Rapeseed oil production line real-time monitoring method and system

Abstract

The invention relates to the technical field of image recognition, in particular to a rapeseed oil production line real-time monitoring method and system, and the method comprises the steps: obtaining an oil cake time sequence image sequence in the operation process of a rapeseed oil production line; static forming parameters of the oil cakes in the multiple oil cake images within the preset time period are determined; based on the static forming parameters corresponding to the multiple oil cake images in the preset time period, dynamic forming parameters of the oil cake in the preset time period are determined; determining the forming state of the oil cake based on the dynamic forming parameters; and when the forming state is an unformed state, performing maintenance alarm on the rapeseed oil production line. According to the method, through a double-layer feature system in which the static forming parameters and the dynamic forming parameters are matched, judgment of the forming state of the oil cake is upgraded from single-frame instantaneous estimation to continuous-period accumulated credibility evaluation, the misjudgment rate caused by instantaneous texture or illumination disturbance is reduced, and therefore the accuracy of anomaly detection of the rapeseed oil production line is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image recognition, and in particular to a rapeseed oil production line real-time monitoring method and system. BACKGROUND

[0002] The rapeseed oil production line refers to a complete set of processing equipment and process flow for realizing automatic and continuous production from rapeseed raw materials to finished edible oil, covering cleaning, pressing or leaching, refining, packaging and other links. In the rapeseed oil production line, each process link needs to be closely connected and cooperated. Among them, the forming quality of oil cake as a key node directly determines the pressing efficiency and residual oil rate. Loose, fragmentation or uneven thickness of the oil cake, etc. unmolding phenomenon can map the upstream link problems such as improper rapeseed moisture control, deviation of frying time and temperature, abnormal inner wall of the machine, uneven feeding or operation parameter drift. These problems will further cause problems such as increased residual oil rate and decreased oil yield, and even cause equipment blockage, which seriously affects the production continuity and oil quality. Therefore, real-time monitoring of the forming state of the oil cake is the core requirement to ensure the efficient and high-quality operation of the rapeseed oil production line.

[0003] The existing problem: the monitoring means of the existing rapeseed oil automatic production line mainly relies on independent monitoring of each process link, and realizes overall control through a multi-parameter cooperative judgment mechanism. Once an abnormality is determined, the troubleshooting is usually carried out by stopping the line. This monitoring method is due to the fact that the parameters of each link are mutually fragmented and lack direct online perception of the forming state of the oil cake, which leads to the inability to continuously and accurately identify the full-link abnormality mapped by the unmolding of the oil cake in the production process, resulting in frequent shutdown for maintenance and serious resource waste. SUMMARY

[0004] The present application provides a rapeseed oil production line real-time monitoring method and system to solve the existing problems.

[0005] The rapeseed oil production line real-time monitoring method and system provided by the present application adopts the following technical scheme: An embodiment of the present application provides a rapeseed oil production line real-time monitoring method, which comprises: acquiring a time sequence of oil cake images of a rapeseed oil production line in a running process; determining static forming parameters of oil cakes in a plurality of oil cake images in a preset period; wherein the static forming parameters are used to represent the oil cake forming probability of a single oil cake image at its acquisition time; determining dynamic forming parameters of the oil cakes in the preset period based on the static forming parameters corresponding to the plurality of oil cake images in the preset period; wherein the dynamic forming parameters are used to represent the cumulative credibility of the oil cakes maintaining the forming state in the preset period; determining the forming state of the oil cakes based on the dynamic forming parameters; and when the forming state is unmolding, performing rapeseed oil production line maintenance alarm.

[0006] Further, the static forming parameters of the oil cake in the single oil cake image are determined by performing superpixel segmentation on the oil cake image to obtain a plurality of tile segmentation regions, extracting area features and brightness features of each tile segmentation region, respectively, and determining regional static forming parameters of each tile segmentation region based on the area features and the brightness features, and performing mean value processing on the regional static forming parameters of the plurality of tile segmentation regions to obtain the static forming parameters of the oil cake in the oil cake image.

[0007] Further, the regional static forming parameters of the single tile segmentation region are determined by determining a first local forming probability based on the area features of the tile segmentation region, wherein the first local forming probability is used to represent a local forming probability of the oil cake in the tile segmentation region determined based on the area features, determining a second local forming probability based on the brightness features of the tile segmentation region, wherein the second local forming probability is used to represent a local forming probability of the oil cake in the tile segmentation region determined based on the brightness features, determining a first weight value based on an area feature distribution of all tile segmentation regions in the oil cake image where the tile segmentation region is located, wherein the first weight value is used to represent an overall forming probability of the oil cake in the oil cake image determined based on the area feature distribution, determining a second weight value based on a brightness feature distribution of all tile segmentation regions in the oil cake image where the tile segmentation region is located, wherein the second weight value is used to represent an overall forming probability of the oil cake in the oil cake image determined based on the brightness feature distribution, and performing weighted summation on the first local forming probability and the second local forming probability based on the first weight value and the second weight value to obtain the regional static forming parameters of the tile segmentation region.

[0008] Further, the dynamic forming parameters of the oil cake in the preset time period are determined based on the static forming parameters corresponding to the plurality of oil cake images in the preset time period, by determining a time sequence difference index between two adjacent oil cake images based on the static forming parameters corresponding to the plurality of oil cake images in the preset time period, wherein the time sequence difference index is used to represent a mutation degree of the static forming parameters between the two oil cake images, clustering the time sequence difference index to determine an abnormal cluster, dividing the preset time period into a plurality of continuous segments based on the abnormal cluster, and determining the dynamic forming parameters of the oil cake in the preset time period based on oil cake morphological features of the oil cake in the plurality of continuous segments.

[0009] Further, the clustering of the time sequence difference indicators to determine the abnormal cluster comprises: clustering a plurality of the time sequence difference indicators to obtain a first clustering cluster and a second clustering cluster; determining an abnormal cluster and a normal cluster in the first clustering cluster and the second clustering cluster; wherein a clustering cluster containing a smaller number of samples is the abnormal cluster, and a clustering cluster containing a larger number of samples is the normal cluster.

