Automatic sorting system and method for different color down based on computer vision

Abstract

The application provides an automatic sorting system and method for different color down based on computer vision, and relates to the field of intelligent detection. The system extracts features from the images of raw down taken by an industrial camera to obtain color features of the raw down, strengthens the obtained color features of the raw down in space, optimizes the features after the space strengthening, and then outputs the position coordinate information of the different color down through a generator based on the optimized features. Based on the position coordinate information, a jet pump is started, and the different color down corresponding to the coordinate points is sucked out by the vacuum generated by the suction nozzle of the jet pump. In this way, automatic sorting of different color down can be realized, and the sorting efficiency and quality are greatly improved.

Description

Technical Field

[0001] The present application relates to the field of intelligent detection, and more specifically, to a system and method for automatically sorting different-colored down based on computer vision. Background Art

[0002] The market demand for pure white down is high. Although down resources are abundant, the high content of mismatched down in raw white down far exceeds market demand. Currently, sorting mismatched down can only be done manually. This method is labor-intensive and inefficient. Due to worker fatigue and subjective factors, sorting quality cannot be fundamentally guaranteed.

[0003] Therefore, an optimized computer vision-based automatic sorting solution for heterochromatic down is needed. Summary of the Invention

[0004] This application addresses the shortcomings of the existing technology and provides a computer vision-based automatic sorting system and method for heterochromatic down. This system extracts features from raw down images taken by an industrial camera to obtain raw down color features, spatially enhances the obtained raw down color features, and then optimizes the spatially enhanced features. The optimized features are then passed through a generator to output the position coordinate information of the heterochromatic down. Based on this position coordinate information, a jet pump is activated, and the vacuum generated by its suction nozzle sucks out the heterochromatic down at the corresponding coordinate point. This allows for automated heterochromatic down sorting, significantly improving sorting efficiency and quality.

[0005] A computer vision-based automatic sorting system for down of different colors, comprising:

[0006] An image acquisition module is used to obtain images of raw velvet entering the inspection site captured by an industrial camera;

[0007] An image preprocessing module, configured to preprocess the raw wool image to obtain a noise-reduced raw wool detection image;

[0008] A color feature encoding module is used to perform feature encoding on the noise-reduced raw wool detection image to obtain a raw wool color space enhancement feature map;

[0009] A position coordinate information generating module, configured to generate position coordinate information of different-colored down based on the color space enhancement feature map of the raw down;

[0010] The heterochromatic down suction control module is used to start the jet pump based on the heterochromatic down position coordinate information, and use the vacuum generated by its suction nozzle to suck out the heterochromatic down at the corresponding position coordinates.

[0011] A method for automatically sorting down feathers of different colors based on computer vision, comprising:

[0012] Acquire images of raw velvet entering the inspection site captured by an industrial camera;

[0013] Preprocessing the raw material wool image to obtain a noise-reduced raw material wool detection image;

[0014] Performing feature coding on the noise-reduced raw wool detection image to obtain a raw wool color space enhancement feature map;

[0015] Generating position coordinate information of different-colored down based on the raw down color space enhancement feature map;

[0016] Based on the coordinate information of the position of the heterochromatic down, the jet pump is turned on, and the vacuum generated by the suction nozzle is used to suck out the heterochromatic down at the corresponding position coordinates.

[0017] This application has significant technical effects due to the adoption of the above technical solutions:

[0018] This application provides a computer vision-based automatic down sorting system and method. This system extracts features from raw down images captured by an industrial camera to obtain color features. These features are then spatially enhanced, optimized, and then transmitted through a generator to output the coordinates of the locations of the down. Based on this coordinate information, a jet pump is activated, and the vacuum generated by its nozzle sucks out the down at the corresponding coordinates. This allows for automated down sorting, significantly improving both efficiency and quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The above and other purposes, features, and advantages of the present application will become more apparent through a more detailed description of the embodiments of the present application in conjunction with the accompanying drawings. The accompanying drawings are intended to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0020] Figure 1 The figure is a flowchart of a method for automatically sorting different-colored down based on computer vision according to an embodiment of the present application.

[0021] Figure 2 2 is an architectural diagram of a method for automatically sorting different-colored down based on computer vision according to an embodiment of the present application.

