An automatic detection system, method and remote monitoring platform for particles

Abstract

The application provides an automatic detection system, method and remote monitoring platform for particles, wherein a particle conveying pipeline of the automatic detection system is provided with a particle detection pipeline and a light source access pipeline; a particle detection device is installed on the particle detection pipeline, and a light source supplementing device is installed on the light source access pipeline; when receiving a particle detection instruction to be executed, a control host first starts the light source supplementing device and the particle detection device, so that the particle detection device collects image data of flowing particles in the particle detection pipeline under the light irradiation of the light source supplementing device; secondly, when the particles in the collected image are processed and it is identified that there are abnormal particles, the abnormal particles are removed; and then the length-diameter ratio of a target particle in a two-dimensional plane is calculated, and according to the length-diameter ratio and a relationship model of the lengths of existing particles and the average length-diameter ratio, the length of the target particle in a three-dimensional space is determined. Therefore, the application can improve product quality consistency, reduce production cost and ensure production safety.

Description

Technical Field

[0001] This application relates to the field of chemical product manufacturing technology, and in particular to an automatic detection system, method and remote monitoring platform for particles. Background Technology

[0002] In certain chemical manufacturing industries, regular particulate matter serves as a core intermediate or final product, and its geometric dimensions (especially length and particle size distribution) directly determine the product's processing performance, production stability, and final application efficiency. For example, in key processes such as polymer granulation and fine chemical powder synthesis, the length uniformity and particle size distribution range of regular particulate matter directly affect the smoothness of subsequent molding and processing, the consistency of product mechanical properties, and key indicators such as dissolution rate and reactivity in end-application scenarios. Therefore, accurate and efficient detection of the geometric dimensions of regular particulate matter is a crucial step in ensuring the quality of such chemical products and a key requirement for achieving refined control of the production process and enhancing industry competitiveness.

[0003] However, the traditional quality inspection system currently widely used in this industry has two fundamental flaws in practical application, which seriously restrict the improvement of product quality and the optimization of production efficiency, as follows: First, the testing process is highly dependent on manual labor and carries significant safety risks. Traditional testing methods rely on operators in special process environments (often accompanied by high temperatures, high pressures, dust, or toxic and harmful gases) to obtain small samples through timed sampling, and then take the samples off the production line for offline measurement. On the one hand, harmful substances in special process environments can pose potential health threats to the respiratory system and skin of operators, significantly increasing production safety management costs. On the other hand, limited by the frequency and number of samples obtained through manual sampling, the obtained test samples have extremely limited scope, failing to comprehensively cover the product status in different areas and at different times of the production line. This makes it difficult for the test results to accurately reflect the overall, real-time operating status of the production line, easily leading to quality hazards and missed detections.

[0004] Secondly, quality control suffers from significant lag and is implemented in a crude manner. Traditional testing, based on sparse sampling and offline measurement, results in a substantial time lag between test data and the actual production process. This means that by the time quality deviations are detected, a large number of defective products have already been generated on the production line, leading to severe material waste and increased production costs. Furthermore, adjustments to production parameters for quality deviations (such as cutter speed and material conveying rate in granulation) rely heavily on operators' past experience, lacking accurate and quantifiable test data, resulting in insufficient accuracy and timeliness in parameter adjustments. This not only makes it difficult to guarantee the consistency of product dimensions and leads to significant quality fluctuations between batches, but also further exacerbates material losses during production, reduces production efficiency, and fails to meet the current demands of the chemical industry for refined production and high-quality development.

[0005] In summary, existing traditional quality inspection systems are no longer adequate for the precise control of the geometric dimensions of regular particulate matter required by specific chemical product manufacturing industries, and there is an urgent need for a technical solution that can achieve real-time, online, and accurate detection. Summary of the Invention

[0006] This application provides an automatic detection system, method, and remote monitoring platform for particles to address the shortcomings of existing technologies, such as reliance on manual labor, high safety risks, lagging quality control, and extensive methods. It provides data support for precise control of the production process, thereby improving product quality consistency, reducing production costs, and ensuring production safety.

[0007] The technical solution provided in this application includes: In a first aspect, embodiments of this application provide an automatic detection system for particles, the automatic detection system comprising: The particle conveying pipeline has a particle detection pipeline and a light source access pipeline running longitudinally through both sides of the set position. A particle detection device is installed on the particle detection pipeline; A light source supplementation device is installed on the light source inlet pipe; The control host, electrically connected to the particle detection device and the light source supplementation device, is used to control the light source supplementation device and the particle detection device to turn on when a particle detection command is received, so that the particle detection device can take pictures of particles passing through the particle detection pipe under the illumination of the light source supplementation device. If abnormal particles are identified in the captured particle pictures, the abnormal particles are removed. For each filtered target particle, the aspect ratio of the target particle in the two-dimensional plane is calculated. Based on the aspect ratio and the relationship model between the actual length and average aspect ratio of existing particles in three-dimensional space, the length of the target particle in three-dimensional space is determined.

[0008] Secondly, this application also provides an automatic detection method for particles, which is applied to the control host of the automatic detection system described in any embodiment of the first aspect embodiment. The automatic detection method includes: When a particle detection command is received, the light source supplement device and the particle detection device are turned on, so that the particle detection device can take pictures of the particles passing through the particle detection pipe when the light source supplement device emits light. If abnormal particles are found in the particle pictures, the abnormal particles are removed. For each target particle after filtration, calculate the aspect ratio of the target particle in a two-dimensional plane; Based on the aspect ratio and the relationship model between the actual length of existing particles in three-dimensional space and the average aspect ratio, the length of the target particle in three-dimensional space is determined.

[0009] Thirdly, this application also provides a remote monitoring platform for particles. The remote monitoring platform includes a remote monitoring management host, an automatic detection system as described in any embodiment of the first aspect, a wireless communication module, and a cutting device. The remote monitoring management host is connected to the automatic detection system and the cutting device respectively through the wireless communication module to send interactive information to the automatic detection system and the cutting device.

[0010] As can be seen from the above technical solutions, this application provides an automatic detection system, method, and remote monitoring platform for particles. The automatic detection system includes: a particle detection pipe and a light source access pipe running longitudinally through opposite sides of a particle conveying pipe at a set position; a particle detection device installed on the particle detection pipe; a light source supplement device installed on the light source access pipe; when the control host receives an instruction to perform particle detection, it controls the light source supplement device and the particle detection device to turn on, so that the particle detection device can take pictures of particles passing through the particle detection pipe under the illumination of the light source supplement device. If abnormal particles are found in the particle pictures, the abnormal particles are removed. For each filtered target particle, the aspect ratio of the target particle in the two-dimensional plane is calculated. Based on the aspect ratio and the relationship model between the actual length and average aspect ratio of existing particles in three-dimensional space, the length of the target particle in three-dimensional space is determined. As can be seen, the automatic detection system provided in this application embodiment has a simple structure and achieves real-time online detection of particles without relying on manual labor. It can estimate the length of each particle in the particle image in real time and determine the particle quality online in real time based on the estimated particle length, so as to achieve the purpose of timely and accurate control. It not only effectively reduces safety risks, but also avoids the lack of accuracy and timeliness of parameter adjustment caused by the lag and roughness of quality control. It provides data support for the precise control of the production process, thereby improving the consistency of product quality, reducing production costs, and ensuring production safety. Attached Figure Description

[0011] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0012] Figure 1 A schematic diagram of the structure of an automatic particle detection system provided in this application; Figure 2 A schematic diagram of the particle detection device provided in this application; Figure 3(a) is a schematic diagram of a particle containing foreign matter provided in this application; Figure 3(b) is a schematic diagram of the results of screening out foreign objects in Figure 3(a) according to this application; Figure 4 A schematic diagram of the test result output interface provided in this application; Figure 5 This application provides a flowchart illustrating an automatic particle detection method. Figure 6 A schematic diagram of an automatic particle detection device provided in this application; Figure 7 This is a schematic diagram of the structure of an electronic device provided in this application. Detailed Implementation

[0013] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the invention as detailed in the appended claims.

[0014] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The singular forms “a,” “the,” and “the” used in this invention and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0015] It should be understood that although the terms first, second, third, etc., may be used in this invention to describe various information, this information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, first information may also be referred to as second information without departing from the scope of this invention, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to a determination."

