Sealed fermentation tank image acquisition method, system, equipment and medium

Abstract

The invention discloses a sealed fermentation tank image acquisition method, system and device and a medium, and relates to the field of image acquisition, the method comprises the following steps: acquiring an initial image in a sealed fermentation tank according to acquisition parameters; determining whether the initial image meets a set image requirement or not; when the initial image does not meet the set image requirements, adjusting the acquisition parameters, and obtaining the initial image again; and when the initial image meets a set image requirement, light reflection and light spot artifacts in the initial image are removed, and a final sealed fermentation tank image is obtained. According to the invention, clear and complete images can be collected in real time under the conditions that the light of the fermentation tank is insufficient and the material of the fermentation tank is easy to reflect light.

Description

Technical Field

[0001] The present application relates to the field of image acquisition, and in particular to a method, system, device and medium for acquiring images of a sealed fermentation tank. Background Art

[0002] As the main place for biochemical reactions such as microbial fermentation, real-time monitoring of the internal fermentation process is particularly important. The sealing of the fermentation tank is the key to ensuring the fermentation effect, preventing external pollution, and maintaining the stability of the fermentation environment. However, this sealing also brings challenges to the monitoring of the situation inside the tank. Considering the huge volume and metal structure of the fermentation tank, the traditional acquisition method faces many difficulties and often cannot provide accurate and complete monitoring data. First, due to the height of the fermentation tank, the bottom and middle areas are often blocked by the upper materials or bubbles generated by fermentation. The light gradually attenuates when penetrating the material, resulting in insufficient light in the bottom area, affecting the clarity and quality of the image; secondly, the metal characteristics of the fermentation tank also bring difficulties to image acquisition. Since the internal environment of the fermentation tank is usually dim, and the metal surface reflects light, glare or shadows are generated during shooting, and the liquid level fluctuations during the fermentation process make the collected images very easy to have light spots; finally, when the single imaging device collects images, due to the limitation of the viewing angle, it is impossible to fully cover all areas inside the fermentation tank, resulting in shooting blind spots. These problems make the traditional sealed fermentation tank image acquisition method unable to meet the requirements of modern factory production for data accuracy and real-time performance.

[0003] Therefore, how to collect clear and complete images in real time when the fermentation tank is insufficiently lit and its material is easily reflective is a difficult problem that needs to be solved urgently. Summary of the invention

[0004] The purpose of the present application is to provide a method, system, device and medium for acquiring images of a sealed fermentation tank, which can acquire clear and complete images in real time when the fermentation tank is insufficiently lit and its material is easily reflective.

[0005] To achieve the above objectives, this application provides the following solutions: In a first aspect, the present application provides a method for acquiring images of a sealed fermentation tank, comprising: An initial image of the interior of the sealed fermentation tank is collected according to collection parameters; the collection parameters include: image collection parameters, light source brightness and light source irradiation angle; the image collection parameters include collection frequency and shooting parameters; the shooting parameters include: shooting angle, exposure time, white balance and focal length; Determining whether the initial image meets set image requirements; When the initial image does not meet the set image requirements, adjusting the acquisition parameters and reacquiring the initial image; When the initial image meets the set image requirements, a removal model is used to remove reflection and light spot artifacts in the initial image to obtain a final sealed fermentation tank image; the removal model is constructed based on a dynamic reversible double residual network; Determine whether the liquid level in the sealed fermentation tank changes based on the image of the interior of the sealed fermentation tank acquired at the current acquisition frequency and the image of the interior of the sealed fermentation tank acquired at the previous acquisition frequency; When the liquid level changes, the grayscale entropy function is used to determine the position with the maximum image clarity evaluation value to achieve automatic focusing.

[0006] Optionally, the grayscale entropy function is expressed as: ; In the formula, represents the image clarity evaluation value, represents the gray level of the image, Indicates that the gray value of the image in the histogram is The probability of Represents binary bit entropy.

[0007] Optionally, the dynamic reversible dual residual network includes multiple residual modules; reversible residual connections are performed between the multiple residual modules.

