysis module, the influence of the color of the fabric surface on light, especially visible light, is eliminated, making the spectral data more accurate and the final analysis result more accurate.

[0033] 3. By region division and area weight ratio calculation of corresponding types of fabric, the total weight ratio of main component of the fabric is obtained, so that various sorting strategies can be developed, and the execution of the sorting strategy also has strong data support. BRIEF DESCRIPTION OF DRAWINGS

[0034] In order to facilitate those skilled in the art to understand, the present application will be further described below with reference to the drawings.

[0035] Fig. 1 The system structure block diagram of the present application is shown in the figure.

[0036] Fig. 2 The sorting system method flow chart of the present application is shown in the figure. DETAILED DESCRIPTION

[0037] In order to further illustrate the technical means and effects adopted by the present application to achieve the predetermined inventive purpose, the specific embodiments, structures, features and effects thereof according to the present application are described in detail below in combination with the drawings and preferred embodiments.

[0038] Referring to Figs. 1-2 As shown in the drawings, a full-automatic waste fabric recognition and sorting control system comprises an automatic conveying module, a sorting action module, a recognition and analysis module, a data prediction module, a control module and a storage module.

[0039] Specifically, the automatic conveying module is one or more transmission devices for orderly conveying waste fabrics, and the transmission device has the function of flattening and conveying the waste fabrics.

[0040] The sorting action module comprises a plurality of sorting mechanical hands, which are sequentially arranged according to the transmission direction of the automatic conveying module. The sorting action module controls the actions of the plurality of mechanical hands according to the instructions of the control module, and performs grabbing and sorting actions on the corresponding waste fabrics.

[0041] The recognition and analysis module collects and recognizes the information of the fabrics on the surface of the conveying device. The recognition and analysis module uses a combination of machine vision technology and near-infrared spectroscopy technology to obtain and analyze the type, color, composition and fiber fineness of the waste fabrics on the surface of the conveying device. The fabric composition mainly includes polyester, wool and polyamide, cotton fabric, polyester fabric and blended fabric.

[0042] Specifically, the type data and color data of the waste fabrics are extracted and recorded by using machine vision technology. A fabric visual recognition model based on a CNN convolutional neural network is set in the storage module, which mainly identifies the fabric type and fabric color of the waste fabrics.

[0043] The CNN convolutional neural network is a mature machine learning network architecture, which is not described in detail in this paper. The present application uses the CNN convolutional neural network to distribute a large amount of historical data of waste fabrics of different types and different colors in a ratio of 9:1 as training set and test set respectively, so as to train the above-mentioned fabric visual recognition model. The above-mentioned type specifically refers to different types of previous use scenarios of the waste fabrics, such as bedding, clothes of different seasons (including shirts or trousers), knitted bags, etc. The determination of different types is conducive to the accuracy of subsequent analysis and prediction results. The above-mentioned color refers to the division of color regions on the surface of the fabric, and the color value is obtained to facilitate parameter adjustment and correction in the subsequent near-infrared spectroscopy analysis process, so as to eliminate the error phenomenon of the composition analysis result caused by different colors.

[0044] The near-infrared spectrum technology can be applied to identification and component prediction of the waste textiles. The essence is that different substances contain different chemical groups, different chemical groups have different near-infrared spectra, and the one-to-one correspondence between different groups and different spectra can be used to realize identification and component prediction of the fiber types. Meanwhile, although the waste textiles have been used, the chemical structure and molecular composition of the fibers do not change, and the fiber composition and content of the original textiles have little difference. Therefore, it is feasible to identify and predict the components of the waste textiles by using the near-infrared technology.

[0045] In the type judgment of the waste textiles, the following aspects are mainly considered, including the fabric shape, the fabric thickness, the wire diameter of the fabric wire and the mesh size;

[0046] In the identification and analysis module, the high-definition CCD industrial camera acquires the fabric image of the waste textile on the transmission device, extracts the corresponding type characteristic data according to the above type characteristics (shape, thickness, wire diameter and mesh size), and inputs the type characteristic data into the fabric visual identification model, so as to determine the type of the waste textile through the fabric visual identification model;

[0047] After determining the type, the identification and analysis module performs regional division on the fabric image according to the corresponding type, takes the sleeve, collar, hem and corresponding parts of the human chest as the set regions, and extracts the color data of the set regions;

[0048] Subsequently, the near-infrared spectrum analysis is performed on the above set regions:

[0049] S1: In the above set region, continuous wavelength light is irradiated to the surface of the waste textile in the set region, so as to generate a vibration spectrum;

[0050] It should be noted that the molecular vibration absorbs the incident light, and different molecules have different absorption capacities for different wavelengths of light. A molecule can only absorb radiation that causes its own vibration change. When near-infrared light acts on the molecules in the substance, the molecules absorb the absorption spectrum of the specific wavelength of infrared light, which is called vibration spectrum, and appears in the form of spectral band;

[0051] S2: Since the collected spectrum data often has much random noise, background interference and other useless information, the existence of these interferences will affect the accuracy of the component analysis result, and the spectrum data needs to be preprocessed:

[0052] The spectrum data is processed by using a data enhancement algorithm to increase the difference of the spectrum data corresponding to different fiber materials. In this embodiment, the data enhancement algorithm adopts mean centering.