[0010] Further, the dividing of the preset time period into a plurality of continuous segments based on the abnormal cluster comprises: taking a starting time of the preset time period as a starting point, and sequentially determining a clustering cluster to which the time sequence difference indicators belong; when the time sequence difference indicators belonging to the abnormal cluster are determined for the first time, taking a previous time of the time sequence difference indicators as an end point of a current continuous segment; taking a next time of the time sequence difference indicators as a starting point of a next continuous segment, and determining an end point of the continuous segment; and repeating the above process until the entire preset time period is traversed, to obtain a plurality of the continuous segments.

[0011] Further, the determination of the dynamic forming parameter of the oil cake in the preset time period based on the oil cake morphological feature in a plurality of the continuous segments comprises: for each of the continuous segments, performing tile segmentation on the oil cake image at each time in the continuous segment to obtain a plurality of tile regions; for each of the tile regions, calculating tile similarity of the tile region and each of the tile regions in a target time to obtain a morphological parameter of each of the tile regions; wherein the target time is used to represent all subsequent times of the oil cake image corresponding to the time at which the tile region is located in the continuous segment; the morphological parameter is a tile similarity average of the tile region; based on a similarity weighting weight, the morphological parameters of each of the tile regions in the continuous segment are weighted and summed to obtain the dynamic forming parameter of the continuous segment; wherein the similarity weighting weight is used to represent a duration influence degree of the oil cake image in which the tile region is located in the continuous segment; based on an oil cake morphological consistency weight, the dynamic forming parameters of each of the continuous segments are weighted and summed and averaged to determine the dynamic forming parameter of the oil cake in the preset time period; wherein the morphological consistency weight is used to represent an oil cake morphological consistency degree between two adjacent continuous segments.

[0012] Further, the determination of the forming state of the oil cake based on the dynamic forming parameter comprises: performing normalization processing on the dynamic forming parameter of the oil cake in the preset time period; if the dynamic forming parameter after the normalization processing is greater than a preset forming threshold, it is determined that the forming state of the oil cake is formed; and if the dynamic forming parameter after the normalization processing is not greater than the preset forming threshold, it is determined that the forming state of the oil cake is not formed.

[0013] Further, after the image sequence of the oil cakes in the running process of the rapeseed oil production line is acquired, the method further comprises: pre-processing the oil cake images in the image sequence of the oil cakes; wherein the pre-processing comprises at least one of filtering processing, gray processing and threshold segmentation processing.

[0014] Another embodiment of the present application provides a real-time monitoring system for a rapeseed oil production line, comprising: an image acquisition device and a host computer, wherein: The image acquisition device is arranged at a cake discharging port of the rapeseed oil production line, and is configured to acquire real-time images of the oil cakes and send the images of the oil cakes to the host computer. The host computer is configured to acquire an image sequence of the oil cakes in the running process of the rapeseed oil production line based on the images of the oil cakes acquired by the image acquisition device, determine static forming parameters of the oil cakes in a plurality of the images of the oil cakes within a preset period, wherein the static forming parameters are used to represent forming probabilities of the oil cakes in the images of the oil cakes at their acquisition time points, determine dynamic forming parameters of the oil cakes within the preset period based on the static forming parameters corresponding to the plurality of the images of the oil cakes within the preset period, wherein the dynamic forming parameters are used to represent cumulative reliabilities of the oil cakes in the forming state within the preset period, determine a forming state of the oil cakes based on the dynamic forming parameters, and perform rapeseed oil production line maintenance alarm when the forming state is unformed.

[0015] The technical scheme of the present application has the following beneficial effects: In the embodiment of the present application, an image sequence of the oil cakes in the running process of the rapeseed oil production line is acquired, static forming parameters of the oil cakes in a plurality of the images of the oil cakes within a preset period are determined, wherein the static forming parameters are used to represent forming probabilities of the oil cakes in the images of the oil cakes at their acquisition time points, dynamic forming parameters of the oil cakes within the preset period are determined based on the static forming parameters corresponding to the plurality of the images of the oil cakes within the preset period, wherein the dynamic forming parameters are used to represent cumulative reliabilities of the oil cakes in the forming state within the preset period, a forming state of the oil cakes is determined based on the dynamic forming parameters, and rapeseed oil production line maintenance alarm is performed when the forming state is unformed.

[0016] So far, the application upgrades the discrimination of oil cake forming state from single-frame instantaneous estimation to cumulative reliability evaluation in a continuous period by the double-layer feature system of static forming parameters and dynamic forming parameters, reduces the misjudgment rate caused by instantaneous texture or light disturbance, and thus improves the accuracy of rapeseed oil production line anomaly detection. On the other hand, the oil cake image is converted into a tile segmentation region by superpixel segmentation, and the regional static forming parameters are constructed on two low-dimensional features of area and brightness, and then coupled by global weight, which not only compresses the data volume, but also retains the key morphological information, reduces the computing load and storage demand of the upper computer. On the other hand, the preset period is continuously segmented by the abnormal cluster obtained based on the time sequence difference index and clustering, and the time-decreasing similarity weighting weight is introduced in each segment, which can effectively distinguish between temporary disturbance and continuous anomaly, avoid invalid shutdown, and thus improve the continuous running time of the production line. On the other hand, the automatic discrimination of forming state and unformed state is realized by the normalization comparison of dynamic forming parameters and preset forming threshold, and the maintenance alarm is triggered immediately when it is unformed, which shortens the response time from anomaly occurrence to manual intervention, and reduces the economic loss caused by the risk of increasing residual oil rate and equipment blockage. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description only constitute some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor based on these drawings.