[0008] Optionally, each of the residual modules includes a weight layer, an activation function layer and a shortcut connection structure; The weight layer is used to perform weighted summation operation on the input data; the activation function layer is used to introduce nonlinear characteristics; the shortcut connection structure is used to directly involve the input data of a certain layer structure in the dynamic reversible dual residual network in the output calculation of another layer structure.

[0009] In a second aspect, the present application provides a sealed fermentation tank image acquisition system, comprising: an image acquisition module, an illumination module and a processing terminal; a perspective window is provided on the sealed fermentation tank; The image acquisition module and the lighting module are both arranged on the sealed fermentation tank; the light emitted by the lighting module is emitted into the interior of the sealed fermentation tank through the perspective window; the image acquisition module acquires an image of the interior of the sealed fermentation tank through the perspective window; The image acquisition module and the lighting module are both connected to the processing terminal; The processing terminal is used to implement the sealed fermentation tank image acquisition method provided above, and based on the image of the inside of the sealed fermentation tank acquired by the image acquisition module, generates an adjustment instruction and obtains a final sealed fermentation tank image; the adjustment instruction is used to adjust the image acquisition parameters of the image acquisition module and the brightness of the lighting module and the illumination angle of the lighting module; the image acquisition parameters include acquisition frequency and shooting parameters; the shooting parameters include: shooting angle, exposure time, white balance and focal length.

[0010] Optionally, the image acquisition module includes multiple imaging devices and adjustment units corresponding to the multiple imaging devices one by one; the multiple imaging devices are arranged in an array; the imaging device is installed on the adjustment unit; the imaging device and the adjustment unit are both connected to the processing terminal; the processing terminal adjusts the shooting angle of the imaging device by adjusting the adjustment unit.

[0011] Optionally, the lighting module includes: multiple light sources and adjustment units corresponding to the multiple light sources one by one; the multiple light sources are arranged in an array; the light sources and the adjustment units are both connected to the processing terminal; the processing terminal adjusts the illumination angle of the lighting module by adjusting the adjustment unit.

[0012] In a third aspect, the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the sealed fermentation tank image acquisition method provided above.

[0013] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the sealed fermentation tank image acquisition method provided above.

[0014] In a fifth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of the above-mentioned sealed fermentation tank image acquisition method.

[0015] According to the specific embodiments provided in this application, this application has the following technical effects: The present application provides a sealed fermentation tank image acquisition method, system, device and medium, which determines whether an initial image meets the set image requirements. When the initial image does not meet the set image requirements, the acquisition parameters are adjusted until the acquired initial image meets the set image requirements, and the reflection and light spot artifacts in the initial image are removed to obtain the final sealed fermentation tank image, thereby being able to acquire a clear and complete image in real time when the fermentation tank is insufficiently lit and its material is easily reflective. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 A schematic diagram of a process for collecting images of a sealed fermentation tank provided in one embodiment of the present application; Figure 2 A structural design diagram of a sealed fermentation tank image acquisition system provided in one embodiment of the present application; Figure 3 This is a block diagram of the auto-focus principle provided by an embodiment of the present application; Figure 4 A schematic diagram of a mathematical model of a residual block in a deep residual network provided in an embodiment of the present application; Figure 5 A schematic diagram of the structure of a dynamic reversible double residual network provided in an embodiment of the present application; Figure 6 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0019] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0020] In an exemplary embodiment, a method for collecting images of a sealed fermentation tank is provided, such as Figure 1 As shown, including: Step 100: Acquire an initial image of the interior of the sealed fermentation tank according to acquisition parameters. The acquisition parameters include: image acquisition parameters, light source brightness, and light source irradiation angle. The image acquisition parameters include acquisition frequency and shooting parameters. The shooting parameters include: shooting angle, exposure time, white balance, and focal length.

[0021] Step 101: Determine whether the initial image meets the set image requirements.

[0022] Step 102: When the initial image does not meet the set image requirements, adjust the acquisition parameters and reacquire the initial image.