[0016] Firstly, see [the following] Figure 1 , Figure 1 The present application provides a schematic diagram of an automatic particle detection system, which includes: a particle conveying pipeline 1, a particle detection device 2, a light source supplementation device 3, and a control host.

[0017] The particle conveying pipe 1 has a particle detection pipe 11 and a light source access pipe 12 running longitudinally through opposite sides at a set position. The particle detection device 2 is installed on the particle detection pipe 11, and the light source supplementation device 3 is installed on the light source access pipe 12. The control host is electrically connected to the particle detection device 2 and the light source supplementation device 3. When a particle detection command is received, the control host will turn on the light source supplementation device 3 and the particle detection device 2 so that the particle detection device 2 can take pictures of the particles passing through the particle detection pipe 11 under the illumination of the light source supplementation device 3. If abnormal particles are found in the captured particle pictures, the abnormal particles are removed. For each filtered target particle, the aspect ratio of the target particle in the two-dimensional plane is calculated. Based on the aspect ratio and the established relationship model between the actual length and the average aspect ratio of the particle in the three-dimensional space, the length of the target particle in the three-dimensional space is determined.

[0018] In this embodiment, the particle conveying pipeline 1 is provided with a longitudinally penetrating particle detection pipeline 11 and a light source access pipeline 12 at predetermined positions on opposite sides along the length direction. The particle detection pipeline 11 and the light source access pipeline 12 are respectively provided with a particle detection device 2 and a light source supplementation device 3, so that the light source emitted by the light source supplementation device 3 can provide sufficient background light source for the particle detection device 2, so that the particle detection device 2 can clearly photograph the particles conveyed by the particle conveying pipeline 1.

[0019] In practical applications, considering the high-speed movement of chemical particles and the requirements for explosion-proof operation, this automatic detection system uses an industrial area array camera. The particle detection device 2 can be a camera equipped with a 50mm fixed-focus lens, and the light source supplementation device 3 can be a high-brightness LED backlight. This camera, by configuring an ultra-short exposure mode, effectively "freezes" the moment of particle movement, overcoming motion blur.

[0020] As an example, such as Figure 1 As shown, the particle detection device 2 includes a cavity housing 23 that is closed at one end and open at the other end, a camera mounting base 21, and a camera 22; the closed end of the cavity housing 23 is provided with a wire interface for electrical connection with the control host. The camera mounting base 21 is installed inside the cavity housing 23. The camera 22 is mounted on the camera mounting base 21 with its lens facing the open end of the cavity housing 23. The open end of the cavity housing 23 is connected to the particle detection pipe 11. There are various ways to connect the open end of the cavity housing 23 to the particle detection pipe, such as using a threaded connection, a flange connection, or a fixed welding method. This embodiment is not limited to any of these methods.

[0021] As another embodiment, such as Figure 2 As shown, the particle detection device 2 includes a cavity housing 23 with one end closed and the other open, a camera mounting base 21, an explosion-proof housing 24, and a camera 22. The closed end of the cavity housing 23 is provided with a wire interface for electrical connection with the control host. One end of the explosion-proof housing 24 is open and the other end is provided with a transparent explosion-proof window 31. The explosion-proof housing 24 is installed inside the cavity housing 23 with its open end abutting against the closed end of the cavity housing 23 and isolated from the cavity housing 23. The camera mounting base 21 is installed inside the explosion-proof housing 24, and the camera 22 is installed on the camera mounting base 21 with its lens facing the transparent explosion-proof window. The open end of the cavity housing 23 is connected to the particle detection pipeline 11. In this embodiment, since the camera 22 transmits particle image data to the control host at a high frequency and with high data packets, a direct wire connection is used for transmission to improve image transmission speed and stability. However, this embodiment is not limited to wireless transmission; the transmission method can be selected according to the actual application scenario. To prevent damage to the particle detection device 2, an explosion-proof housing 24 encloses the camera 22, providing explosion protection. In some embodiments, the particle detection device 2 can transmit high-definition images of 2448×2048 resolution to the control host in real time via gigabit Ethernet at a rate of 25 frames per second, providing a high-quality data source for subsequent data analysis and processing. As an example, the explosion-proof housing 24 is structurally adapted to the camera 22. The housing 24 contains an aluminum shell for easy heat dissipation and heat sinks that are fitted to the aluminum shell. Multiple heat dissipation holes facing the heat sinks are provided throughout the housing 24 and the aluminum shell, guiding the heat generated by the camera 22 to the heat sinks and dissipating it through the aluminum shell. This heat then dissipates through the heat dissipation holes, entering the sealed gap between the explosion-proof housing 24 and the hollow housing 23. For heat dissipation, such as... Figure 2As shown, the cavity housing 23 is further provided with a cold water vapor inlet 232 and a water vapor outlet 231 at opposite positions on the outer shell side for the entry of cold air or cold water. This allows the cold water vapor inlet 232 to be opened when heat dissipation is required, allowing cold air and / or cold water to enter the sealed gap. Heat is then carried away through the cold air and / or cold water from the water vapor outlet 231, thus achieving the purpose of heat dissipation for the camera 22. In this embodiment, the cold water vapor inlet 232 and the water vapor outlet 231 are located on opposite sides of the outer shell of the cavity housing 23, so that cold air or cold water enters the sealed gap and the heat flows out along with the cold air or cold water from the opposite water vapor outlet. In other embodiments, the cavity housing 23 is further provided with a gas inlet 234 and a transparent cover 14. An airflow cavity housing 26 for conveying compressed air extends outwards for a predetermined length from the side of the cavity housing 23 near the transparent explosion-proof window. A concave housing 27 extends from the edge of the airflow cavity housing 26 towards the outer edge of the transparent explosion-proof window, and a predetermined gap is reserved between the concave housing 27 and the outer side of the transparent explosion-proof window. The gas inlet 234 is located at the airflow cavity housing 26, allowing gas to enter the cavity formed by the concave housing 27 and the airflow cavity housing 26 through the gas inlet 234, be compressed, and then flow out through the gap to clear water mist from the transparent explosion-proof window 31. The transparent cover 14 is fitted to the end of the airflow cavity housing 26 to prevent liquid from entering the gap while allowing the camera to clearly capture particle images. In this embodiment, the gas inlet 234 and the cold water vapor inlet 232 are located on the same side.

[0022] As one embodiment, the gas inlet 234, the cold water vapor inlet 232, and the water vapor outlet 231 are each equipped with a cap, which is open during use. As another embodiment, the gas inlet 234, the cold water vapor inlet 232, and the water vapor outlet 231 are each equipped with a one-way valve 25, all of which are electrically connected to the control unit. In the cold water mode, where cooling is required by cold water, the one-way valve 25 at the cold water vapor inlet 232 and the one-way valve 25 at the water vapor outlet 231 are opened; otherwise, the one-way valve 25 at the cold water vapor inlet 232 and the one-way valve 25 at the water vapor outlet 231 are closed. In the cold air mode, where cooling is required by cold air, the valves at the gas inlet 234 and the water vapor outlet 231 are opened; otherwise, the one-way valve 25 at the gas inlet 234 and the valve at the water vapor outlet 231 are closed.

[0023] As an example, the transparent explosion-proof window 31 can be a transparent body made of explosion-proof material, which can be explosion-proof glass.

[0024] In another embodiment, a temperature sensor and a humidity sensor are installed inside the explosion-proof housing 24. The humidity sensor is located inside the transparent explosion-proof window 31 of the explosion-proof housing 24, but does not obstruct the edge of the transparent explosion-proof window 31. The temperature sensor is located inside the explosion-proof housing 24. Both the temperature sensor and the humidity sensor are electrically connected to the control host. When the temperature measured by the temperature sensor is higher than the upper limit of the temperature threshold, the one-way valve 25 at the cold water vapor inlet 232 is opened to cool the air through cold water or cold air mode. When the humidity measured by the humidity sensor is higher than the upper limit of the water mist humidity threshold, the one-way valve 25 corresponding to the gas inlet 234 is opened, so that the incoming gas forms an air curtain from the aforementioned gaps to remove water mist. In this way, the camera 22 can be quickly cooled while removing water mist, so that it can take clear pictures through the transparent explosion-proof window 31, improve the foreign object recognition rate, and thus improve the data accuracy.