[0018] Figure 1 The flowchart of the rapeseed oil production line real-time monitoring method provided by the embodiment of the present application; Figure 2 The structure diagram of the rapeseed oil production line real-time monitoring system provided by the embodiment of the present application. DETAILED DESCRIPTION

[0019] In order to further illustrate the technical means and effects adopted by the present application to achieve the predetermined invention purpose, the following will combine the drawings and preferred embodiments to specifically describe the specific implementation, structure, features and effects of the rapeseed oil production line real-time monitoring method and system according to the present application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0021] Before the specific scheme of the rapeseed oil production line real-time monitoring method and system is introduced, the principle on which the rapeseed oil production line real-time monitoring method and system realizes real-time monitoring of the rapeseed oil production line is introduced: It should be noted that the rapeseed oil production line to which the embodiments of the present application are directed adopts a general pressing process, and the technical path is as follows: the rapeseed after cleaning and conditioning is separated from oil by physical extrusion; under the action of high pressure, the oil flows out of the oil outlet of the press chamber to form crude oil, and the residual solids are pressed into solid flakes or block cake embryos and continuously discharged from the cake discharge port, which are called oil cakes. The geometric shape, structural integrity and surface optical properties of the oil cake directly reflect the running state of each process link in the upstream, therefore, by monitoring the oil cake in real time, the health status of the rapeseed oil production line can be non-invasively deduced.

[0022] The oil cake can be clearly distinguished into two categories of "shaped oil cake" and "unshaped oil cake" in terms of morphology. Among them, the shaped oil cake is subjected to plastic deformation of rapeseed particles under the action of continuous pressure, the cell wall is broken, the oil is released, and a continuous and stable gel porous structure is formed between the particles through residual lipids; macroscopically, it presents as large-size flakes or block-shaped structure, the surface is dense, the hardness is high, and the deformation amount after being separated from the press chamber is extremely small. The unshaped oil cake is mainly subjected to elastic deformation, the particles rebound and expand after the pressure is removed, the pore network is discontinuous and invalid, and it macroscopically presents as a powder or small block structure, has a high oil content, a soft texture, and is easy to be secondarily broken or collapsed after being separated from the press chamber.

[0023] From the perspective of oil quality, the residual oil content of the shaped oil cake is low, and the subsequent leaching or refining loss is low; due to the back suction effect, the unshaped oil cake has an increased content of residual oil and impurities in the cake, and thus needs to increase the chemical refining load. From the perspective of equipment reliability, the shaped oil cake has a complete structure, which can reduce the wear of the press screw and reduce the risk of blockage of the conveying device; the unshaped oil cake is loose and easy to break, which aggravates the wear of the press screw and the conveying auger, and is easy to form a bridge in the oil tank and other parts, forcing the production line to be non-planned shutdown.

[0024] The fundamental reasons for the unshaped oil cake can be summarized as follows: (1) Abnormality in the roasting seed link: temperature or moisture control failure leads to insufficient plasticity or excessive lubrication of the rapeseed; (2) Abnormality in the pressing link: lack of low-pressure pre-pressing stage or local insufficient pressure in the pressure curve, resulting in insufficient bonding of the particles; (3) Abnormality in the press chamber structure: smooth surface of the press screw without sawtooth, poor sealing of the press cage, resulting in failure to establish an effective pressure gradient; (4) Abnormality in the pretreatment of raw materials: incomplete cleaning introduces impurities such as sand and straw to damage the particle bonding surface.

[0025] The rapeseed oil production line real-time monitoring method and system provided by the embodiment of the present application can realize real-time monitoring of the rapeseed oil production line based on the above mechanism. The specific scheme of the rapeseed oil production line real-time monitoring method and system provided by the present application will be described in detail below with reference to the drawings.

[0026] Please refer to Figure 1 which shows a rapeseed oil production line real-time monitoring method provided by an embodiment of the present application, comprising: Step S110: Obtain the oil cake time sequence image sequence of the rapeseed oil production line in the running process.

[0027] It should be noted that the image acquisition device for collecting the oil cake image can be arranged directly above the cake discharge port of the rapeseed oil production line. The image acquisition device can use an industrial-grade high-definition network camera, and the lens optical axis is perpendicular to the cake discharge direction to ensure that the imaging plane is parallel to the oil cake surface, thereby reducing the perspective distortion. The camera outputs the code stream in real time through the gigabit Ethernet interface, and the frame rate is 1 fps, that is, the time interval between adjacent two images is 1 s. The sampling frequency is set according to the actual extrusion speed of the oil cake: according to the field measurement, the oil cake extrusion speed at the cake discharge port is not particularly fast, and the sampling period of 1 s is sufficient to reduce data redundancy without missing key morphological information. The collected original image sequence is denoted as:

[0028] Among them, represents the acquisition time of the image.

[0029] Preferably, in an embodiment of the present application, after the above step S110, the method can further comprise: pre-processing the oil cake images in the oil cake time sequence image sequence; wherein the pre-processing comprises at least one of filtering processing, grayscale processing and threshold segmentation processing.

[0030] It should be noted that the above filtering processing can use a mean filter to filter the image The main purpose is to eliminate Gaussian noise in the image. The main purpose of the grayscale processing of the image is to reduce the number of channels of the image, reduce the data volume, and facilitate subsequent image processing with reduced difficulty. The threshold segmentation processing of the image can use a threshold segmentation method such as Otsu threshold segmentation method, and the purpose is to make only the oil cake information exist in the original image, reduce the interference and invalid calculation of other device information.

[0031] It should be further explained that the above-mentioned Otsu threshold segmentation method is an adaptive threshold selection algorithm based on the image grayscale histogram. Its core idea is to divide the image pixels into two categories, foreground (target) and background, by traversing all possible grayscale thresholds, so that the inter-class variance between the two categories is maximized and the intra-class variance is minimized. Ultimately, the grayscale level that maximizes the inter-class variance is determined as the optimal segmentation threshold, thereby achieving adaptive binary segmentation without the need for manual parameter setting. It can be understood that the above-mentioned Otsu threshold segmentation method is a relatively mature and well-known technology in image processing technology. For its specific implementation method, please refer to the relevant technology, and the embodiments of the present invention will not be repeated.

[0032] Step S120: determining static forming parameters of the oil cakes in the plurality of oil cake images within a preset time period; wherein the static forming parameters are used to characterize the probability of an oil cake being formed in a single oil cake image at the time of its acquisition.