[0023] Step 103: When the initial image meets the set image requirements, a removal model is used to remove reflection and light spot artifacts in the initial image to obtain the final sealed fermentation tank image. The removal model is constructed based on a dynamic reversible dual residual network (Dr²Net).

[0024] The deep residual network is composed of a series of residual blocks plus some other layers. The deep residual network includes multiple residual blocks. The residual block in the deep residual network is composed of a direct mapping part and a residual part. The mathematical model is as follows Figure 4 As shown in the figure, the dynamic reversible dual residual network is designed based on the deep residual network, which has more stable and efficient training, better network performance and accuracy, and stronger model versatility and applicability.

[0025] The dynamic reversible dual residual network includes multiple residual modules (e.g. Figure 5 The F1, F2 and F3 modules shown in the figure); reversible residual connections are performed between multiple residual modules. The F1, F2 and F3 modules play a crucial role in optimizing the performance of the network. Figure 5 In the example, the dynamic adjustment coefficient α is used for the original residual connection, and its value is (0, 1). When α=1, the network initialization is close to a non-reversible network, which is convenient for using pre-trained parameters. During training, α gradually decreases, the influence of the original residual connection weakens, and the network is driven to transform into a reversible structure. Another dynamic adjustment coefficient β is applied to the newly added residual connection, and its value is (0, 1). It is set to 0.1 at initialization to avoid the problem of reverse calculation and make the structure close to a non-reversible network. During training, the value of β gradually increases, which enhances the role of the new residual connection, helps the network approach reversible characteristics, and controls reversibility and information flow. H(x)=F(x)+x is the core idea of ​​the residual connection, which allows the input x to directly participate in the calculation of the output, which helps to alleviate the gradient vanishing problem encountered by deep neural networks during training, so that the network can learn and train more effectively. Among them, F(x) is the result after processing by the normal weight layer and activation function layer, x is the input, and H (x) is the output.

[0026] Based on the above description, in the training process of deep neural networks, problems such as gradient disappearance are often encountered, which seriously affects the training efficiency and effect of the network. In order to solve this problem, each of the residual modules includes a weight layer, an activation function layer and a shortcut connection structure; The weight layer is composed of multiple neurons, and features are extracted by performing a weighted sum operation on the input data. In this process, the weights, as parameters that are constantly adjusted during the training process, can perform a linear transformation on the input data, thereby capturing key feature information in the data.

[0027] The activation function layer follows the weight layer, and its purpose is to introduce nonlinear characteristics to the dynamic reversible double residual network. In practical applications, if there is no activation function, the multi-layer neural network will degenerate into a single-layer linear network and cannot process complex data patterns. Based on this, a suitable activation function, such as ReLU, is used in the residual module. By performing nonlinear transformation on the output of the weight layer, the network's expression ability is greatly enhanced, enabling it to fit more complex functional relationships.

[0028] The introduction of the shortcut connection structure is to effectively alleviate the gradient vanishing problem in the dynamic reversible dual residual network. In the residual module, the shortcut connection structure allows the input of a certain layer to directly participate in the output calculation of the subsequent layer, specifically, the output can be expressed in the form of H (x) = F (x) + x. During the back-propagation process, the gradient can be directly propagated back to the previous layer through the shortcut connection, avoiding the gradient attenuation caused by the multiplication of multiple layers of weight matrices, ensuring that the dynamic reversible dual residual network can be trained more efficiently.

[0029] In addition, dynamic adjustment coefficients α and β can be introduced in the dynamic reversible double residual network, and these dynamic adjustment coefficients are associated with the connections in the residual module. During the training process of the dynamic reversible double residual network, the network performance can be further optimized by dynamically adjusting the values ​​of the dynamic adjustment coefficients α and β. Specifically, in the early stage of training, the initial values ​​of the dynamic adjustment coefficients α and β can be set according to a specific strategy, so that the dynamic reversible double residual network starts learning from a relatively simple state. As the training progresses, the dynamic adjustment coefficients α and β are dynamically adjusted according to the performance and training situation of the dynamic reversible double residual network. For example, the dynamic adjustment coefficient α can control the degree of influence of certain weight layers, and the dynamic adjustment coefficient β can adjust the strength of the shortcut connection structure. In this way, the structure and parameters of the dynamic reversible double residual network itself can be continuously optimized during the training process, and the configuration that best suits the current data can be found, thereby improving the convergence speed and accuracy of the network.