[0053] The SG convolution smoothing method is used to smooth random noise. The fitted value of the data within a certain interval before and after the smoothed point is used to replace the original data of the smoothed point. The fitted value is fitted by the polynomial least squares method. In this embodiment, the polynomial degree of SG smoothing is set to the window width of 5, the polynomial degree of the second derivative of SG is set to 3, and the moving window width is set to 7.

[0054] The SG convolution derivative method is used to solve the baseline shift problem that is common in spectral data.

[0055] After performing multivariate scattering correction on the spectral data, the baseline shift and offset of each spectrum are corrected with reference to the standard spectrum, thereby improving the signal-to-noise ratio of the spectrum.

[0056] S3: Feature extraction from spectral data based on binary particle swarm optimization algorithm.

[0057] In the binary particle swarm optimization algorithm, the position component of each particle is set to 0 or 1, and the particle's velocity represents the probability that the particle's position is 1.

[0058] When filtering feature spectral data, the vector length of each particle is set to the number of wavelength points in the original spectrum. The particle vector corresponds to the number of wavelength variables in the original spectrum. The probability of a particle taking the value 0 or 1 depends on its position component. In the binary particle swarm optimization algorithm, the particle velocity is mapped using the sigmoid function, and the mapping result serves as the probability of the position component taking the value 1.

[0059] The mapping formula for the sigmoid function is as follows:

[0060] ;

[0061] After obtaining the probability, the probability value is compared with the probability setting threshold. The position component is set to 1 if and only if the probability value is not lower than the probability setting threshold, and is set to 0 in all other cases.

[0062] When the particle component is 1, it means that the wavelength point corresponding to the vector is selected, and the spectral data at that wavelength is the feature spectrum. When it is 0, it means that the corresponding variable is not selected, thus completing the feature extraction of the entire spectral data.

[0063] S4: Since the color of the waste fabric itself can affect the spectral data, this invention also performs color value correction on the spectral data after feature extraction:

[0064] The spectral data extracted from each defined region are labeled as follows: Where i represents the number of the defined region, and j represents the number of data points in the spectral data;

[0065] Color data of the ith set region is marked as , and the corresponding fiber without dyeing is taken as a reference, a color value correction coefficient of the current color data is calculated by comparison, and the formula is: ;

[0066] The formula is dimensionless calculation; wherein, is the color value correction coefficient;

[0067] C represents the color data of the undyed fiber, and the color data here is calculated by weighted average of the color data of the fibers themselves;

[0068] represents the wavelength data corresponding to the color of the undyed fiber, and the wavelength data here is calculated by weighted average of the wavelength data corresponding to the color of the fibers themselves;

[0069] represents the correlation degree of the color value correction coefficient and the corresponding wavelength data, and when the wavelength is long, the correlation degree is large, and when the wavelength is short, the correlation degree is small;

[0070] The spectral data and the color value correction coefficient are substituted into the calculation formula to obtain the corrected spectral data ;

[0071] S5: generating an NIR spectrum (i.e. near infrared spectrum) online according to the corrected spectral data in S4, and then establishing a segmented regression model of the component to be measured by principal component analysis and least square method, to classify and identify the fibers in the waste fabric, and obtain the component composition of the fibers in the set region;

[0072] The component composition obtained by identification and analysis is transmitted to a data prediction module, and the data prediction module predicts and calculates the component composition and component proportion of the entire waste fabric:

[0073] First, the area of the region divided by the identification and analysis module is calculated, and the overall weight of the waste fabric is weighed, and the overall weight is divided according to the position of the set region and the type of the waste fabric, the division is based on the area and the type characteristic distribution proportion coefficient, and the weight of each set region is obtained according to the proportion coefficient and the overall weight;

[0074] Then, based on the weight of each set region and the component composition of the corresponding region, since the component composition contains the component proportion, the component proportion of the corresponding region is multiplied by the weight of the region to obtain the weight of the related component of the corresponding region;

[0075] Finally, the weights of the same components of the entire waste fabric are summed up to obtain the composition ratio of all components in the waste fabric and the weight of the corresponding component, and a data table is generated and sent to the control module.