[0033] Preferably, in one embodiment of the present invention, the above-mentioned step S120 may include: performing superpixel segmentation on the oil cake image to obtain multiple tile segmentation areas; respectively extracting the area characteristics and brightness characteristics of each tile segmentation area, and determining the regional static forming parameters of each tile segmentation area based on the area characteristics and brightness characteristics; averaging the regional static forming parameters of multiple tile segmentation areas to obtain the static forming parameters of the oil cake in the oil cake image.

[0034] Preferably, in one embodiment of the present invention, determining the regional static forming parameters of a single tile segmentation area includes: determining a first local forming probability based on the area characteristics of the tile segmentation area; wherein the first local forming probability is used to characterize the local forming probability of the oil cake in the tile segmentation area determined based on the area characteristics; determining a second local forming probability based on the brightness characteristics of the tile segmentation area; wherein the second local forming probability is used to characterize the local forming probability of the oil cake in the tile segmentation area determined based on the brightness characteristics; determining a first weight value based on the area feature distribution of all tile segmentation areas in the oil cake image where the tile segmentation area is located; wherein the first weight value is used to characterize the overall forming probability of the oil cake in the oil cake image determined based on the area feature distribution; determining a second weight value based on the brightness feature distribution of all tile segmentation areas in the oil cake image where the tile segmentation area is located; wherein the second weight value is used to characterize the overall forming probability of the oil cake in the oil cake image determined based on the brightness feature distribution; and performing weighted summation of the first local forming probability and the second local forming probability based on the first weight value and the second weight value to obtain the regional static forming parameters of the tile segmentation area.

[0035] It should be noted that the raw data of the oil cake image at each moment contains a relatively short and fixed moment when the rapeseed oil residue is squeezed out of the production line into oil cakes, which can be understood as a static moment. From the above mechanism analysis, it can be concluded that there is a static difference between the unformed and formed oil cakes, so this feature can be used to analyze the precipitation shape of the oil cake at the static moment. Take the moment as an example, as shown below: Step 1: The original image at the moment is used as the basis, and the super pixel segmentation algorithm is used to perform tile segmentation to obtain the All tile segmentation areas at a moment; Step 2: After superpixel segmentation, we can obtain All the moments Tile segmentation area; First Take the tile segmentation area as an example ( ), the corresponding regional static forming parameters The steps to obtain include: (1) First, All the moments The area of ​​each tile segmentation area is obtained by For example, the corresponding area is The number of all pixels in the region is recorded as ; (2) Then proceed to All areas The color of each tile segmentation area is obtained, and the For example, the corresponding color is The mean grayscale value of all pixels in the region is recorded as ; (3) Get the The minimum area of ​​the tile segmentation region at a moment and The average area of ​​multiple tile segmentation regions at a time ; (4) Get the The maximum brightness of all tile segmentation areas at a moment and The average brightness of all tile segmentation areas at the moment ; (5) Use all the parameters that have been obtained to perform the The static shaping parameters of the tile segmentation area The calculation formula is:

[0036] wherein, is normalized.

[0037] It should be further explained that the above regional static forming parameters are calculated by using the coupling method of weight-difference value, mainly including: (1) Global forming probability weight calculation: taking the normalized value of area dispersion and the normalized value of brightness dispersion as the global forming probability weight, and the smaller the value, the more the whole frame of oil cake tends to be unformed, and the weight value is reduced, thereby suppressing the local features.

[0038] (2) Local forming probability calculation: taking the area difference value of the first tile segmentation region as the first local forming probability of the tile segmentation region, and taking the brightness difference value as the second local forming probability.

[0039] (3) Coupling output.

[0040] It can be understood that the current comprehensive possibility of the unformed extrusion of the oil cake at the th moment, the formed oil cake has larger area and lower brightness compared with the unformed oil cake in the extrusion process, but even if it is unformed extrusion, there is a part of the area with smaller brightness and larger area, so the state of the whole moment of the oil cake needs to be used as a weight to limit it. Therefore, the smaller the value, the smaller the value, the greater the possibility of the whole extrusion oil cake at the th moment being unformed. Then, the area difference between the th tile region and the smallest region and the brightness difference between the th tile region and the maximum brightness of all regions are used to determine the possibility of the th region being an unformed region. The greater the area difference between the th region and the smallest region, and the greater the brightness difference between the th region and the maximum brightness region, the greater the possibility of the th region being a formed region. The purpose of the global forming probability weight for weighting the main part is: even if the main part value is large, which shows that the region is formed, but the whole extrusion at the th moment is an unformed region, then the weight is used to limit the th region, and vice versa.

[0041] The larger the area static forming parameter of the first tile segmentation area at the first moment is, the more likely the first tile segmentation area is to be a formed oil cake, and vice versa. The larger the area static forming parameter of the first tile segmentation area at the first moment is, the more likely the first tile segmentation area is to be a formed oil cake, and vice versa. The larger the area static forming parameter of the first tile segmentation area at the first moment is, the more likely the first tile segmentation area is to be a formed oil cake, and vice versa.

[0042] Step three, the area static forming parameters of all tile areas at the first moment are obtained by the above method, the static parameters of each tile area can be obtained, and then the static forming parameter at the first moment is obtained by using the static parameters of all tile areas, and the specific calculation formula is as follows:

[0043] The above scheme represents the static forming parameter of the overall oil cake extrusion at the first moment by the average value of the area static forming parameters of all tile areas at the first moment. The larger the value is, the more tile areas at the first moment show the characteristics of large area and low brightness, that is, the more likely the oil cake in the oil cake image at the first moment is a formed oil cake, and vice versa.

[0044] At this point, during the production of rapeseed oil on the production line, the oil cake static forming parameter at the first moment is obtained by monitoring and analyzing the oil cake discharge at the first moment. It can be understood that the superpixel segmentation algorithm is a kind of calculation method for clustering image pixels into local homogeneous regions with similar color, brightness or texture characteristics. The superpixel segmentation algorithm is a relatively mature known technology in image processing technology, and its specific implementation mode is described in the related art, and will not be described here.

[0045] It can be understood that the superpixel segmentation algorithm is a kind of calculation method for clustering image pixels into local homogeneous regions with similar color, brightness or texture characteristics. The superpixel segmentation algorithm is a relatively mature known technology in image processing technology, and its specific implementation mode is described in the related art, and will not be described here.