[0030] It can be seen that the dynamic reversible dual residual network provided in this application can effectively solve the key problems in the deep neural network training process and improve the overall performance.

[0031] Based on the above description, the removal model is constructed to remove the reflection and spot artifacts in the initial image, which has the following advantages: 1. Unique residual connection and reversible design.

[0032] (11) Dual residual connections: The dynamic reversible dual residual network contains two types of residual connections, one that maintains the traditional residual structure in the pre-trained model and the other that makes the network reversible. Traditional residual connections facilitate the rapid transmission of information and the effective back propagation of gradients, allowing the network to better learn the feature information in the image, including the features of reflections and light artifacts. The reversible residual connection allows the network to not lose any information during the forward propagation process, so that the gradient can be accurately calculated without recalculating the activation function during back propagation, which helps to more accurately adjust the network parameters to reduce reflections and light artifacts.

[0033] (12) Dynamic training strategy: During the training process, two dynamically adjustable coefficients are used for the two residual connections respectively, and a dynamic training strategy is introduced to enable the pre-trained model to seamlessly transition to a reversible network with higher numerical accuracy. This dynamic adjustment can flexibly adjust the weights of different residual connections according to the progress of training and the characteristics of the image, better adapting to the needs of eliminating reflection and spot artifacts.

[0034] 2. Efficient feature extraction and fusion: (21) Multi-scale feature extraction: The dynamic reversible dual residual network can extract features from images at different scales. Through the combination of multiple convolutional layers and pooling layers, it can capture different levels of features of reflections and light spot artifacts in the image, including information such as the size, shape, and position of the reflections and light spot artifacts.

[0035] (22) Feature fusion: The features extracted at different scales are fused so that the dynamic reversible double residual network can fully understand the reflection and light spot artifact information in the image, thereby more accurately locating and removing the reflection and light spot artifacts. During the fusion process, the dynamic reversible double residual network performs weighted fusion based on the importance and relevance of the features, highlighting the features related to reflection and light spot artifacts while suppressing the interference of other irrelevant features.

[0036] 3. Correction learning based on non-reflective and light spot artifact areas: (31) Distinguishing between reflection and light spot artifacts and non-reflection and light spot artifact areas: During the training process, the dynamic reversible double residual network learns how to distinguish between reflection and light spot artifact areas and non-reflection and light spot artifact areas in the image. By training on a large number of image pairs with and without reflection and light spot artifacts, the dynamic reversible double residual network can automatically identify the characteristic patterns of reflection and light spot artifacts, so as to accurately locate the reflection and light spot artifact areas when processing new images.

[0037] (32) Correcting the reflective and light-spot artifact areas using the non-reflective and light-spot artifact areas: According to the pixel values ​​of the pixels in the non-reflective and light-spot artifact areas, the pixel values ​​of the pixels in the reflective and light-spot artifact areas are adjusted. For example, the dynamic reversible dual residual network can learn the pixel value distribution rules of the non-reflective and light-spot artifact areas, and then correct the pixel values ​​of the reflective and light-spot artifact areas according to these rules, so that they are closer to the pixel values ​​of the non-reflective and light-spot artifact areas, thereby achieving the effect of reducing the reflective and light-spot artifacts.

[0038] In the exemplary embodiment of the present application, when collecting the internal image of the sealed fermentation tank, the liquid level changes caused by the fermentation of the liquid inside the sealed fermentation tank, resulting in a problem of blurred image. Therefore, in order to achieve automatic and accurate focusing, the sealed fermentation tank image collection method provided by the present application also includes: Step 104: determining whether the liquid level in the sealed fermentation tank changes based on the image of the interior of the sealed fermentation tank acquired at the current acquisition frequency and the image of the interior of the sealed fermentation tank acquired at the previous acquisition frequency.