[0025] It should be noted that in this embodiment, since the particles flow in the pipe with the water flow, and the particles in the water flow are usually not parallel to the observation plane, their posture is randomly distributed. The measured aspect ratio in two-dimensional space is usually smaller than that in three-dimensional space, and therefore the particle length cannot be directly estimated. Based on this, this embodiment establishes a relationship model between the aspect ratio in two-dimensional space and the particle length through a large number of experiments and simulations.

[0026] To improve image processing capabilities, the control host in this embodiment includes at least two processors: a CPU for reducing the sampling frequency to match the processing speed, and a GPU for high-speed image processing. In some embodiments, the CPU can place the acquired image data in a memory area directly accessible by the GPU.

[0027] Step 1: Acquire image data from camera 22 and load the image data into the CPU's general memory buffer, either as a NumPy array or a raw C buffer. Step 2: The inference thread (or before the data enters the queue) calls functions such as PyTorch's tensor.pin_memory() or the CUDA API to copy the cached image data to a pre-allocated paged memory area. In this step, once data enters the page-locked memory, its physical address is locked, preparing it for subsequent high-speed transfers. High-speed data transfer is where page-locked memory plays its most significant role, aiming to eliminate secondary copying during transmission.

[0028] Step 3: Transfer image data from the CPU's locked memory to the GPU's video memory; In this step, DMA takes over the transfer task, directly transferring data from the CPU's paged memory to the GPU's VRAM via the PCIe bus. The inference program initiates the transfer command; since the data resides in paged memory, the GPU's DMA (Direct Memory Access) controller is activated. DMA then takes over the transfer task, directly transferring data from the CPU's paged memory to the GPU's VRAM via the PCIe bus.

[0029] In this process, there is zero CPU intervention: the CPU hardly needs to participate in data copying, reducing the load. The non-blocking setting (non_blocking=True) allows the CPU to immediately execute subsequent CPU tasks (such as processing the next frame's data, reading the next frame's locked memory address, etc.) during DMA data transfer, achieving parallelism between CPU computation and data transfer.

[0030] Based on this, the GPU in this embodiment has inference and result processing capabilities, so that the GPU computing power can be fully utilized.

[0031] Step 4: After the data arrives at the GPU's video memory (VRAM), the image data is preprocessed, and the preprocessed image is input into the image recognition model that has been placed in the image data to identify the edges of objects in the image data, so as to output the inference result of the object contour in the image. Common preprocessing such as normalization, color space conversion, and image scaling are also completed on the GPU side (e.g., through CUDA / TensorRT).

[0032] Model execution: Image data is fed into the inference engine (such as TensorRT) to obtain the final inference result.

[0033] Step 5: The inference results are transferred from the GPU back to the CPU memory for business logic processing. Once the inference for this frame is complete, the paged memory block used to store the image data is immediately released, or more efficiently reused to receive the next frame of data captured by camera 22, avoiding the overhead of repeated memory allocation.

[0034] Regarding the image storage solution, as an example: memory passthrough – recommended for single-machine operation; in this example, after obtaining the image data captured by camera 22, it is directly stored in the memory queue. Image data is retrieved from the queue and directly sent to the GPU for inference, completely bypassing the hard drive. If it is necessary to retain the base image, a dedicated thread is started to handle retrieving images from the queue and writing them to the hard drive, utilizing the operating system's write cache to avoid blocking the inference thread.

[0035] In other embodiments, the open end of the cavity housing 23 is threadedly connected to the particle detection pipe 11, and the light source supplementation device 3 is threadedly connected to the light source access pipe 12. In this embodiment, the outer end of the open end of the cavity housing 23 is provided with an external thread, and the particle detection pipe 11 is provided with an internal thread; or, the open end port of the cavity housing 23 is provided with an internal thread, and the particle detection pipe 11 is provided with an external thread. Correspondingly, the light source supplementation device 3 is provided with an internal thread, and the light source access pipe 12 is provided with an external thread; or the light source supplementation device 3 is provided with an external thread, and the light source access pipe 12 is provided with an internal thread. Alternatively, two threaded pipes with external threads at both ends are provided, through which the open end of the cavity housing 23 is threadedly connected to the particle detection pipe 11, and the light source supplementation device 3 is threadedly connected to the light source access pipe 12.

[0036] In other embodiments, the automatic detection system further includes a clamping connector equipped with a reverse spring. One end of the clamping connector is detachably connected to a first mounting member connected to the particle detection device 2, and the other end is detachably connected to a second mounting member connected to the particle detection device 2. When particle detection is required, the clamping connector, under the action of the reverse spring, continuously and uniformly applies a clamping force to the light source supplementation device 3 and the particle detection device 2, so that the particle detection device 2 is firmly connected to the particle detection pipe 11, and the light source supplementation device 3 is firmly connected to the light source inlet pipe 12. In this embodiment, the first mounting member is named to distinguish it from subsequent mounting members and does not specifically refer to any particular mounting member. Correspondingly, the second mounting member is named to distinguish it from the aforementioned mounting members and does not specifically refer to any particular mounting member. The clamping connector is equipped with a reverse spring. When the clamping connector is unfolded, the reverse spring generates a reverse elastic force, i.e., a clamping force, and simultaneously applies equal clamping forces to both the first and second mounting members, so that the particle detection device 2 is firmly installed on the particle detection pipe 11 under the action of the clamping connector. As one embodiment, the clamping connector includes a first clamping member, a second clamping member, and a reverse spring. The first and second clamping members are movably connected in a cross configuration. One end of the reverse spring is connected to the mounting end inside the first clamping member, and the other end is connected to the mounting end inside the second clamping member. When the first and second clamping members are opened through the small opening end, the reverse spring generates a reverse elastic force and applies a clamping force to the large opening end of the first and second clamping members. In this embodiment, the first clamping member is named to distinguish it from subsequent mounting members and does not specifically refer to any particular clamping member. Similarly, the second clamping member is named to distinguish it from the aforementioned mounting members and does not specifically refer to any particular clamping member.

[0037] The control host can receive detection commands sent by remote third-party electronic devices indicating the execution of particle detection, or it can respond to user-triggered detection commands indicating the execution of particle detection, and execute the steps of controlling the activation of the light source supplementation device 3 and the particle detection device 2. After activation, the particle detection device 2, under the illumination emitted by the light source supplementation device 3, photographs the particles passing through the particle detection pipe 11, obtains particle images, and performs particle identification on the photographed particle images. If abnormal particles are identified in the particle images, the abnormal particles are removed, and filtered target particles are obtained. For each target particle, the aspect ratio of the target particle in the two-dimensional plane is calculated, and the aspect ratio is input into the existing relationship model between particle length and average aspect ratio, outputting the length of the target particle. Then, based on the length, it is determined whether the target particle meets the requirements or whether the cutting device needs to be adjusted. In some embodiments, referring to Figures 3(a) and (b), the control host includes a particle recognition unit for identifying abnormal particles in the particle image. The particle recognition unit performs morphological feature recognition on the particle contour based on contour analysis to obtain the particle contour feature parameters; it then determines whether the contour feature parameters belong to a range of uniform contour feature parameters used to characterize particles. If they do, the particle is determined to be a normal particle; otherwise, it is determined to be an abnormal particle. In this embodiment, after obtaining the initial segmentation mask, the automatic detection system does not directly calculate the size of all objects. Instead, it introduces a morphological feature filtering layer based on contour analysis, as shown in Figure 3(a), to accurately distinguish between target particles and abnormal particles. These abnormal particles can be foreign objects such as bubbles, water droplets, or imaging noise. By deeply analyzing the contour of each segmented object, its key morphological feature parameters are extracted. These feature parameters can effectively characterize the differences in shape characteristics between different categories of objects: target particles usually have relatively regular and consistent contour features, while abnormal particles (such as bubbles, water droplets, etc.) exhibit significantly different contour characteristics. This automatic detection system constructs a robust classification decision boundary through a preset feature threshold range, automatically identifies and removes non-target foreign objects, i.e., abnormal particles, as shown in Figure 3(b), ensuring the purity of the input data for subsequent calculations and fundamentally improving the accuracy and reliability of the measurement.