[0046] Step S130: determining a dynamic forming parameter of the oil cake in the preset period based on the static forming parameters corresponding to the plurality of oil cake images in the preset period; wherein the dynamic forming parameter is used to represent the cumulative credibility of the oil cake keeping the forming state in the preset period.

[0047] ​​​​​​​​​​​​Preferably, in an embodiment of the present application, the step S130 can include: determining a time sequence difference index between two adjacent oil cake images based on the static forming parameters corresponding to the plurality of oil cake images in the preset period; wherein the time sequence difference index is used to represent the mutation degree of the static forming parameters between the two oil cake images; clustering the time sequence difference index to determine an abnormal cluster; dividing the preset period into a plurality of continuous segments based on the abnormal cluster; and determining the dynamic forming parameters of the oil cake in the preset period based on the oil cake morphological features of the plurality of continuous segments.

[0048] Preferably, in an embodiment of the present application, the clustering of the time sequence difference index to determine the abnormal cluster includes: clustering the plurality of time sequence difference indexes to obtain a first clustering cluster and a second clustering cluster; determining an abnormal cluster and a normal cluster in the first clustering cluster and the second clustering cluster; wherein the clustering cluster containing a smaller number of samples is the abnormal cluster, and the clustering cluster containing a larger number of samples is the normal cluster.

[0049] Preferably, in an embodiment of the present application, the dividing of the preset period into a plurality of continuous segments based on the abnormal cluster includes: taking the starting time of the preset period as the starting point, and sequentially determining the clustering cluster to which the time sequence difference index belongs; when the time sequence difference index belonging to the abnormal cluster is determined for the first time, taking the previous time of the time sequence difference index as the end point of the current continuous segment; taking the next time of the time sequence difference index as the starting point of the next continuous segment, and determining the end point of the continuous segment; and repeating the above process until the entire preset period is traversed to obtain the plurality of continuous segments.

[0050] Preferably, in an embodiment of the present application, the determining of the dynamic forming parameters of the oil cake in the preset period based on the oil cake morphological features of the plurality of continuous segments includes: for each continuous segment, performing tile segmentation on the oil cake images at each time in the continuous segment to obtain a plurality of tile regions; for each tile region, calculating the tile similarity between the tile region and the tile region at a target time to obtain the morphological parameters of each tile region; wherein the target time is used to represent all subsequent times of the oil cake image corresponding to the time at which the tile region is located in the continuous segment; the morphological parameters are the average tile similarity of the tile region; based on a similarity weighting weight, the morphological parameters of each tile region in the continuous segment are weighted and summed to obtain the dynamic forming parameters of the continuous segment; wherein the similarity weighting weight is used to represent the continuous influence degree of the oil cake image in which the tile region is located in the continuous segment; based on an oil cake morphological consistency weight, the dynamic forming parameters of each continuous segment are weighted and summed and averaged to determine the dynamic forming parameters of the oil cake in the preset period; wherein the morphological consistency weight is used to represent the oil cake morphological consistency degree between two adjacent continuous segments.

[0051] It should be noted that the above scheme captures the corresponding morphological characteristics of the oil cake at a static moment. However, when monitoring the current rapeseed oil production line using only the morphological characteristics of the oil cake at a static moment, it is easy to fall into the short-term error and one-sided analysis. The short-term error specifically refers to the fact that during the rapeseed oil production process, the oil cake only presents a loose shape for a short period of time, which is not sustained and has little impact on overall production. One-sided analysis refers to the fact that the rapeseed oil cake has its own texture. Using static analysis methods, there is a possibility that this texture will be detected as oil cake lumps, which may lead to inaccurate judgment of rapeseed oil production line abnormalities.

[0052] During the continuous extrusion process, the morphology of the oil cake exhibits significant temporal consistency differences: formed oil cakes, due to their dense, porous gel structure, possess high rigidity and minimal deformation under subsequent material flow; unformed oil cakes, on the other hand, suffer from weak inter-particle bonding and a loose structure, making them susceptible to secondary fragmentation or collapse under continued thrust. Therefore, static forming parameters extracted from a single static frame can be misjudged due to transient texture interference. By introducing temporal consistency constraints, static results can be dynamically corrected, thereby improving the accuracy and robustness of forming state identification.

[0053] Based on this logic, the changes of the oil cake image at consecutive moments can be used to obtain the dynamic shaping parameters of the oil cake within the preset period. From the moment to the Taking a moment as an example, the steps for obtaining the dynamic forming parameters of the oil cake within the preset period mainly include: Step 1: First, use the method of obtaining the static forming parameters of the oil cake to obtain the To All the moments Static forming parameters at a certain moment.

[0054] Step 2: Use the To All the moments Get all the static molding parameters at each moment The timing difference of static forming parameters at each moment is Take the moment as an example, the corresponding time series difference index The calculation method is as follows:

[0055] in, 、 and Respectively represent , No. Hedi The static forming parameters of the oil cake at each moment, .

[0056] The above time series difference indicators The calculation logic is: The time series difference index at the moment is used to describe the The shape change of the oil cake at a certain moment in the previous and subsequent moments, The larger the value, the The shape of the oil cake at the moment is significantly different from that at the preceding and following moments. The static forming parameters of the oil cake at each moment may be inaccurately described. The smaller the value, the The moment is when the shape of the oil cake is more stable.

[0057] By processing the static forming parameters of all oil cakes at time T' using the above method, the time series difference indexes of the static forming parameters at all T' times can be obtained.

[0058] Step 3: Perform cluster analysis on the time series difference indicators of all static molding parameters to obtain abnormal clusters. The main steps include: (1) With all The timing difference indicators of the static forming parameters corresponding to each moment constitute a timing difference indicator sequence, which is as follows: ; (2) Use K-means clustering to cluster the time series difference indicator sequence. The number of clusters can be set to two. After clustering, two clusters can be obtained. The cluster with fewer samples in the cluster is regarded as an abnormal cluster, and the other cluster is regarded as a normal cluster. The meaning of the above clusters is: at time It is not necessarily the case that the oil cakes at every moment are formed or unformed, and there is a possibility of shape change. Therefore, by clustering the time series difference sequence at consecutive moments, if all the If there is a morphological change at a certain moment, then there must be a corresponding time series difference value that has a size change with the rest of the time series difference values, and the K-means clustering algorithm can be used to distinguish them; if there is no morphological change of the oil cake in this process, the gap between the abnormal cluster and the normal cluster after clustering is significantly smaller.