[0022] Figure 3 The present invention provides a flowchart for performing feature encoding on the noise-reduced raw down detection image to obtain a raw down color space enhancement feature map in the method for automatically sorting different-colored down based on computer vision according to an embodiment of the present application.

[0023] Figure 4 The present invention provides a flowchart for generating position coordinate information of different-colored down based on the color space enhancement feature map of the raw down in the method for automatically sorting different-colored down based on computer vision according to an embodiment of the present application.

[0024] Figure 5 The present invention provides a flowchart for performing depth correlation distribution correction based on depth anchoring features on the raw down color space enhancement feature map in the computer vision-based automatic sorting method for different-colored down according to an embodiment of the present application to obtain an optimized raw down color space enhancement feature map.

[0025] Figure 6 The present invention provides a flowchart for performing depth correlation distribution correction based on depth anchoring features on the raw down color space enhancement feature vector in the computer vision-based automatic sorting method for different-colored down according to an embodiment of the present application to obtain an optimized raw down color space enhancement feature vector.

[0026] Figure 7 This is a system block diagram of a computer vision-based automatic sorting system for different-colored down according to an embodiment of the present application.

[0027] Figure 8 is a block diagram of an electronic device 10 according to an embodiment of the present application. DETAILED DESCRIPTION

[0028] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described herein.

[0029] Traditionally, people rely on manual sorting of down with different colors. However, this method has many drawbacks. First, manual sorting is labor-intensive, low in efficiency, and leads to high costs. Second, due to worker fatigue and subjective factors, sorting quality is difficult to guarantee, with high error rates and even the possibility of serious quality problems, such as the inclusion of large amounts of different colors, which affects the overall quality of the product.

[0030] In recent years, deep learning and neural networks have been widely applied in fields such as computer vision, natural language processing, and text signal processing. Furthermore, deep learning and neural networks have demonstrated capabilities approaching or even surpassing human performance in areas such as image classification, object detection, semantic segmentation, and text translation. The development of deep learning and neural networks has provided new solutions and approaches for the automated sorting of heterochromatic down feathers based on computer vision.

[0031] Figure 1The figure is a flowchart of a method for automatically sorting different-colored down based on computer vision according to an embodiment of the present application. Figure 2 FIG is an architecture diagram of a method for automatically sorting down feathers of different colors based on computer vision according to an embodiment of the present application. Figure 1 and Figure 2 As shown, the automatic sorting method of heterochromatic down based on computer vision according to the embodiment of the present application includes: S110, obtaining an image of raw down entering the detection site taken by an industrial camera; S120, preprocessing the raw down image to obtain a denoised raw down detection image; S130, feature encoding the denoised raw down detection image to obtain a raw down color space enhancement feature map; S140, generating heterochromatic down position coordinate information based on the raw down color space enhancement feature map; S150, based on the heterochromatic down position coordinate information, starting the jet pump and using the vacuum generated by its suction nozzle to suck out the heterochromatic down at the corresponding position coordinates.

[0032] Specifically, the technical solution of this application first requires obtaining images of raw down entering the inspection site, captured by an industrial camera. It should be understood that computer vision technology processes and interprets image information by simulating the human visual system. To identify specific objects in an image (in this case, heterochromatic down feathers), a clear image input is required. The raw down image provides the raw data for visual analysis and processing. The raw down image contains rich color information, which can be used to help distinguish heterochromatic down from normal down. Subsequent feature extraction steps allow the heterochromatic down to be identified and processed. Compared to manual sorting, automated sorting systems based on computer vision can process large quantities of raw down quickly and accurately. Raw down images are captured by industrial cameras, which typically have higher resolution, higher frame rates, and better image quality, enabling them to capture clearer and more accurate images of the raw down. In contrast, traditional cameras may not meet these high-performance requirements. Furthermore, industrial cameras are designed and manufactured with greater emphasis on stability and reliability, enabling them to operate stably for extended periods in harsh industrial environments.

[0033] In practical applications, due to various factors (such as lighting, equipment performance, and transmission interference), the raw down images captured by industrial cameras may contain noise, which can affect image clarity and accuracy. To effectively remove noise from the original raw down images and improve image quality, this application processes the raw down images through an image denoiser to produce a de-noised raw down detection image. The presence of noise can interfere with image processing algorithms, reducing their robustness and accuracy. Noise reduction can reduce the impact of noise on the algorithm, improve its robustness and accuracy, and make the identification of heterochromatic down more accurate and reliable.