[0038] As an example, the constructed FastSAM model is used to perform morphological feature recognition on the particle contour to obtain the contour feature parameters of the particle. The FastSAM model has a morphological feature filtering layer and a feature recognition layer. This embodiment abandons the traditional and inefficient manual segmentation method and introduces the advanced FastSAM model. This model is built on the YOLOv8 architecture, and its advantages are: 1) High accuracy and zero-shot capability: Even without extensive training on specialized chemical particle data, it can achieve excellent zero-shot segmentation performance by leveraging general visual priors learned on large datasets (such as SA-1B).

[0039] 2) Real-time performance: Compared to its predecessor SAM (Segment Anything Model), FastSAM significantly reduces computational requirements while maintaining competitive segmentation accuracy. On an NVIDIA RTX 3090 graphics card, processing speed can be increased by tens of times, fully meeting the real-time requirements of industrial online inspection.

[0040] 3) Fully automatic segmentation: It can automatically output a precise pixel-level mask for each particle instance in the image without manual intervention.

[0041] As can be seen, the technical solution provided in this embodiment addresses the limitations of existing FastSAM models in recognizing multi-class targets with zero samples. The next stage will introduce a category-aware domain-adaptive technique. By finely labeling and specifically fine-tuning multi-class image data under different materials, lighting conditions, and scenes, the model's output layer and feature extractor will be reconstructed, enabling it not only to achieve pixel-level segmentation but also to accurately distinguish between various semantic categories such as target particles, bubbles, and impurities. This aims to build a robust visual perception core capable of accurate recognition and segmentation across media, scales, and categories, significantly improving the system's generalization ability and practical value in complex multi-target scenarios such as chemical and food processing.

[0042] In some embodiments, the control host includes an average aspect ratio determination unit for determining the average aspect ratio and an aspect ratio calculation unit for calculating the aspect ratio of the target particle in a two-dimensional plane. The average aspect ratio determination unit is used to obtain particle samples that have been abnormally removed within a set time period; for each particle sample, it calculates the minimum bounding rectangle of the particle sample, and calculates the aspect ratio of the particle sample in a two-dimensional plane based on the minimum bounding rectangle; and determines the average aspect ratio based on the aspect ratios of all particle samples calculated within the set time period and the number of particle samples.

[0043] The aspect ratio calculation unit is used to calculate the minimum bounding rectangle of the target particle, and to calculate the aspect ratio of the target particle in a two-dimensional plane based on the minimum bounding rectangle.

[0044] In this embodiment, directly estimating the length of a three-dimensional particle from a two-dimensional image is a typical "soft measurement" problem. This application innovatively proposes a solution based on aspect ratio and piecewise interpolation, specifically: Feature extraction: For each particle mask that passes through the filter, i.e., the particle sample, calculate the minimum bounding rectangle of the particle sample, and define the length-diameter ratio (LDR) as LDR = Major Axis Length / Minor Axis Length, and calculate the average length-diameter ratio ALDR within a set time period, ALDR = (sum of LDR of each particle sample within the set time period) / number of particle samples within the set time period.

[0045] Model Calibration: In the initial calibration phase of this automatic detection system, a large number of particle images and their corresponding manually measured lengths L at different times are acquired simultaneously, forming a calibration dataset {ALDR_i, L_i}. In this embodiment, the manually measured length refers to the actual length of the particle measured using a dedicated tool, where i represents the particle number.

[0046] In other embodiments, the control host further includes a relational model determination unit for constructing a relational model. The average aspect ratio determination unit is used to obtain the average aspect ratio of particle samples in particle images collected within a set time period; for each particle sample, the actual length of the particle sample at the corresponding time point is measured, and the actual average length of each particle sample within the set time period is calculated; based on the average aspect ratio and actual average length of each particle sample, a piecewise interpolation method is used to fit a mapping function from the aspect ratio of each particle sample to its corresponding actual length, and the mapping function is determined as the relational model between particle length and average aspect ratio.

[0047] In this embodiment, data analysis of the calibrated dataset reveals a significant nonlinear relationship between the average aspect ratio (ALAR) and the actual length L. Based on this, a piecewise interpolation method is used to accurately fit the mapping function L = f(LDR) from LDR to L, according to the ALAR and the actual average length. This method can better capture local variations in the relationship curve and has higher accuracy than simple linear regression or global polynomial fitting.

[0048] Based on this, after determining the aspect ratio and relationship model of the particle, the length of the particle can be estimated online according to the relationship model without the need for actual measurement of the particle, thus reducing the time and cost of measuring thousands of particles. Specifically, when the automatic detection system is running online, for each newly detected particle, it is only necessary to calculate the ALDR of the particle, and the length L of the particle can be estimated in real time through the established piecewise interpolation model, i.e., the relationship model f(ALDR).

[0049] In other embodiments, such as Figure 4As shown, the control host also includes a particle screening unit. This unit compares the length of each acquired target particle with a qualified length threshold. If the length of the target particle exceeds the upper limit of the qualified length threshold, it is marked as a large particle and its quantity is accumulated. If the length of the target particle is less than the lower limit of the qualified length threshold, it is marked as a small particle and its quantity is accumulated. If the length of the target particle falls within the qualified length threshold range, it is marked as a qualified particle and its quantity is accumulated. Based on the accumulated values ​​of large particles, small particles, and qualified particles, and the total number of target particles, the proportion of large particles, the proportion of small particles, and the particle qualification rate are determined.

[0050] In this embodiment, the acceptable length threshold range can be understood as a length lower limit greater than or equal to the acceptable length threshold, but less than the acceptable length upper limit. When the screened particles are within this acceptable length threshold range, it indicates that the particles meet the acceptable requirements; otherwise, the particles are marked accordingly. If the proportion of large particles is greater than the acceptable threshold for large particle proportion, it indicates that the current cutting speed of the industrial particles is too low, and too many large particles are being cut, which does not meet industrial requirements; if the proportion of small particles is greater than the acceptable threshold for small particle proportion, particle control information indicating a reduction in the cutting speed is sent to the cutting device used to cut the particles, so that the cutting device reduces the cutting speed of the particles according to the particle control information. Accordingly, if the particle acceptance rate is within the target acceptable threshold, it indicates that the particles are acceptable and meet the acceptable requirements, and feedback is sent to the cutting device to maintain the current cutting speed of the cutting device. In some embodiments, the control host further includes a cutting control unit, which is configured to send particle control information indicating an increase in the cutting speed to the cutting device for cutting particles if the proportion of large particles is above a qualified threshold, so that the cutting device increases the cutting speed of particles according to the particle control information; and to send particle control information indicating a decrease in the cutting speed to the cutting device for cutting particles if the proportion of small particles is above a qualified threshold, so that the cutting device decreases the cutting speed of particles according to the particle control information.

[0051] In this embodiment, the cutting blade in the cutting device can be made of laser or a solid blade; this embodiment is not limited to either.

[0052] As an example, the particle screening unit calculates the average length of each target particle within a set time interval. When the average length is greater than the upper limit threshold and the proportion of large particles is greater than the large particle proportion threshold, it sends particle control information to the cutting device for cutting the particles, indicating an increase of a first set percentage based on the current cutting speed. This first set percentage can be 10% or 5%. When the average length is within the target length range but the proportion of small particles is greater than the small particle proportion threshold (which can be 15%), it sends particle control information to the cutting device for cutting the particles, indicating a decrease of a second set percentage based on the current cutting speed (which can be 2%). When the average length is within the target length range but the distribution width is greater than the distribution width threshold, it sends control information to the cutting device for cutting the particles, indicating a fine adjustment of the current cutting speed to find the final distribution of particles. Based on the large particle proportion value and small particle proportion value corresponding to the next set time interval, it returns the calculation of the average length of each target particle within the set time interval until the average length is within the target length range but the distribution width is within the distribution width threshold. Here, fine adjustment can be understood as an adjustment of less than 2% to increase or decrease the current cutting speed. When the average length is greater than the upper limit threshold and the proportion of large particles is less than or equal to the large particle proportion threshold, or when the average length is within the target length range but the proportion of small particles is less than or equal to the small particle proportion threshold, particle control information indicating the maintenance of the current cutter speed is sent to the cutting device used to cut the particles. In this embodiment, the distribution width can be understood as the sum of the proportion of large particles and the proportion of small particles, and the distribution width threshold can be 20%.