[0059] Step 4: Obtain the dynamic molding parameters of oil cake extrusion using abnormal clusters and oil cake morphological characteristics in continuous time. The main steps include: (1) First, the time series difference sequence is divided by using abnormal clusters to obtain continuous segments of the time series difference sequence. The specific division method is: Sequence The first data The counting starts until the previous data of the time difference value in the abnormal cluster is met, and then the counted time difference value paragraph is the first continuous segment. Then the time difference value in the abnormal cluster starts, and the counting is restarted until the previous data of the time difference value in the new abnormal cluster is met, and the counted time difference value paragraph is regarded as a new continuous segment. The time difference sequence can be segmented by using the above method, and continuous segments can be obtained.

[0060] (2) Then, the dynamic forming parameters of the oil cake extrusion of the current continuous segment are obtained by using the morphological characteristics of the oil cake in each continuous segment, mainly including: Taking the th continuous segment as an example, the acquisition method of the dynamic morphological characteristics of the oil cake is as follows: first, the tile segmentation algorithm is used to obtain the tile region at each time in the continuous segment; then, the th time in the th continuous segment is taken as the start, the similarity of the tile region of the th time and all the remaining times in the th continuous segment is detected (the similarity can be obtained by using the perceptual hashing algorithm), and the average of all the similarities is taken as the morphological parameter of the first time . Then, the morphological parameter of each time is calculated in the above manner to obtain the morphological parameter of each time. It should be noted that each time is only calculated backwardly. Taking the th time as an example, when calculating the morphological parameter, it does not calculate the similarity with the times before the th time, but only calculates the similarity with the times after the th time (where , and represents the total number of all times in the th continuous segment).

[0061] The above method can obtain morphological parameters, and then the is obtained by using all the morphological parameters. The calculation method is as follows:

[0062] wherein, represents the morphological parameter of the oil cake extrusion of the th time in the th continuous segment.

[0063] The calculation logic of the above oil cake dynamic morphological characteristics is as follows: the morphological parameters of the times in the continuous segment As a basis, introduce time inverse proportional weight (i.e. similarity weighted weight) , so that the earlier the morphological parameters appear, the greater the weight. The weight design is based on the following mechanism: if the morphology at an earlier time remains high similarity, it indicates that the oil cake is continuously stable and has minimal deformation in that period, so it needs to be strengthened; if the morphology similarity at an earlier time is low, it implies that the oil cake has deformed during continuous extrusion, and this state is maintained in the subsequent period, so it needs to be compensated. The quantifies the morphological stability of the entire continuous segment: The greater, the more stable and the smaller the change in the morphology of the oil cake in that period; otherwise, it indicates that the morphology fluctuates dramatically.

[0064] Using the above method, the dynamic morphological characteristics of each continuous segment can be obtained.

[0065] Then, the dynamic forming parameters are obtained, taking the first continuous segment as an example, the corresponding dynamic forming parameters are obtained as follows: First, obtain the static forming parameters of all time points in the first continuous segment, taking the first time point as an example, the corresponding static forming parameters are : second, using all possible static forming parameters in the first continuous segment to obtain the mean value of the static forming parameters of all time points in the first continuous segment; finally, use and to calculate , the calculation method is:

[0066] The calculation logic of the above calculation method is: The greater, the greater the possibility of the larger time point in the first continuous segment under static observation; The greater, the more stable the morphology of the oil cake at each time point in the first continuous segment in the subsequent time. Therefore, the greater the two, the greater the possibility of oil cake forming in the first continuous segment, and the better the continuity.

[0067] Finally, the dynamic forming parameters of the current preset period are obtained through all continuous segments, and the specific acquisition method is:

[0068] wherein, represents the static forming parameter of the i th moment in the j th continuous segment; represents the dynamic forming parameter of the j th continuous segment.

[0069] The calculation logic of the above calculation formula is: the dynamic forming parameter of each continuous segment is weighted to obtain the mean value. Taking the j th continuous segment as an example, the consistency weight can be represented as The purpose is to determine whether the oil cake morphology in the j th continuous segment is consistent with the i th continuous segment. The first moment in the j th continuous segment is the data in the abnormal cluster, that is, there is a difference in the static morphology of the oil cake between the moment in the j th continuous segment and the i th continuous segment, so the static forming parameter of the first moment in the j th continuous segment is compared with the mean value of all static forming parameters in the i th continuous segment. The closer the ratio is to 1, the closer the morphology of all moments in the j th continuous segment is to the morphology of the i th moment, and the consistency weight Therefore, the greater the value of the dynamic forming parameter

[0070] It should be noted that since the last continuous segment is the current moment, the weight value cannot be calculated and compared, so the default weight value is 1.

[0071] By using the above method, the dynamic forming parameters of the rapeseed oil production line in different time periods during production can be obtained.

[0072] At this point, the dynamic forming parameters of the oil cake in the rapeseed oil production process are obtained.

[0073] ​​​​​​​​​​​​​​​​It needs to be further explained that the K-means clustering is an unsupervised iterative optimization algorithm. The tile segmentation algorithm refers to a calculation method for dividing the oil cake image into a plurality of local homogeneous regions (tiles), such as a superpixel segmentation algorithm. The perceptual hashing algorithm is an algorithm for mapping image content to a compact binary or integer fingerprint, which is commonly used for image deduplication, copyright detection, and similarity evaluation. The K-means clustering, tile segmentation algorithm, and perceptual hashing algorithm are all relatively mature and well-known technologies, and their specific implementation methods can be found in related technologies, which will not be described hereinafter.

[0074] Step S140: determining the molding state of the oil cake based on the dynamic molding parameter.