[0039] Step 105: When the liquid level changes, the focus depth algorithm is used to realize automatic focus of image acquisition. According to Shannon's information theory, the amount of information is the largest when the entropy is the largest.

[0040] Compared with blurry images, clear images have more diverse grayscale distribution and more information. Assume that the image grayscale is According to the image grayscale distribution histogram, the grayscale value of the image in the histogram is calculated to be The probability of , then the grayscale entropy function of the image is expressed as:

[0041] In the formula, Indicates The logarithm with base is used to calculate the "information content" or "uncertainty" of each gray level probability. Represents binary bit entropy, usually .

[0042] When the grayscale entropy (i.e., image clarity evaluation value ) takes the maximum value, the image information is the largest, the image edge information and detail information are the richest, so the clearest imaging position can be obtained. Based on this, in this embodiment, when the liquid level changes, the grayscale entropy function is used to determine the position with the largest image clarity evaluation value to achieve automatic focus.

[0043] In the exemplary embodiment of the present application, in order to further improve the clarity of the image, the image reflection removal method based on position perception and the light spot artifact elimination method based on data synthesis can be used to eliminate the reflection and light spot artifacts in the initial image. In order to further improve the efficiency of image processing, a deep learning network can be used to implement the image reflection removal method based on position perception and the light spot artifact elimination method based on data synthesis. Specifically, the deep learning network receives the initial image as input and outputs an image after reflection removal and light spot elimination processing, so as to realize the monitoring of the fermentation tank. Based on this, the implementation process of the above step 103 of the present application includes: Step 1031: construct a removal model based on a deep learning network, for example: The reflection confidence map is used as a clue for the network (e.g., a deep learning network) to learn how to adaptively encode the reflection information and control the feature flow when predicting the reflection information and transmission information in the image. Integrating position information into the network can significantly improve the quality of the reflection removal results. Among them, reflection information refers to the information generated by the reflection of light. In the fermentation tank image acquisition, these reflection information needs to be distinguished in order to more accurately monitor the fermentation process. The formation and utilization of feature flows depend on the design and training of the network. The network forms these feature flows by learning how to extract useful information from the input image. These feature flows are then used for operations such as image quality assessment, focus adjustment, and reflection removal to achieve accurate monitoring of the fermentation process. Position information can be obtained through position encoding of image processing techniques (e.g., feature point recognition or deep learning models).

[0044] In addition, a set of learnable Laplace kernel parameters are introduced to facilitate the extraction of distinguishable Laplace features for reflection detection. The network is designed as a recurrent network to gradually optimize the reflection removal results of each iteration, and finally achieve the purpose of reflection removal. Among them, the introduction of learnable Laplace kernel parameters in the network is to enhance the network's ability to distinguish between reflection and transmission in the image. By learning these parameters, the network can better extract features from the image, especially when dealing with image problems with complex lighting conditions, such as reflection problems in fermentation tanks. Laplace features are closely related to the image data used in the network. They are key information extracted from the image or graph structure to enhance the network's feature recognition and processing capabilities.

[0045] The removal of the speckle artifacts uses a data synthesis method, which generates training inputs by randomly synthesizing natural images without flares and flare images, and trains the convolutional neural network to restore the flare-free scene. After the prediction is completed, the input light source is mixed into the output image. Among them, natural images without flares can be sampled from public datasets, and flare images can be modeled and generated through a physically realistic semi-synthetic pipeline.

[0046] Step 1032: Input the initial image into the removal model to remove reflection and light spot artifacts in the initial image to obtain the final sealed fermentation tank image.

[0047] Based on the same inventive concept, the embodiment of the present application also provides a sealed fermentation tank image acquisition system for implementing the sealed fermentation tank image acquisition method involved above. The implementation solution provided by the system to solve the problem is similar to the implementation solution recorded in the above method, so the specific limitations in one or more sealed fermentation tank image acquisition system embodiments provided below can refer to the limitations of the sealed fermentation tank image acquisition method above, and will not be repeated here.