[0053] As an example, the qualified length threshold is determined as follows: Based on the current time point, particle images taken within a set time interval are acquired. This set time interval is a continuous time period formed by superimposing a historical set time interval and a new time interval. The historical set time interval is greater than or equal to the new time interval. For example, if the time interval is 5 minutes, then the historical set time interval is 4 minutes, and the new time interval is 1 minute. The historical time interval belongs to both the previous historical set time interval and the current historical set time interval, while the new time interval belongs only to the current set time interval. If the current time point is 10:00, then the set time interval is 5 minutes, the historical time interval is the 4 minutes formed by 9:56 and 10:00, and the new time interval is the 1 minute formed by 10:00 and 10:01. The 4 minutes and 1 minute are then superimposed to form 5 minutes. In this way, in practical applications, statistical indicators can be displayed in real time in the form of a dashboard, providing a decision-making basis for closed-loop control. Continuous time points can be formed, which can improve the inspection accuracy and provide observers with continuous observation of the particle status. For example, by using a sliding time window to improve response speed, shooting time length 1 corresponds to 1 to 5 minutes of data, and shooting time length 2 corresponds to 2 to 6 minutes of data. As you can see, there is an overlap of 4 minutes between length 1 and length 2.

[0054] In this embodiment, at the system layer: detection is moved forward and dynamic process modeling is used to overcome time lag, fundamentally solving the control lag problem. This embodiment reconstructs the system architecture; as one example, the automatic detection system is moved forward to a position near the cutter exit in the cutting device. This significantly shortens the detection delay and provides near real-time feedback for the control strategy. Based on this, a high-order dynamic process model integrating material transport dynamics is established, and a model predictive control algorithm is used to achieve advanced and precise control of particle length. As one example, the obtained large particle ratio and small particle ratio values ​​are input into a specified model predictive control model to obtain a control value indicating whether control is needed. If the control value is greater than the control threshold, particle control information indicating that the current cutter speed needs to be adjusted is sent to the cutting device used to cut the particles; if the control value is less than or equal to the control threshold, it remains unchanged. In some embodiments, the predictive control model is U1 = PID_L(e_L) + K_Pl * sat(P_large-threshold_Pl) + K_Ps * sat(P_smal-threshold_Ps), where PID_L( . ) is the length control function, e_L is the length difference between the actual length and the desired length of the particle, K_Pl is the adjustment coefficient for large particles, and sat( .) is the saturation limiting function, P_large is the proportion of large particles, threshold_Pl is the threshold for the proportion of large particles, K_Ps is the adjustment coefficient for small particles, threshold_Ps is the threshold for the proportion of small particles, and U1 is the adjustment amount of the cutting speed in the cutting device. This embodiment mainly uses length control. When the proportion of large and small particles exceeds a certain threshold, an additional control amount is superimposed, targeting sat( . The function only takes effect when the corresponding threshold is exceeded.

[0055] In other embodiments, during the operation of the working mode, for each particle image, if the number of particles in the image is less than the particle threshold, the non-compliance index, indicating that the number of particles is below the standard, is incremented by 1. When the continuously accumulated non-compliance index is greater than or equal to the non-compliance threshold, a frequency adjustment command is sent to the camera to reduce the camera's shooting frequency, causing the camera to shoot images in a semi-dormant state (i.e., at a low frequency). The non-compliance index is reset to zero, and an alarm device is triggered to provide an alarm prompt. If the number of particles in the image is greater than or equal to the particle threshold, the non-compliance index is reset to zero, and the normal working mode continues. During the semi-dormant working mode, for each grain image, if the number of grains in the image is greater than or equal to a grain threshold, the compliance index, indicating that the number of grains meets the standard, is incremented by 1. When the cumulative compliance index exceeds the compliance threshold, a recovery frequency command is sent to the camera, indicating that the camera's initial frequency has been restored. This causes the camera to resume operation from the semi-dormant state and capture images at the initial frequency, while the compliance index is reset to zero. If the number of grains in the image is less than the grain threshold, the compliance index is reset to zero, and the semi-dormant working mode continues. The initial frequency is greater than the low frequency. In this embodiment, the non-compliance threshold refers to the number of consecutive grain images in which the number of grains is less than the grain threshold. The set number can be 10 frames, and the corresponding non-compliance threshold can be 10. In practical applications, the non-compliance threshold and the grain threshold can be set by the user according to the actual situation; this embodiment does not limit this. The particle threshold refers to the number of particles detected in each image, which is 4. For example, if the number of particles in 10 consecutive particle images is less than 4, the device (camera and control host) can enter a semi-sleep mode (i.e., reduce the camera frame rate and image inference frame rate to 0.2 FPS), while simultaneously alarming to confirm whether the camera and other equipment are malfunctioning, and resetting the non-compliance index to zero, and then re-counting new particle images. The compliance threshold refers to the number of consecutive particle images in which the number of particles is greater than or equal to the particle threshold. The set number can be 10 frames, and the corresponding compliance threshold can be 4. In practical applications, the compliance threshold can be set by the user according to the actual situation, and this embodiment is not limited to this. Based on this, in the semi-sleep mode, if the number of particles in 10 consecutive frames is greater than or equal to 4, it switches to the working state, and the compliance index is reset to zero. By intelligently starting and resting the device, the failure rate of the camera and industrial control computer can be reduced, energy can be saved, and the equipment life can be extended.

[0056] For the control layer of this automatic detection system: multi-source information fusion and intelligent collaborative decision-making, to achieve a higher level of optimized control, this embodiment will break through the single control variable and introduce the key operating condition parameter of "remaining material weight". By constructing a multivariate coupled model of cutter speed-material height-particle length, designing a multi-objective optimization function, and developing an optimal collaborative control strategy that can simultaneously take into account quality consistency, production efficiency, and raw material utilization, a complete production closed loop with intelligent decision-making capabilities will be formed.

[0057] As can be seen, by applying the technical solution provided in this application, the automatic detection system can achieve autonomous quality control. Through the Modbus TCP / IP industrial communication protocol, it establishes a real-time connection with the cutting device on the production line, comparing the calculated average length and the proportion of large particles with the set target requirements. Using the dead-zone compensation closed-loop control algorithm PID (Proportional-Integral-Derivative Controller), meaning that if the particle length exceeds the acceptable length threshold, the cutting speed is reduced, and feedforward compensation based on the statistical proportion of large and small particles is introduced, thus achieving predictive suppression of major production disturbances and effectively reducing the impact of time delay. Furthermore, the technical solution provided in this embodiment realizes an innovative path of deep integration between cutting-edge artificial intelligence, computer vision technology, and traditional industrial manufacturing. It is not merely a "replacement of the human eye," but rather, through a series of core technologies such as FastSAM segmentation, multimodal feature filtering, and soft measurement modeling, it constructs an intelligent system capable of "perception-understanding-decision-action." This technical solution successfully transforms quality control from an offline, sampling, and post-event remedial model to an online, full-scale, real-time prevention and control model, greatly improving the level of intelligence in chemical production, product quality consistency, and safety. It provides a highly valuable technology for the digital transformation of high-end manufacturing.

[0058] This application aims to upgrade the system from a single detection and control system to a comprehensive intelligent platform capable of sensing complex working conditions and making predictions and optimization decisions. In practical applications, it is necessary for the control host to "see" the future state. The Smith predictor: The control host has a built-in simulator that predicts that "if there is no delay, the cutting device can respond promptly to the particle adjustment information sent by the control host." Based on this predicted state, the cutting device can cut a large number of qualified particles. Therefore, the Smith predictor allows the control host to "see" the state τ seconds after the cutter adjustment (pipeline delay time), rather than the actual measured delay, thus enabling it to make correct decisions in advance. After multiple experimental verifications, the Smith predictor can predict the new particle length after a set time period = a × old particle length - b × control quantity, where a is the historical particle length adjustment system and b is the cutter speed control quantity adjustment system, which can be determined from historical data. Based on this, a = 0.9, b = 0.1. The essence of the Smith predictor is to use intelligent prediction to compensate for physical delays, so that the control host is no longer "led by the nose" by the pipeline length, but can anticipate the future and act in advance.