[0034] The essence of the image denoiser in this application is a denoiser based on an autoencoder. An autoencoder is an unsupervised learning algorithm used to learn the coded representation of data from input data. It consists of two parts: an encoder and a decoder. The encoder is responsible for compressing the input data into a low-dimensional coded representation, while the decoder is responsible for recovering the original data from this coded representation. The reason why the autoencoder can be used for image denoising is that the encoder can learn useful feature representations from noisy images, while the decoder can recover a noise-free image from this feature representation. During the training process, the autoencoder learns how to extract useful information from noisy images and remove noise, so it can be used for image denoising. Through the training and optimization of the autoencoder, a denoiser model that can effectively remove image noise can be obtained. In this application, by inputting a noisy raw velvet image, a noise-free raw velvet detection image can be obtained after denoising.

[0035] In an embodiment of the present application, one possible implementation method of passing the raw wool image through an image denoiser to obtain a denoised raw wool detection image may be: inputting the raw wool image into an encoder of the image denoiser, wherein the encoder uses a convolutional layer to perform explicit spatial encoding on the raw wool image to obtain raw wool features; inputting the raw wool features into a decoder of the image denoiser, wherein the decoder uses a deconvolution layer to perform deconvolution processing on the raw wool features to obtain the denoised raw wool detection image.

[0036] Figure 3 This is a flow chart of the method for automatically sorting down feathers with different colors based on computer vision according to an embodiment of the present application, in which the feature encoding of the noise-reduced raw down detection image is performed to obtain a raw down color space enhancement feature map. Figure 3 As shown, the feature encoding of the denoised raw wool detection image to obtain a raw wool color space enhancement feature map includes: S131, passing the denoised raw wool detection image through a raw wool color feature encoder to obtain a raw wool color detection feature map; S132, passing the raw wool color detection feature map through a spatial distribution feature enhancer to obtain the raw wool color space enhancement feature map.

[0037] The denoised raw down detection image contains key information such as the color and texture of the down, which is crucial for the identification and sorting of different-colored down. Specifically, the denoised image removes noise, making the color, texture and other features of the down clearer and more accurate. In order to extract relevant features from the denoised image, it is necessary to pass the denoised raw down detection image through a raw down color feature encoder to obtain a raw down color detection feature map. The essence of the raw down color feature encoder in this application is a convolutional neural network model as a feature extractor. Those of ordinary skill in the art should know that the convolutional neural network model (CNN) has excellent performance in the field of image processing. The convolution layer of the model in the convolutional neural network adopts a local connection method, which means that each neuron is only connected to a small area in the input image. This local connection can effectively extract local features in the image, including color features. In addition, the convolutional neural network model also has invariance such as translation, rotation and scaling, which makes it more robust to changes in position, angle and size in the down image.

[0038] In an embodiment of the present application, a possible implementation method of passing the denoised raw wool detection image through a raw wool color feature encoder to obtain a raw wool color detection feature map may be: each layer of the raw wool color feature encoder performs the following on the input data during the forward transfer process of the layer: convolution processing is performed on the input data based on a two-dimensional convolution kernel to obtain a convolution feature map; mean pooling processing is performed on the convolution feature map based on a local feature matrix to obtain a pooling feature map; and nonlinear activation is performed on the pooling feature map to obtain an activation feature map; wherein the output of the last layer of the raw wool color feature encoder is the raw wool color detection feature map, and the input of the first layer of the raw wool color feature encoder is the denoised raw wool detection image.

[0039] Although the raw down color detection feature map can capture the basic color information of the raw down, it may lack spatial sensitivity to color distribution and emphasis on heterochromatic areas. In order to further extract and strengthen the distribution characteristics of colors in space, especially those areas with significant differences in color from the surrounding areas (i.e., heterochromatic down), the raw down color detection feature map is passed through a spatial distribution feature enhancer in this application to obtain a raw down color space enhancement feature map. Through the spatial distribution feature enhancer, it is possible to more accurately locate areas with abnormal colors, i.e., heterochromatic down, thereby improving the accuracy of sorting. The essence of the spatial distribution feature enhancer in this application is a convolutional neural network model using a spatial attention mechanism. The spatial attention mechanism can guide the model to focus on specific areas in the input data, i.e., those areas related to the task. In the task of heterochromatic down sorting, areas with large color differences from the surrounding areas (heterogeneous down) are the objects that need to be focused on. The convolutional neural network model using the spatial attention mechanism can adaptively learn which areas are more important for identifying heterochromatic down and enhance these areas in the feature map. By enhancing the distribution characteristics of colors in space, the model can more accurately locate the position of heterochromatic down. It is also worth mentioning that since the model can adaptively learn which areas are important, it can focus on these areas more efficiently when processing data, thereby reducing unnecessary calculations.