[0048] In an exemplary embodiment, a sealed fermentation tank image acquisition system is provided, including: an image acquisition module, an illumination module, and a processing terminal. A perspective window is provided on the sealed fermentation tank. This perspective window is a special perspective window with the characteristics of high light transmittance, corrosion resistance, and high temperature resistance. For example, materials with excellent optical and chemical stability such as quartz glass are selected, combined with precision polishing and anti-reflective coating technology to improve light transmittance, and a high temperature and corrosion resistant sealing design and an easy-to-clean surface treatment are adopted to ensure that the situation inside the sealed fermentation tank can be clearly observed during the fermentation process, while ensuring the long-term stability and corrosion resistance of the perspective window. In addition, correct installation and regular maintenance are also key to ensuring the performance of the perspective window.

[0049] The image acquisition module and the lighting module are both arranged on the sealed fermentation tank. The light emitted by the lighting module is projected into the interior of the sealed fermentation tank through the perspective window. The image acquisition module acquires the image of the interior of the sealed fermentation tank through the perspective window.

[0050] The image acquisition module and the lighting module are both connected to the processing terminal.

[0051] The processing terminal is used to implement the sealed fermentation tank image acquisition method provided above, and based on the image of the inside of the sealed fermentation tank acquired by the image acquisition module, an adjustment instruction is generated and a final sealed fermentation tank image is obtained. The adjustment instruction is used to adjust the image acquisition parameters of the image acquisition module, the brightness of the lighting module, and the illumination angle of the lighting module. The image acquisition parameters include acquisition frequency and shooting parameters. The shooting parameters include: shooting angle, exposure time, white balance, and focal length.

[0052] As an optional embodiment, the image acquisition module includes a plurality of imaging devices (such as cameras) and an adjustment unit corresponding to the plurality of imaging devices. The plurality of imaging devices are arranged in an array. The imaging devices are mounted on the adjustment unit. The imaging devices and the adjustment unit are both connected to a processing terminal. The processing terminal adjusts the shooting angle of the imaging device by adjusting the adjustment unit.

[0053] As an optional embodiment, the lighting module includes: a plurality of light sources and an adjustment unit corresponding to the plurality of light sources. The plurality of light sources are arranged in an array. The light sources and the adjustment unit are both connected to a processing terminal. The processing terminal adjusts the illumination angle of the lighting module by adjusting the adjustment unit.

[0054] For example, placing long- and short-focus cameras and a cylindrical light source on the top of a sealed fermentation tank ensures that clear and detailed images can be captured even in the long-distance, low-light internal environment of the sealed fermentation tank, ensuring that the bottom of the tank or the liquid surface can be photographed.

[0055] In an exemplary embodiment of the present application, taking the sealed fermentation tank image acquisition system provided by the present application as an example, the implementation process of the sealed fermentation tank image acquisition method provided by the present application is described, including: Step 1: Initialize the acquisition system and preset standard shooting parameters.

[0056] Step 1.1: Detect the camera status and obtain the camera's current working status information.

[0057] Step 1.2: Detect the status of the light source to obtain the current working status information of the light source.

[0058] Step 1.3: Set the camera's exposure time, white balance, focal length and other shooting parameters.

[0059] Step 2: Collect the initial image inside the sealed fermentation tank through a special perspective window.

[0060] Step 2.1: Ensure that the high light transmittance, corrosion resistance, and high temperature resistance transparent window is correctly installed in the designated position of the sealed fermentation tank.

[0061] Step 2.2: Fix the long-focus camera and short-focus camera outside the perspective window respectively, adjust the position, angle and focal length of the camera to ensure that the camera field of view completely covers the area inside the sealed fermentation tank.

[0062] Step 2.3: Enable and adaptively adjust the light source so that the light evenly illuminates the area inside the sealed fermentation tank through the perspective window to solve the problem of uneven image brightness caused by metal reflection or liquid surface light spots.

[0063] Step 3: Automatically adjust the brightness and orientation of the light source according to the quality of the acquired initial image.