[0059] In some embodiments, the particle detection device captures images at a speed greater than the control host detects particle images. The control host further includes an image detection preprocessing unit, which is used for: The particle images collected within a set time period are grouped according to the location of the timestamp corresponding to each particle image at a preset time interval. For each group of particle images, sort them according to the timestamps corresponding to the particle images in that group; According to the sorting order of the particle images, the first particle image in each group is extracted as the current particle image, and the current particle image is detected to determine whether there are abnormal particles in the captured particle image. Then, according to the sorting order of the particle images, the next particle image in each group is extracted as the current particle image, and the step of detecting the current particle image is performed until all particle images in each group have been detected.

[0060] In this embodiment, the core objective of the image processing scheduling mechanism is to address the problem that the image acquisition speed far exceeds the processing speed of the control host for each individual image. This embodiment employs an intelligent priority ranking method to ensure that, even when not all images can be processed, the control host's inference results still evenly cover all time periods, avoiding processing only images concentrated within a short period. For example, First, all images awaiting processing are grouped according to the time they were acquired, accurate to the second. Images acquired within the second 13:00:30 are considered a group. Then, within each time group (per second), the system assigns a relative ranking to the images based on the order in which they were acquired; the first image in that second is ranked 1, the second 2, and so on. When the control host executes the inference task, it prioritizes selecting the image with the highest relative ranking from all the pending particle images.

[0061] First, the image ranked 1st across all time periods is selected. Only after the first image of each second has been selected and processed will the selection of the second-ranked image across all seconds begin. It is evident that the technical solution provided in this embodiment covers data for each time interval, balancing the speed difference between high-speed acquisition and limited inference, thus facilitating particle length estimation.

[0062] In another embodiment, the image detection preprocessing unit calculates the information entropy of each particle image acquired within a set time period. If the information entropy of the particle image is higher than or equal to a quality threshold, the particle image is detected to determine whether there are abnormal particles in the captured image; otherwise, the particle image is deleted. In this embodiment, the image's information entropy is used as the primary quality indicator. A higher information entropy value means that the image contains richer details and texture information, has higher contrast, and is more suitable for accurate inference of particle length in particle images. By applying the technical solution provided in this embodiment, high-quality samples are selected, balancing the speed difference between high-speed acquisition and limited inference, thus facilitating particle length estimation.

[0063] Therefore, in the technical solution provided in this application, the automatic detection system includes a particle detection pipe 11 and a light source access pipe 12 that are longitudinally connected on both sides of a particle conveying pipe 1 at a set position; a particle detection device 2 is installed on the particle detection pipe 11; a light source supplement device 3 is installed on the light source access pipe 12; when the control host receives an instruction to perform particle detection, it controls the light source supplement device 3 and the particle detection device 2 to be turned on, so that the particle detection device 2 can take pictures of the particles passing through the particle detection pipe 11 under the illumination of the light source supplement device 3. If abnormal particles are found in the particle images after identification, the abnormal particles are removed; and for each target particle after filtering, the aspect ratio of the target particle in the two-dimensional plane is calculated; the length of the target particle is determined according to the aspect ratio and the relationship model between the existing particle length and the average aspect ratio. As can be seen, the automatic detection system provided in this application embodiment has a simple structure. When detecting particles, it can achieve real-time online detection without relying on manual labor. It can estimate the length of each particle in the particle image in real time, and determine the particle quality online in real time based on the estimated particle length, so as to achieve the purpose of timely and accurate control. It can not only effectively reduce safety risks, but also avoid the lack of accuracy and timeliness of parameter adjustment caused by the lag and roughness of quality control. It provides data support for the precise control of the production process, thereby improving the consistency of product quality, reducing production costs, and ensuring production safety.

[0064] Secondly, such as Figure 5 As shown in the figure, this application also relates to a flowchart of an automatic particle detection method, which is applied to the control host described in any embodiment of the first aspect. The automatic detection method includes the following steps: Step 101: Upon receiving a command to perform particle detection, control the light source supplementation device 3 and the particle detection device 2 to turn on, so that the particle detection device 2 can take a picture of the particles passing through the particle detection pipe 11 when the light source supplementation device 3 emits light. If abnormal particles are identified in the particle picture, then proceed to step 102.

[0065] Step 102: Remove abnormal particles, and for each filtered target particle, calculate the aspect ratio of the target particle in the two-dimensional plane. If the aspect ratio is based on the relationship model between the length of the existing particles in three-dimensional space and the average aspect ratio, proceed to step 103.

[0066] Step 103: Determine the length of the target particle in three-dimensional space.

[0067] As one embodiment, the method for identifying abnormal particles in the particle image includes: Morphological features of the particle contour are identified based on contour analysis to obtain the contour feature parameters of the particle. Determine whether the contour feature parameters belong to the range of uniform contour feature parameters used to characterize particles. If they do, the particle is determined to be a normal particle; otherwise, the particle is determined to be an abnormal particle.

[0068] As one embodiment, the implementation of the average aspect ratio determination unit and the method for calculating the aspect ratio of the target particle in a two-dimensional plane includes: Obtain particle samples that have been abnormally removed within a set time period; For each particle sample, calculate the minimum bounding rectangle of the particle sample, and calculate the aspect ratio of the particle sample in the two-dimensional plane based on the minimum bounding rectangle. The average aspect ratio is determined based on the aspect ratio and number of particle samples of all particle samples calculated within the set time period. The aspect ratio calculation unit is used for: Calculate the minimum bounding rectangle of the target particle, and based on the minimum bounding rectangle, calculate the aspect ratio of the target particle in the two-dimensional plane.

[0069] As an example, the implementation of constructing a relational model includes: Obtain the average aspect ratio of particle samples in particle images collected within a set time period; For each particle sample, the actual length of the particle sample at the corresponding time point is measured, and the actual average length of each particle sample within the set time period is calculated. Based on the average aspect ratio and the actual average length, a piecewise interpolation method is used to fit a mapping function from the aspect ratio of each particle sample to its corresponding actual length, and the mapping function is determined as the relationship model between particle length and average aspect ratio.

[0070] As one embodiment, the automatic detection method further includes: For each target particle whose length has been obtained, its length is compared with a qualified length threshold. If the length of the target particle is greater than the upper limit of the qualified length threshold, the target particle is marked as a large particle and its quantity is accumulated. If the length of the target particle is less than the lower limit of the qualified length threshold, the target particle is marked as a small particle and its quantity is accumulated. If the length of the target particle is within the qualified length threshold range, the target particle is marked as a qualified particle and its quantity is accumulated. Based on the cumulative values ​​of large particles, small particles, qualified particles, and the total number of target particles, the proportion of large particles, the proportion of small particles, and the particle qualification rate are determined.

[0071] As one embodiment, the automatic detection method further includes: If the proportion of large particles is higher than the qualified threshold for the proportion of large particles, then particle control information indicating an increase in the cutting speed is sent to the cutting device used to cut the particles, so that the cutting device increases the cutting speed of the particles according to the particle control information. If the proportion of small particles is higher than the qualified threshold for the proportion of small particles, then particle control information indicating a reduction in the cutting speed is sent to the cutting device used to cut the particles, so that the cutting device reduces the cutting speed of the particles according to the particle control information. As one embodiment, the particle detection device 2 includes a cavity housing 23 that is closed at one end and open at the other end, an explosion-proof housing 24, a camera mounting base 21, and a camera 22; the closed end of the cavity housing 23 is provided with a wire interface for electrical connection with the control host, and the explosion-proof housing 24 is open at one end and provided with a transparent explosion-proof window 31 at the other end; The explosion-proof housing 24 is installed inside the cavity housing 23 with its open end abutting against the closed end of the cavity housing 23 and isolated from the cavity housing 23. The camera mounting base 21 is installed inside the explosion-proof housing 24. The camera 22 is installed on the camera mounting base 21 with the camera lens facing the transparent explosion-proof window 31. The open end of the cavity housing 23 is connected to the particle detection pipe 11.