[0075] Preferably, in an embodiment of the present application, the step S140 can include: normalizing the dynamic molding parameter of the oil cake in a preset period; if the normalized dynamic molding parameter is greater than a preset molding threshold, determining that the molding state of the oil cake is molding; and if the normalized dynamic molding parameter is not greater than the preset molding threshold, determining that the molding state of the oil cake is unmolding.

[0076] Step S150: performing a rapeseed oil production line maintenance alarm when the molding state is unmolding.

[0077] It needs to be explained that after obtaining the static molding parameter and the dynamic molding parameter of the oil cake in the rapeseed oil production process, the dynamic molding parameter can be used for real-time monitoring of the rapeseed oil production line, and the main steps include: Step one, judging the shape of the oil cake by the dynamic molding parameter in the rapeseed oil production process in a preset period, and the specific judgment method is as follows: first, normalizing the dynamic molding parameter of the rapeseed oil cake in a preset period; and then establishing a parameter corresponding oil cake shape model based on experience, and the experience model in a certain application scenario is shown in Table 1: Table 1 Experience model

[0078] Then, the threshold method is used for judgment, and the molding threshold used in the embodiment of the present application is 0.85. When the dynamic molding parameter of the oil cake in a preset period is less than 0.85, it is considered that the oil cake is not molded.

[0079] For the molded oil cake, it is determined that the rapeseed moisture, roasting time and temperature, machine inner wall, uniformity of the feed, and operation parameters in the rapeseed oil production line are normal.

[0080] For the unmolding oil cake, it is considered that the rapeseed moisture, roasting time and temperature, machine inner wall, uniformity of the feed, and operation parameters in the rapeseed oil production line are abnormal, and the rapeseed oil production line needs to be maintained.

[0081] Referring to Figure 2 It shows a rapeseed oil production line real-time monitoring system 200 provided by another embodiment of the application, which comprises an image acquisition device 210 and a host computer 220, wherein: The image acquisition device 210 is arranged at a cake discharging port of the rapeseed oil production line, and is used to acquire real-time oil cake images and send the oil cake images to the host computer 220; The host computer 220 is used to acquire a time sequence of oil cake images of the rapeseed oil production line in a running process based on the oil cake images acquired by the image acquisition device, determine static forming parameters of the oil cake in multiple oil cake images in a preset period, wherein the static forming parameters are used to represent forming probabilities of a single oil cake image at an acquisition time of the single oil cake image, determine a dynamic forming parameter of the oil cake in the preset period based on the static forming parameters corresponding to the multiple oil cake images in the preset period, wherein the dynamic forming parameter is used to represent an accumulated credibility of the oil cake keeping a forming state in the preset period, determine the forming state of the oil cake based on the dynamic forming parameter, and perform rapeseed oil production line maintenance alarm when the forming state is unformed.

[0082] Thus, the application is completed.

[0083] In summary, in the embodiment of the present application, the oil cake time sequence image sequence in the running process of the rapeseed oil production line is acquired; the static forming parameters of the oil cake in the multiple oil cake images in a preset period are determined; wherein the static forming parameters are used to represent the oil cake forming probability of a single oil cake image at the acquisition time; based on the static forming parameters corresponding to the multiple oil cake images in the preset period, the dynamic forming parameters of the oil cake in the preset period are determined; wherein the dynamic forming parameters are used to represent the cumulative credibility of the oil cake keeping the forming state in the preset period; based on the dynamic forming parameters, the forming state of the oil cake is determined; when the forming state is unformed, the rapeseed oil production line maintenance alarm is performed. The double-layer feature system of the static forming parameters and the dynamic forming parameters makes the oil cake forming state judgment upgrade from single-frame instantaneous estimation to continuous period cumulative credibility evaluation, reduces the misjudgment rate caused by instantaneous texture or light disturbance, and improves the accuracy of the rapeseed oil production line abnormality detection; on the other hand, the oil cake image is converted into a tile segmentation region by using superpixel segmentation, and the regional static forming parameters are constructed on two low-dimensional features of area and brightness, and then coupled by global weight, which not only compresses the data volume, but also retains the key morphological information, reduces the computing load and storage demand of the upper computer; on the other hand, the preset period is continuously divided into segments based on the time sequence difference index and the abnormal cluster obtained by clustering, and the time-decreasing similarity weighting weight is introduced in each segment, which can effectively distinguish between temporary disturbance and continuous abnormality, avoid invalid shutdown, and improve the continuous running time of the production line; on the other hand, by normalizing the comparison between the dynamic forming parameters and the preset forming threshold, the automatic judgment of the forming state and the unformed state is realized, and the maintenance alarm is triggered immediately when it is unformed, which shortens the response time from abnormality occurrence to manual intervention, reduces the economic loss caused by the risk of increasing residual oil rate and equipment blockage.

[0084] The above only describes the preferred embodiments of the present application and should not be used to limit the present application. Any modification, equivalent replacement, improvement, etc. within the principles of the present application should be included in the protection scope of the present application.

Claims

1. A rapeseed oil production line real-time monitoring method, characterized in that, The method comprises: Obtain a time-series image sequence of oil cakes during the operation of a rapeseed oil production line; Determining static forming parameters of oil cakes in a plurality of oil cake images within a preset time period; wherein the static forming parameters are used to characterize the probability of oil cake formation of a single oil cake image at the time of its acquisition; Determining a dynamic forming parameter of the oil cake within the preset time period based on the static forming parameters corresponding to the plurality of oil cake images within the preset time period; wherein the dynamic forming parameter is used to represent the cumulative reliability of the oil cake maintaining the formed state within the preset time period; determining a forming state of the oil cake based on the dynamic forming parameters; When the molding state is unmolded, a rapeseed oil production line maintenance alarm is performed.

2. The rapeseed oil production line real-time monitoring method according to claim 1, wherein Determining the static forming parameters of the oil cake in a single oil cake image includes: Performing superpixel segmentation on the oil cake image to obtain multiple tile segmentation regions; respectively extracting area features and brightness features of each of the tile segmentation regions, and determining regional static shaping parameters of each of the tile segmentation regions based on the area features and the brightness features; Average processing is performed on the regional static forming parameters of a plurality of the tile segmentation regions to obtain the static forming parameters of the oil cake in the oil cake image.