[0040] In an embodiment of the present application, a possible implementation method of passing the raw velvet color detection feature map through a spatial distribution feature enhancer to obtain a raw velvet color space enhancement feature map may be: using each layer of the spatial distribution feature enhancer to perform the following on the input data in the forward pass of the layer: performing convolution processing based on a convolution kernel on the input data to obtain a convolution feature map; passing the convolution feature map through a spatial attention unit to obtain a spatial attention map; calculating the positional multiplication of the convolution feature map and the spatial attention map to obtain a spatial attention feature map; inputting the spatial attention feature map into a nonlinear activation unit to obtain an activation feature map; wherein the input of the first layer of the spatial distribution feature enhancer is the raw velvet color detection feature map, and the output of the last layer of the spatial distribution feature enhancer is the raw velvet color space enhancement feature map.

[0041] Figure 4 This is a flow chart of generating the coordinate information of the position of different-colored down based on the color space enhancement feature map of the raw down in the method for automatically sorting different-colored down based on computer vision according to an embodiment of the present application. Figure 4As shown, the method of generating the position coordinate information of different-colored down based on the raw down color space enhancement feature map includes: S141, performing depth correlation distribution correction based on depth anchoring features on the raw down color space enhancement feature map to obtain an optimized raw down color space enhancement feature map; S142, passing the optimized raw down color space enhancement feature map through a position information generator to obtain a generation result, and the generation result is the position coordinate information of different-colored down.

[0042] In particular, in the technical solution of this application, the raw down color space enhancement feature map is a key step in identifying and locating off-color down, used to guide the jet pump's suction nozzle to accurately remove unwanted off-color down. Improving the feature saliency of this feature map and avoiding feature redundancy are crucial to ensuring the accuracy and efficiency of the down detection and cleaning process. Specifically, prominent features help more accurately identify off-color down, ensuring that only unwanted down is removed and avoiding accidental aspiration. Distinctive features can improve the system's ability to identify off-color down, especially when down colors are similar or when ambient lighting is complex. Distinctive features help the system maintain stable performance under varying shooting conditions and varying raw down conditions. Furthermore, removing redundant features reduces processing time, improves system response speed, and speeds up the cleaning process. Redundant features can degrade system performance and affect identification and cleaning accuracy. Avoiding feature redundancy helps maintain high system performance. Avoiding feature redundancy helps more efficiently utilize computing and storage resources, reducing system operating costs. Furthermore, distinctive features facilitate precise cleaning, ensuring that only off-color down is removed and minimizing damage to qualified down. Avoiding feature redundancy can reduce unnecessary calculations, speed up the cleaning process, and improve production efficiency. Distinctive features and avoiding redundancy help reduce jet pump malfunctions caused by misidentification and improve system reliability. Based on this, the raw velvet color space enhancement feature map is corrected using a depth-correlated distribution correction based on depth anchoring features to obtain an optimized raw velvet color space enhancement feature map.

[0043] Specifically, Figure 5 The present invention provides a flowchart of the method for automatically sorting down feathers with different colors based on computer vision according to an embodiment of the present application, wherein the depth correlation distribution correction based on the depth anchor feature is performed on the color space enhancement feature map of the raw down to obtain an optimized color space enhancement feature map of the raw down. Figure 5As shown, the depth correlation distribution correction based on the depth anchoring feature of the raw velvet color space enhancement feature map to obtain the optimized raw velvet color space enhancement feature map includes: S1411, feature flattening the raw velvet color space enhancement feature map to obtain a raw velvet color space enhancement feature vector; S1412, depth correlation distribution correction based on the depth anchoring feature of the raw velvet color space enhancement feature vector to obtain an optimized raw velvet color space enhancement feature vector; S1413, reverse feature aggregation of the optimized raw velvet color space enhancement feature vector in accordance with the feature flattening method to obtain the optimized raw velvet color space enhancement feature map.