[0072] As an example, the automatic detection method further includes: the open end of the cavity shell 23 is threadedly connected to the particle detection pipe 11, and the light source supplement device 3 is threadedly connected to the light source access pipe 12.

[0073] In another embodiment, the automatic detection system further includes a clamping connector with a reverse spring. One end of the clamping connector is detachably connected to the first mounting component connected to the particle detection device 2, and the other end is detachably connected to the particle detection device 2. When particle detection is required, the clamping connector continuously and uniformly applies clamping force to the light source supplementation device 3 and the particle detection device 2 under the action of the reverse spring, so that the particle detection device 2 is firmly connected to the particle detection pipe 11, and the light source supplementation device 3 is firmly connected to the light source inlet pipe 12.

[0074] For the method embodiments, since they basically correspond to the device embodiments, the relevant parts can be referred to in the description of the method embodiments. The method embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this application according to actual needs.

[0075] Therefore, in the technical solution provided in this application, the automatic detection method includes controlling the light source supplementation device 3 and the particle detection device 2 to be turned on when a particle detection command is received. This allows the particle detection device 2 to capture images of particles passing through the particle detection pipe 11 under the illumination of the light source supplementation device 3. If abnormal particles are identified in the captured particle image, these abnormal particles are removed. For each filtered target particle, the aspect ratio in a two-dimensional plane is calculated. Based on the aspect ratio and the relationship model between the existing particle length and the average aspect ratio, the length of the target particle is determined. It is evident that the automatic detection system provided in this application has a simple structure and achieves real-time online detection of particles without relying on manual labor. It can estimate the length of each particle in the particle image in real time and determine the particle quality online in real time based on the estimated particle length, achieving timely and accurate control. This not only effectively reduces safety risks but also avoids the inaccuracy and timeliness of parameter adjustments caused by lagging and crude quality control. It provides data support for precise control of the production process, thereby improving product quality consistency, reducing production costs, and ensuring production safety.

[0076] Thirdly, this embodiment also provides a remote monitoring platform for particles, characterized in that the remote monitoring platform includes a remote monitoring management host, an automatic detection system as described in any embodiment of the first aspect, a wireless communication module, and a cutting device. The remote monitoring management host is connected to the control host and the cutting device respectively through the wireless communication module to send interactive information to the automatic detection system and the cutting device to achieve remote and precise adjustment.

[0077] Fourthly, see Figure 6 , Figure 6 This application provides a schematic diagram of the structure of an automatic particle detection device 500, which is applied to the control host described in any embodiment of the first aspect. The automatic detection device includes: The particle identification unit 501 is used to control the light source supplementation device 3 and the particle detection device 2 to be turned on when a particle detection command is received, so that the particle detection device 2 can take a picture of the particles passing through the particle detection pipe 11 when the light source supplementation device 3 emits light. If abnormal particles are identified in the particle picture, the abnormal rejection unit 502 is triggered.

[0078] The anomaly removal unit 502 is used to remove abnormal particles and calculate the aspect ratio of each target particle in the two-dimensional plane for each filtered target particle. If the aspect ratio is based on the relationship model between the length of the existing particles in the three-dimensional space and the average aspect ratio, the length estimation unit 503 is triggered.

[0079] The length estimation unit 503 is used to determine the length of the target particle in three-dimensional space.

[0080] As an example, the particle identification unit of the automatic detection device is used to: identify the morphological features of the particle contour based on the morphological features of contour analysis to obtain the contour feature parameters of the particle; determine whether the contour feature parameters belong to the range of uniform contour feature parameters used to characterize the particle; if they do, determine that the particle belongs to a normal particle; if they do not, determine that the particle belongs to an abnormal particle.

[0081] As one embodiment, the anomaly rejection unit 502 includes an average aspect ratio determination unit for determining the average aspect ratio and an aspect ratio calculation unit for calculating the aspect ratio of the target particle in a two-dimensional plane. The average aspect ratio determination unit is used for: Obtain particle samples that have been abnormally removed within a set time period; For each particle sample, calculate the minimum bounding rectangle of the particle sample, and calculate the aspect ratio of the particle sample in the two-dimensional plane based on the minimum bounding rectangle. The average aspect ratio is determined based on the aspect ratio and number of particle samples of all particle samples calculated within the set time period. The aspect ratio calculation unit is used for: Calculate the minimum bounding rectangle of the target particle, and based on the minimum bounding rectangle, calculate the aspect ratio of the target particle in the two-dimensional plane.

[0082] As one embodiment, the anomaly removal unit 502 further includes a relation model determination unit for constructing a relation model, the relation model determination being used for: Obtain the average aspect ratio of particle samples in particle images collected within a set time period; For each particle sample, the actual length of the particle sample at the corresponding time point is measured, and the actual average length of each particle sample within the set time period is calculated. Based on the average aspect ratio and the actual average length, a piecewise interpolation method is used to fit a mapping function from the aspect ratio of each particle sample to its corresponding actual length, and the mapping function is determined as the relationship model between particle length and average aspect ratio.

[0083] As one embodiment, the automatic detection device further includes a particle screening unit, which is used for: For each target particle whose length has been obtained, its length is compared with a qualified length threshold. If the length of the target particle is greater than the upper limit of the qualified length threshold, the target particle is marked as a large particle and its quantity is accumulated. If the length of the target particle is less than the lower limit of the qualified length threshold, the target particle is marked as a small particle and its quantity is accumulated. If the length of the target particle is within the qualified length threshold range, the target particle is marked as a qualified particle and its quantity is accumulated. Based on the cumulative values ​​of large particles, small particles, qualified particles, and the total number of target particles, the proportion of large particles, the proportion of small particles, and the particle qualification rate are determined.

[0084] As one embodiment, the automatic detection device further includes a cutting control unit, which is used for: If the proportion of large particles is higher than the maximum value required by the generation target, then particle control information indicating an increase in the cutting speed is sent to the cutting device used to cut the particles, so that the cutting device increases the cutting speed of the particles according to the particle control information; If the proportion of small particles is higher than the minimum value required for generating the target, particle control information indicating a reduction in the cutting speed is sent to the cutting device used to cut the particles, so that the cutting device reduces the cutting speed of the particles according to the particle control information.

[0085] As one embodiment, the particle detection device 2 includes a cavity housing 23 that is closed at one end and open at the other end, a camera mounting base 21, and a camera 22; the closed end of the cavity housing 23 is provided with a wire interface for electrical connection with the control host. The camera mounting base 21 is installed inside the cavity housing 23, and the camera 22 is installed on the camera mounting base 21 with the camera lens facing the open end of the cavity housing 23. The open end of the cavity housing 23 is connected to the particle detection pipe 11.

[0086] As an example, the automatic detection method further includes: the open end of the cavity shell 23 is threadedly connected to the particle detection pipe 11, and the light source supplement device 3 is threadedly connected to the light source access pipe 12.

[0087] In another embodiment, the automatic detection system further includes a clamping connector with a reverse spring. One end of the clamping connector is detachably connected to the first mounting component connected to the particle detection device 2, and the other end is detachably connected to the particle detection device 2. When particle detection is required, the clamping connector continuously and uniformly applies clamping force to the light source supplementation device 3 and the particle detection device 2 under the action of the reverse spring, so that the particle detection device 2 is firmly connected to the particle detection pipe 11, and the light source supplementation device 3 is firmly connected to the light source inlet pipe 12.

[0088] Therefore, in the technical solution provided in this application, the automatic detection device includes controlling the light source supplementation device 3 and the particle detection device 2 to be turned on when receiving an instruction to perform particle detection. This allows the particle detection device 2 to capture images of particles passing through the particle detection pipe 11 under the illumination of the light source supplementation device 3. If abnormal particles are identified in the captured particle image, these abnormal particles are removed. For each filtered target particle, the aspect ratio in a two-dimensional plane is calculated. Based on the aspect ratio and the relationship model between the existing particle length and the average aspect ratio, the length of the target particle is determined. It is evident that the automatic detection system provided in this application has a simple structure. When detecting particles, it eliminates the need for manual sampling and can estimate the particle length in the particle image in real time. Based on the estimated particle length, the particle quality can be determined online in real time, achieving timely and accurate control. This not only effectively reduces safety risks but also avoids the inaccuracy and timeliness of parameter adjustments caused by lagging and crude quality control. It provides data support for precise control of the production process, thereby improving product quality consistency, reducing production costs, and ensuring production safety.