3. The rapeseed oil production line real-time monitoring method according to claim 2, wherein Determining the regional static shaping parameters of a single tile segmentation region includes: Determining a first local formation probability based on the area feature of the tile segmentation region; wherein the first local formation probability is used to characterize the local formation probability of the oil cake in the tile segmentation region determined based on the area feature; Determining a second local formation probability based on a brightness feature of the tile segmentation region; wherein the second local formation probability is used to represent a local formation probability of the oil cake in the tile segmentation region determined based on the brightness feature; determining a first weight value based on the area feature distribution of all the tile segmentation regions in the oil cake image in which the tile segmentation region is located; wherein the first weight value is used to represent the overall formation probability of the oil cake in the oil cake image determined based on the area feature distribution; determining a second weight value based on a brightness feature distribution of all the tile segmentation regions in the oil cake image in which the tile segmentation region is located; wherein the second weight value is used to represent an overall probability of the oil cake being formed in the oil cake image, determined based on the brightness feature distribution; Based on the first weight value and the second weight value, a weighted sum is performed on the first local forming probability and the second local forming probability to obtain the regional static forming parameter of the tile segmentation region.

4. The rapeseed oil production line real-time monitoring method according to claim 1, wherein The determining of the dynamic forming parameters of the oil cake within the preset time period based on the static forming parameters corresponding to the plurality of oil cake images within the preset time period includes: Determining a temporal difference index between two adjacent oil cake images based on the static forming parameters corresponding to the plurality of oil cake images within the preset time period; wherein the temporal difference index is used to characterize the degree of mutation of the static forming parameters between the two oil cake images; Clustering the time series difference indicators to determine abnormal clusters; Based on the abnormal cluster, dividing the preset time period into a plurality of continuous segments; Based on the oil cake morphological characteristics of the oil cakes in a plurality of the continuous segments, dynamic forming parameters of the oil cakes in the preset time period are determined.

5. The rapeseed oil production line real-time monitoring method according to claim 4, wherein Clustering the time series difference indicators to determine abnormal clusters includes: Clustering the plurality of time series difference indicators to obtain a first cluster and a second cluster; An abnormal cluster and a normal cluster are determined in the first cluster and the second cluster; wherein the cluster containing a smaller number of samples is the abnormal cluster, and the cluster containing a larger number of samples is the normal cluster.

6. The rapeseed oil production line real-time monitoring method according to claim 4, wherein The step of dividing the preset time period into a plurality of continuous segments based on the abnormal cluster includes: Taking the start time of the preset time period as the starting point, sequentially determining the clusters to which the time series difference indicators belong; When the time series difference index belonging to the abnormal cluster is determined for the first time, the previous moment of the time series difference index is used as the end point of the current continuous segment; Taking the next moment of the time series difference indicator as the starting point of the next continuous segment, and determining the end point of the continuous segment; The above process is repeated until the entire preset time period is traversed to obtain a plurality of continuous segments.

7. The rapeseed oil production line real-time monitoring method according to claim 4, wherein: The determining of the dynamic forming parameters of the oil cake within the preset time period based on the oil cake morphological characteristics of the oil cake in the plurality of continuous segments includes: For each of the continuous segments, performing tile segmentation on the oil cake image at each moment in the continuous segment to obtain a plurality of tile regions; For each tile region, calculating a tile similarity between the tile region and each tile region at a target moment, and obtaining a morphological parameter of each tile region; wherein the target moment is used to represent all subsequent moments in the continuous segment that are located at a moment corresponding to the oil cake image in the tile region; and the morphological parameter is a mean value of the tile similarity of the tile region; Based on the similarity weighted weight, a weighted sum is performed on the morphological parameters of each tile region in the continuous segment to obtain the dynamic shaping parameter of the continuous segment; wherein the similarity weighted weight is used to represent the degree of continuous influence of the oil cake image in the tile region in the continuous segment; Based on the oil cake morphological consistency weight, the dynamic forming parameters of each continuous segment are weightedly summed and averaged to determine the dynamic forming parameters of the oil cake within the preset time period; wherein, the morphological consistency weight is used to characterize the degree of consistency in the oil cake morphology between two adjacent continuous segments.

8. The method for real-time monitoring of a rapeseed oil production line according to any one of claims 1 to 7, characterized in that: The determining the forming state of the oil cake based on the dynamic forming parameters includes: performing normalization processing on the dynamic forming parameters of the oil cake within the preset time period; If the normalized dynamic forming parameter is greater than a preset forming threshold, determining that the forming state of the oil cake is formed; If the normalized dynamic forming parameter is not greater than the preset forming threshold, it is determined that the forming state of the oil cake is unformed.

9. The method for real-time monitoring of a rapeseed oil production line according to any one of claims 1 to 7, characterized in that: After obtaining the oil cake time-series image sequence of the rapeseed oil production line during operation, the method further includes: The oil cake images in the oil cake time series image sequence are preprocessed; wherein the preprocessing includes at least one of filtering processing, grayscale processing and threshold segmentation processing.

10. A rapeseed oil production line real-time monitoring system, characterized in that: include: Image acquisition device and host computer, including: The image acquisition device is arranged at the cake discharge port of the rapeseed oil production line, and is used to collect real-time oil cake images and send the oil cake images to the host computer; The host computer is used to obtain a time-series image sequence of the oil cakes of the rapeseed oil production line during operation based on the oil cake images captured by the image acquisition device; determine the static forming parameters of the oil cakes in multiple oil cake images within a preset time period; wherein the static forming parameters are used to characterize the probability of oil cake formation of a single oil cake image at the time of its acquisition; determine the dynamic forming parameters of the oil cakes within the preset time period based on the static forming parameters corresponding to multiple oil cake images within the preset time period; wherein the dynamic forming parameters are used to characterize the cumulative credibility of the oil cakes maintaining a formed state within the preset time period; determine the forming state of the oil cakes based on the dynamic forming parameters; and issue a rapeseed oil production line maintenance alarm when the forming state is unformed.

Images