[0044] More specifically, Figure 6 The present invention provides a flowchart of performing depth correlation distribution correction based on depth anchoring features on the raw down color space enhancement feature vector in the method for automatically sorting different-colored down based on computer vision according to an embodiment of the present application to obtain an optimized raw down color space enhancement feature vector. Figure 6 As shown, the depth correlation distribution correction based on the depth anchoring feature of the raw velvet color space enhancement feature vector is performed to obtain the optimized raw velvet color space enhancement feature vector, including: S14121, determining the mean value of the raw velvet color space enhancement feature vector; S14122, accumulating and summing the raw velvet color space enhancement feature vectors to obtain the total value of the raw velvet color space enhancement feature vectors; S14123, subtracting the mean value of the raw velvet color space enhancement feature vector from the total value of the raw velvet color space enhancement feature vector, dividing the result by a predetermined hyperparameter, and then dividing the result by the length of the raw velvet color space enhancement feature vector to obtain a first quotient value; S14124, calculating the exponential function value with a natural constant as the base for the first quotient value to obtain a first natural exponential function value; S14125, multiplying the first natural exponential function value by the raw velvet color space enhancement feature vector by position point to obtain the optimized raw velvet color space enhancement feature vector.

[0045] In an embodiment of the present application, another possible implementation method of performing a depth correlation distribution correction based on a depth anchor feature on the raw velvet color space enhancement feature vector to obtain an optimized raw velvet color space enhancement feature vector may be: performing a depth correlation distribution correction based on a depth anchor feature on the raw velvet color space enhancement feature vector using the following processing formula to obtain an optimized raw velvet color space enhancement feature vector, wherein the processing formula is:

[0046]

[0047] Among them, V1 represents the color space enhancement feature vector of the raw material velvet, v irepresents the eigenvalue of the ith position of the raw material velvet color space enhancement feature vector, represents the mean of the raw velvet color space enhancement feature vector, L represents the length of the raw velvet color space enhancement feature vector, ε represents a predetermined hyperparameter, ⊙ represents position point multiplication, and V1′ represents the optimized raw velvet color space enhancement feature vector.

[0048] It should be well understood that in order to improve the feature significance of the raw velvet color space enhancement feature map and avoid feature redundancy, in the technical solution of the present application, the raw velvet color space enhancement feature map is subjected to a depth correlation distribution correction based on the depth anchor feature, which uses the global mean of the raw velvet color space enhancement feature map as its dynamically formed depth anchor feature, and applies a coherence attention response mechanism based on the depth anchor feature to the original feature distribution, so as to perform position scattering response depth correlation correction in the class divergence space with depth difference, so that the original feature distribution has a highly significant depth correlation distribution characteristic, so as to improve the feature significance of the raw velvet color space enhancement feature map and avoid feature redundancy.

[0049] Next, the optimized raw material down color space enhancement feature map is passed through a position information generator to obtain a generation result, which is the position coordinate information of the heterogeneous down. It should be well understood that although the optimized raw material down color space enhancement feature map can highlight the color and spatial distribution characteristics of the heterogeneous down, it is not sufficient to directly guide the equipment to perform heterogeneous down sorting operations. The role of the position information generator is to convert the key information in the feature map into specific position coordinate information, so that the equipment can accurately find and separate the heterogeneous down according to this coordinate information. The position information generator is usually a deep learning model, which can receive the optimized raw material down color space enhancement feature map as input, and output the position coordinate information of the heterogeneous down through a series of algorithm processing.

[0050] Finally, based on the coordinate information of the different-colored down locations, the jet pump is activated, and the vacuum generated by its suction nozzle is used to suck out the different-colored down at the corresponding coordinates. A jet pump is a device that uses the jet action of a fluid (liquid or gas) to generate negative or high pressure, thereby achieving extraction, mixing, or pressurization. In the scenario of automatic sorting of different-colored down, the main function of the jet pump is to accurately and efficiently suck out the different-colored down at specific coordinates through the vacuum generated by its suction nozzle. The operating principle of the jet pump is based on the Bernoulli equation and the principle of continuity. When high-pressure working fluid is ejected at high speed through the nozzle, the pressure around the nozzle decreases, forming a low-pressure area or vacuum region. This low-pressure area causes the surrounding fluid to be drawn into the jet pump. Subsequently, the sucked-in fluid mixes and exchanges energy with the working fluid in the throat section of the jet pump. As the mixed fluid passes through the diffuser, its velocity gradually decreases, and its kinetic energy is converted into pressure energy, and it is finally discharged through the discharge pipe.