[0089] Fifthly, this application also provides an electronic device. From a hardware perspective, a hardware architecture diagram can be found in [reference needed]. Figure 7 As shown, it includes a machine-readable storage medium and a processor, wherein: the machine-readable storage medium stores machine-executable instructions that can be executed by the processor; the processor is used to execute the machine-executable instructions to implement the automatic detection operation disclosed in the above example.

[0090] The machine-readable storage medium provided in this application embodiment stores machine-executable instructions. When the machine-executable instructions are invoked and executed by a processor, the machine-executable instructions cause the processor to perform the automatic detection operation disclosed in the above examples.

[0091] Here, a machine-readable storage medium can be any electronic, magnetic, optical, or other physical storage device that can contain or store information, such as executable instructions, data, etc. For example, a machine-readable storage medium can be: RAM (Random Access Memory), volatile memory, non-volatile memory, flash memory, storage drives (such as hard disk drives), solid-state drives, any type of storage disk (such as optical discs, DVDs, etc.), or similar storage media, or combinations thereof.

[0092] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer, which can take the form of a personal computer, laptop computer, cellular phone, camera, telephone, smartphone, personal digital assistant, media player, navigation device, email sending and receiving device, game console, tablet computer, wearable device, or any combination of these devices.

[0093] For ease of description, the above devices are described separately by function as various units. Of course, in implementing this application, the functions of each unit can be implemented in one or more software and / or hardware.

[0094] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, embodiments of this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0095] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0096] Furthermore, these computer program instructions can also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in the process. Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0097] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0098] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. An automatic detection system for particles, characterized in that, The automatic detection system includes: The particle conveying pipeline has a particle detection pipeline and a light source access pipeline running longitudinally through both sides of the set position. A particle detection device is installed on the particle detection pipeline; A light source supplementation device is installed on the light source inlet pipe; The control host, electrically connected to the particle detection device and the light source supplementation device, is used to control the light source supplementation device and the particle detection device to turn on when a particle detection command is received, so that the particle detection device can capture images of particles passing through the particle detection pipe under the illumination of the light source supplementation device. If abnormal particles are found in the captured particle images, the abnormal particles are removed. For each filtered target particle, the aspect ratio of the target particle in the two-dimensional plane is calculated. Based on the aspect ratio and the relationship model between the actual length and average aspect ratio of existing particles in three-dimensional space, the length of the target particle in three-dimensional space is determined.

2. The automatic detection system according to claim 1, characterized in that, The control host includes a particle recognition unit for identifying abnormal particles in the particle image, the particle recognition unit being used for: Morphological features of the particle contour are identified based on contour analysis to obtain the contour feature parameters of the particle. Determine whether the contour feature parameters belong to the range of uniform contour feature parameters used to characterize particles. If they do, the particle is determined to be a normal particle; otherwise, the particle is determined to be an abnormal particle.

3. The automatic detection system according to claim 1, characterized in that, The control host includes an average aspect ratio determination unit for determining the average aspect ratio and an aspect ratio calculation unit for calculating the aspect ratio of the target particle in a two-dimensional plane. The average aspect ratio determination unit is used for: Obtain particle samples that have been abnormally removed within a set time period; For each particle sample, calculate the minimum bounding rectangle of the particle sample, and calculate the aspect ratio of the particle sample in the two-dimensional plane based on the minimum bounding rectangle. The average aspect ratio is determined based on the aspect ratio and number of particle samples calculated for all particle samples within the set time period. The aspect ratio calculation unit is used for: Calculate the minimum bounding rectangle of the target particle, and based on the minimum bounding rectangle, calculate the aspect ratio of the target particle in the two-dimensional plane.

4. The automatic detection system according to claim 3, characterized in that, The control host also includes a relational model determination unit for constructing a relational model, the relational model determination unit being used for: Obtain the average aspect ratio of particle samples in particle images collected within a set time period; For each particle sample, the actual length of the particle sample at the corresponding time point is measured, and the actual average length of each particle sample within the set time period is calculated. Based on the average aspect ratio and the actual average length, a piecewise interpolation method is used to fit a mapping function from the aspect ratio of each particle sample to its corresponding actual length, and the mapping function is determined as the relationship model between particle length and average aspect ratio.

5. The automatic detection system according to claim 4, characterized in that, The control host also includes a particle screening unit, which is used for: For each target particle whose length has been obtained, its length is compared with a qualified length threshold. If the length of the target particle is greater than the upper limit of the qualified length threshold, the target particle is marked as a large particle and its quantity is accumulated. If the length of the target particle is less than the lower limit of the qualified length threshold, the target particle is marked as a small particle and its quantity is accumulated. If the length of the target particle is within the qualified length threshold range, the target particle is marked as a qualified particle and its quantity is accumulated. Based on the cumulative values ​​of large particles, small particles, qualified particles, and the total number of target particles, the proportion of large particles, the proportion of small particles, and the particle qualification rate are determined.

6. The automatic detection system according to claim 5, characterized in that, The control host also includes a cutting control unit, which is used for: If the proportion of large particles is higher than the qualified threshold for the proportion of large particles, then particle control information indicating an increase in the cutting speed is sent to the cutting device used to cut the particles, so that the cutting device increases the cutting speed of the particles according to the particle control information. If the proportion of small particles is higher than the acceptable threshold for small particle proportion, particle control information indicating a reduction in the cutting speed is sent to the cutting device used to cut the particles, so that the cutting device reduces the cutting speed of the particles according to the particle control information.

7. The automatic detection system according to claim 3, characterized in that, The particle detection device captures images at a speed greater than the control host's speed for detecting particle images. The control host further includes an image detection preprocessing unit, which is used for: The particle images collected within a set time period are grouped according to the location of the timestamp corresponding to each particle image at a preset time interval. For each group of particle images, sort them according to the timestamps corresponding to the particle images in that group; According to the sorting order of particle images, the highest-ranking particle image in each group is extracted as the current particle image, and the current particle image is inspected to determine whether there are any abnormal particles in the captured particle image. Then, according to the sorting order of particle images, the next particle image in each group is extracted as the current particle image, and the step of inspecting the current particle image is performed until all particle images in each group have been inspected; or Calculate the information entropy of each particle image collected within a set time period. If the information entropy of the particle image is higher than or equal to the quality threshold, then the particle image is detected to determine whether there are abnormal particles in the captured particle image; otherwise, the particle image is deleted.

8. The automatic detection system according to claim 5, characterized in that, The particle detection device includes a cavity housing that is closed at one end and open at the other, a camera mounting base, an explosion-proof housing, and a camera; the closed end of the cavity housing is provided with a wire interface for electrical connection with the control host, and one end of the explosion-proof housing is open and the other end is provided with a transparent explosion-proof window; The explosion-proof housing is installed inside the cavity housing with its open end abutting against the closed end of the cavity housing and in an isolated and sealed manner. The camera mounting base is installed inside the explosion-proof housing. The camera is mounted on the camera mounting base with the camera lens facing the transparent explosion-proof window. The open end of the cavity housing is connected to the particle detection pipeline.

9. An automatic detection method for particles, characterized in that, The automatic detection method is applied to the control host of the automatic detection system according to any one of claims 1 to 8, and the automatic detection method includes: When a particle detection command is received, the light source supplement device and the particle detection device are turned on, so that the particle detection device can take pictures of the particles passing through the particle detection pipe when the light source supplement device emits light. If abnormal particles are found in the particle pictures, the abnormal particles are removed. For each target particle after filtration, calculate the aspect ratio of the target particle in a two-dimensional plane; Based on the aspect ratio and the relationship model between the actual length of existing particles in three-dimensional space and the average aspect ratio, the length of the target particle in three-dimensional space is determined.

10. A remote monitoring platform for particles, characterized in that, The remote monitoring platform includes a remote monitoring management host, an automatic detection system according to any one of claims 1 to 8, a wireless communication module, and a cutting device. The remote monitoring management host is connected to the control host and the cutting device respectively through the wireless communication module to send interactive information to the automatic detection system and the cutting device.

Images