[0076] The control module is provided with a sorting strategy, such as sorting according to the type characteristics of the waste fabric, or sorting control combined with the component weight ratio, and further, in the case of the same type characteristics and / or component weight ratio, further screening and sorting can be made according to the fiber fineness, so as to achieve fine sorting matching different sorting strategies.

[0077] The control module controls the sorting action module to sort the target waste fabric in a targeted manner through wired or wireless network communication.

[0078] More, in the case of a large amount of historical data accumulated in the later period, the data prediction module can also directly predict the sorting of the waste fabric of the corresponding type (such as bedding, single clothes, etc.), that is, in the case of big data support, the related component composition and the corresponding component ratio interval can be directly predicted after determining the type of the waste fabric, so that sorting is directly performed without the need of the recognition analysis module, which can greatly improve the efficiency and ensure a certain degree of accuracy.

[0079] The above is only a preferred embodiment of the present application, and does not limit the present application in any form. Although the present application has been disclosed as above, it is not intended to limit the present application. Any person skilled in the art can make some changes or modifications to the above disclosed technical content to obtain equivalent embodiments with equivalent changes, without departing from the scope of the technical solution of the present application. Any simple modification, equivalent change and modification of the above embodiments made according to the technical essence of the present application shall still fall within the scope of the technical solution of the present application.

Claims

1. A fully automated waste textile identification and sorting control system, comprising an automatic conveying module for orderly conveying waste textiles and a sorting action module for performing gripping and sorting actions on the waste textiles, characterized in that, The system also includes an identification and analysis module. The identification and analysis module uses a combination of machine vision technology and near-infrared spectroscopy technology to acquire and analyze data on the type, color, composition and fiber fineness of waste fabrics on the surface of the automatic conveying module. It obtains the composition and weight distribution of each area of ​​the waste fabrics, generates a data form and sends it to the data module. The control module controls the sorting action module to perform sorting and screening through sorting strategies. The identification and analysis module determines the type of waste fabric by constructing a fabric visual recognition model, and divides the fabric image of the corresponding type into regions according to the corresponding human body position, identifies the color of each region, and records the color value. The identification and analysis module irradiates the divided area with continuous wavelengths to generate raw spectral data and performs preprocessing, feature extraction, and color value correction on the raw spectral data. The operation of color value correction for the spectral data after feature extraction includes: The spectral data extracted from each defined region are labeled as follows: Where i represents the number of the defined region, and j represents the number of data points in the spectral data; Mark the color data of the i-th designated area as... Using undyed fibers as a reference, a color correction factor for the current color data is calculated through comparison. The formula is as follows: ; in, C represents the color value correction factor; C represents the color data of the undyed fiber. This indicates the wavelength data of visible light corresponding to the color of undyed fibers; This indicates the wavelength data of visible light corresponding to the color of undyed fibers; This indicates the correlation between the color value correction coefficient and the corresponding wavelength data; spectral data and color value correction factor Substitute into the calculation formula In the process, corrected spectral data are obtained. ; NIR spectra are generated online based on the corrected spectral data, and the fibers in waste fabrics are classified and identified using a piecewise regression model based on principal component analysis and least squares method. The composition of fibers in the set area is obtained and transmitted to the data prediction module.

2. The fully automated waste textile identification and sorting control system according to claim 1, characterized in that, The steps for preprocessing raw spectral data include: Step 1: Use data augmentation algorithms to augment the original spectral data, enhancing the differences in spectral data between different fabric fibers; Step 2: Smooth the data to eliminate random noise during the data acquisition process; Step 3: Perform derivative processing on the above data to resolve the baseline offset problem; Step 4: Perform multivariate scattering correction on the spectral data to improve the signal-to-noise ratio of the spectrum.

3. The fully automated waste textile identification and sorting control system according to claim 1, characterized in that, When filtering feature spectral data, the vector length of each particle is set to the number of wavelength points in the original spectrum. The particle vector and the number of wavelength variables in the original spectrum are equal and correspond one-to-one. The feature extraction of the spectral data is based on a binary particle swarm optimization algorithm. The particle velocity is mapped using the sigmoid function, and the mapping result is used as a probability value. The probability value is compared with a probability threshold. If and only if the probability value is not lower than the probability threshold, it means that the wavelength point corresponding to the vector is selected, and the spectral data at that wavelength is the feature spectrum.

4. The fully automated waste textile identification and sorting control system according to claim 1, characterized in that, The data prediction module calculates the composition ratio of all fiber components and the weight of each component in the entire waste fabric based on the weight of the entire waste fabric, the area of ​​each region, and the composition of the components, and generates a data form to send to the control module.

5. The fully automated waste textile identification and sorting control system according to claim 1, characterized in that, The control module is equipped with a sorting strategy that controls sorting based on the type of waste fabric and the corresponding weight percentage of the components, or further sorting based on fiber fineness when the type and / or weight percentage of the components are the same.

Images