[0051] In the technical solution of this application, once the coordinates of the down feathers are determined, the system sends a command to the jet pump. Based on the command, the jet pump adjusts the direction and position of its nozzle to align with the coordinates of the down feathers. The jet pump is then turned on, and high-pressure working fluid is ejected through the nozzle at high speed, creating a low-pressure zone that draws the down feathers into the jet pump. In the throat section, the drawn-in down feathers mix and exchange energy with the working fluid before being discharged through the diffuser, completing the extraction process.

[0052] In summary, the computer vision-based automatic sorting method for heterochromatic down feathers in the embodiments of the present application is explained. This method extracts features from raw down feather images taken by an industrial camera to obtain raw down feather color features, spatially enhances the obtained raw down feather color features, and then optimizes the spatially enhanced features. The optimized features are then passed through a generator to output the position coordinate information of the heterochromatic down feathers. Based on this position coordinate information, a jet pump is activated, and the vacuum generated by its suction nozzle is used to suck out the heterochromatic down feathers at the corresponding coordinate points. In this way, automated heterochromatic down feather sorting can be achieved, greatly improving sorting efficiency and quality.

[0053] Figure 7 FIG is a system block diagram of a computer vision-based automatic sorting system for down feathers of different colors according to an embodiment of the present application. Figure 7 As shown, according to an embodiment of the present application, a computer vision-based automatic sorting system 100 for heterochromatic down includes: an image acquisition module 110 for acquiring an image of raw down entering the detection site taken by an industrial camera; an image preprocessing module 120 for preprocessing the raw down image to obtain a noise-reduced raw down detection image; a color feature encoding module 130 for feature encoding the noise-reduced raw down detection image to obtain a raw down color space enhancement feature map; a position coordinate information generation module 140 for generating heterochromatic down position coordinate information based on the raw down color space enhancement feature map; and a heterochromatic down suction control module 150 for starting a jet pump based on the heterochromatic down position coordinate information, and using the vacuum generated by its suction nozzle to suck out the heterochromatic down at the corresponding position coordinates.

[0054] Here, those skilled in the art will appreciate that the specific functions and operations of the various units and modules in the above-mentioned computer vision-based automatic sorting system for different-color down 100 have been described in detail above. Figures 1 to 6 The description of the automatic sorting method of different-colored down based on computer vision has been introduced in detail, and therefore, its repeated description will be omitted.

[0055] In summary, the computer vision-based automatic sorting system 100 for heterochromatic down feathers according to the embodiment of the present application is described. It extracts features from raw down feather images taken by an industrial camera to obtain raw down feather color features, spatially enhances the obtained raw down feather color features, and then optimizes the spatially enhanced features. The optimized features are then passed through a generator to output the position coordinate information of the heterochromatic down feathers. Based on the position coordinate information, the jet pump is turned on, and the vacuum generated by its suction nozzle is used to suck out the heterochromatic down feathers at the corresponding coordinate points. In this way, automated heterochromatic down feather sorting can be achieved, greatly improving sorting efficiency and quality.

[0056] As described above, the computer vision-based automatic sorting system for down and feathers with different colors 100 according to the embodiments of the present application can be implemented in various wireless terminals, such as a server for computer vision-based automatic sorting of down and feathers with different colors. In one example, the computer vision-based automatic sorting system for down and feathers with different colors 100 according to the embodiments of the present application can be integrated into a wireless terminal as a software module and / or a hardware module. For example, the computer vision-based automatic sorting system for down and feathers with different colors 100 can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the computer vision-based automatic sorting system for down and feathers with different colors 100 can also be one of the many hardware modules of the wireless terminal.

[0057] Alternatively, in another example, the computer vision-based automatic sorting system for different-colored down 100 and the wireless terminal may be separate devices, and the computer vision-based automatic sorting system for different-colored down 100 may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0058] Below, reference Figure 8 To describe the electronic device according to the embodiment of the present application. Figure 8 1 is a block diagram of an electronic device 10 according to an embodiment of the present application. For example, the electronic device 10 can be provided as a server or a terminal device. Figure 8 The electronic device 10 includes a processing component 11, which further includes one or more processors, and a memory resource represented by a memory 12 for storing instructions executable by the processing component 11, such as an application. The application stored in the memory 12 may include one or more modules, each corresponding to a set of instructions. In addition, the processing component 11 is configured to execute the instructions to perform the above-described method.

[0059] The electronic device 10 may further include a power supply assembly 13 configured to perform power management of the electronic device 10, a wired or wireless network interface 14 configured to connect the electronic device 10 to a network, and an input / output (I / O) interface 15. The electronic device 10 may operate based on an operating system stored in the memory 12, such as Windows Server™, Mac OSX™, Unix™, Linux™, FreeBSD™, or the like.

[0060] In an embodiment of the present application, a computer-readable storage medium is further provided, such as a memory 12 including computer program instructions. The computer program instructions can be executed by the processing component 11 of the electronic device 10 to complete the above method.

[0061] The present application may be a system, method and / or computer program product. The computer program product may include a computer-readable storage medium carrying computer program instructions for causing a processor to implement various aspects of the present application.

[0062] A computer-readable storage medium can be a tangible device that can hold and store instructions for use by an instruction execution device. A computer-readable storage medium can be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: 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), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, such as a punch card or a raised structure in a groove on which instructions are stored, and any suitable combination thereof. As used herein, a computer-readable storage medium is not to be construed as a transient signal per se, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagating through a waveguide or other transmission medium (e.g., a light pulse through a fiber optic cable), or an electrical signal transmitted through an electrical wire.

[0063] The computer program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer program instructions from the network and forwards the computer program instructions to be stored in the computer-readable storage medium in the respective computing / processing device.

[0064] The computer program instructions for performing the operation of the present application can be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, and conventional procedural programming languages ​​such as "C" language or similar programming languages. The computer program instructions can be executed entirely on the user's computer, partially on the user's computer, executed as an independent software package, partially on the user's computer and partially on a remote computer, or completely on a remote computer or server. In the case of a remote computer, the remote computer can 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 can be connected to an external computer (such as an Internet service provider connected via the Internet). In some embodiments, by utilizing the state information of the computer program instructions to personalize an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA) or a programmable logic array (PLA), the electronic circuit can execute the computer program instructions, thereby realizing various aspects of the present application.

[0065] Various aspects of the present application are described herein with reference to flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present application. It should be understood that each block of the flowcharts and / or block diagrams, and combinations of blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions.

[0066] These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine, so that when these instructions are executed by the processor of the computer or other programmable data processing device, a device is generated to implement the functions / actions specified in one or more blocks in the flowchart and / or block diagram. These computer program instructions can also be stored in a computer-readable storage medium, where these instructions cause the computer, programmable data processing device, and / or other equipment to operate in a specific manner. Thus, the computer-readable medium storing the instructions comprises an article of manufacture, which includes instructions for implementing various aspects of the functions / actions specified in one or more blocks in the flowchart and / or block diagram.

[0067] Computer program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operational steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to implement the functions / actions specified in one or more blocks in the flowchart and / or block diagram.

[0068] For those skilled in the art, it is obvious that the present application is not limited to the details of the above-mentioned exemplary embodiments, and the present application can be implemented in other specific forms without departing from the spirit or basic characteristics of the present application. The scope of the present application is defined by the appended claims rather than the above description, and it is intended that all changes that fall within the meaning and scope of the equivalent elements of the claims are included in the present application, and any associated figure marks in the claims should not be regarded as limiting the claims involved. Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application is described in detail with reference to the preferred embodiments, ordinary technicians in this field should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit of the technical solutions of the present application.

Claims

1. A method for automatically sorting down feathers of different colors based on computer vision, characterized in that: include: Acquire images of raw velvet entering the inspection site captured by an industrial camera; Preprocessing the raw material wool image to obtain a noise-reduced raw material wool detection image; Performing feature coding on the noise-reduced raw wool detection image to obtain a raw wool color space enhancement feature map; Generating position coordinate information of different-colored down based on the raw down color space enhancement feature map; Based on the coordinate information of the different-colored down feather positions, the jet pump is turned on to suck out the different-colored down feathers at the corresponding coordinates using the vacuum generated by the jet pump nozzle; The process of generating the position coordinate information of down feathers of different colors based on the color space enhancement feature map of the raw down comprises: Performing depth correlation distribution correction based on depth anchoring features on the raw velvet color space enhancement feature map to obtain an optimized raw velvet color space enhancement feature map; The optimized raw down color space enhancement feature map is passed through a position information generator to obtain a generation result, wherein the generation result is the position coordinate information of the different-colored down; The process of performing depth correlation distribution correction based on depth anchoring features on the raw velvet color space enhancement feature map to obtain an optimized raw velvet color space enhancement feature map includes: Flattening the raw velvet color space enhancement feature map to obtain a raw velvet color space enhancement feature vector; Performing depth correlation distribution correction on the raw velvet color space enhancement feature vector based on the depth anchor feature to obtain an optimized raw velvet color space enhancement feature vector; Performing reverse feature aggregation on the optimized raw material velvet color space enhancement feature vector according to the feature flattening method to obtain the optimized raw material velvet color space enhancement feature map; The process of performing depth correlation distribution correction based on depth anchoring features on the color space enhancement feature vector of the raw velvet to obtain an optimized color space enhancement feature vector of the raw velvet includes: Determining the mean value of the color space enhancement feature vector of the raw velvet; Accumulating and summing the color space enhancement feature vectors of the raw velvet to obtain a total value of the color space enhancement feature vectors of the raw velvet; subtracting a mean value of the raw velvet color space enhancement feature vector from the sum of the raw velvet color space enhancement feature vectors, dividing the sum by a predetermined hyperparameter, and then dividing the sum by the length of the raw velvet color space enhancement feature vector to obtain a first quotient; Calculating an exponential function value with a natural constant as a base on the first quotient to obtain a first natural exponential function value; The first natural exponential function value is multiplied by the raw material velvet color space enhancement feature vector according to position points to obtain the optimized raw material velvet color space enhancement feature vector.

2. The method for automatically sorting down feathers of different colors based on computer vision according to claim 1, characterized in that: Preprocessing the raw pile image to obtain a noise-reduced raw pile detection image includes: passing the raw pile image through an image noise reducer to obtain the noise-reduced raw pile detection image.

3. The method for automatically sorting down feathers of different colors based on computer vision according to claim 2, characterized in that: Performing feature coding on the noise-reduced raw wool detection image to obtain a raw wool color space enhancement feature map, including: Passing the noise-reduced raw wool detection image through a raw wool color feature encoder to obtain a raw wool color detection feature map; The raw material wool color detection feature map is passed through a spatial distribution feature enhancer to obtain the raw material wool color space enhancement feature map.

4. The method for automatically sorting down feathers of different colors based on computer vision according to claim 3, characterized in that: The image denoiser is an auto-codec-based denoiser, the raw material velvet colour feature encoder is a convolutional neural network model serving as a feature extractor, and the spatial distribution feature enhancer is a convolutional neural network model using a spatial attention mechanism.

5. A computer vision-based automatic sorting system for down of different colors, using the computer vision-based automatic sorting method for down of different colors according to claim 1, characterized in that: include: An image acquisition module is used to obtain images of raw velvet entering the inspection site captured by an industrial camera; An image preprocessing module, configured to preprocess the raw wool image to obtain a noise-reduced raw wool detection image; A color feature encoding module is used to perform feature encoding on the noise-reduced raw wool detection image to obtain a raw wool color space enhancement feature map; A position coordinate information generating module, configured to generate position coordinate information of different-colored down based on the color space enhancement feature map of the raw down; The heterochromatic down suction control module is used to start the jet pump based on the heterochromatic down position coordinate information, and use the vacuum generated by its suction nozzle to suck out the heterochromatic down at the corresponding position coordinates.

6. The computer vision-based automatic sorting system for different-colored down according to claim 5, characterized in that: The color feature coding module includes: a raw material wool color feature generating unit, configured to pass the noise-reduced raw material wool detection image through a raw material wool color feature encoder to obtain a raw material wool color detection feature map; The raw material velvet color space enhancement feature generation unit is used to pass the raw material velvet color detection feature map through a spatial distribution feature enhancer to obtain the raw material velvet color space enhancement feature map.

Images