[0064] Step 3.1: Control the state of the light source array, and adjust the brightness of the light source according to the brightness of the image based on the image quality assessment results. Image quality assessment is usually performed by calculating statistical characteristics of the image, such as contrast signal-to-noise ratio, grayscale clarity Tenengrad function, image brightness mean, and other indicators. When these indicators are lower than the preset threshold, indicating that the image quality does not meet the requirements, the system will automatically adjust the brightness of the light source. The adjustment process will continue until these quality indicators of the image reach or exceed the preset threshold, or the image brightness mean reaches the preset brightness level, at which time the brightness adjustment will stop to ensure that the image is both clear and has appropriate brightness.

[0065] Step 3.2: Adjust the illumination angle or position of the light source according to the brightness distribution or specific features of different areas in the initial image. This specific feature may include average brightness, brightness standard deviation, local brightness change, etc. When there are problems such as uneven brightness, overexposure or underexposure, insufficient lighting in key areas, etc., adjust the illumination angle or position.

[0066] Step 4: Automatically focus based on the liquid level changes to capture the best quality image.

[0067] Step 4.1: After the liquid surface condition is imaged, image acquisition and A / D conversion are completed together.

[0068] Step 4.2: Send the image data to the processing terminal to obtain the processing result, and use the processing result to control the driving adjustment unit to adjust the lens position to achieve the purpose of accurate focusing.

[0069] Step 5: Set the acquisition frequency to collect image data in the sealed fermentation tank in real time and transmit it to the processing terminal.

[0070] Step 5.1: Pre-set the collection frequency through the processing terminal to realize automatic collection of the fermentation process.

[0071] Step 5.2: Use a telephoto camera with a long shooting distance to capture detailed information on the bottom of the fermentation tank.

[0072] Step 5.3: Use a short-focus camera with a wide field of view to photograph the tank body information of the sealed fermentation tank after the liquid level rises, so as to collect complete liquid level information during the fermentation process.

[0073] In summary, the present application provides an efficient, stable and flexible sealed fermentation tank image acquisition system that can shoot the fermentation process in real time. By installing a shooting device (i.e., an image acquisition module) on the top of the sealed fermentation tank, the perspective window can be used to achieve continuous and real-time acquisition of the situation inside the tank without affecting the sealing. In order to capture a complete image inside the tank and solve the problem of metal surface reflection during shooting, a long- and short-focus dual camera and an adjustable light source module are integrated. The telephoto camera can magnify the details inside the fermentation tank and is suitable for shooting a specific area or a tank bottom at a distance. The short-focus camera has a wider viewing angle and can capture more scene information. The light source can be adjusted to provide brightness for a dark environment, and by training a neural network, a position-aware image reflection removal method and a data-synthesis-based spot artifact elimination method are used to reduce shadows and reflections caused by uneven light. In order to cope with the focusing problem caused by liquid level changes, an automatic focus method is introduced, which can monitor the height of the liquid level in real time and automatically adjust the focal length of the camera to ensure that clear and detailed images can be captured at any liquid level.

[0074] In the exemplary embodiment of the present application, Figure 2 As shown, the sealed fermentation tank image acquisition system provided by the present application can be described as a design architecture including a hardware layer and a software layer. Among them, the hardware layer includes an autofocus module, an image acquisition module and a lighting module. The software layer includes a control system, an internally implanted functional algorithm (such as a focus depth algorithm, i.e., an autofocus algorithm) and a user interface. Among them, the control system can include a focusing system, a camera control system and a light source control system.

[0075] The autofocus module uses a focus depth algorithm to set a threshold according to the monitored liquid level changes and automatically adjust the camera's focus. The autofocus module can be integrated with the image acquisition module to form a camera that can achieve automatic zoom. The image acquisition module can also use a high-performance telephoto array camera to ensure that clear and detailed images can be captured even in long-distance, low-light tank environments.

[0076] The lighting module may be designed with a light source group that can provide sufficient and uniform light to ensure that each area in the sealed fermentation tank can be clearly photographed and reduce the generation of light spots and shadows.

[0077] In practical applications, autofocus can be achieved through a focus depth algorithm. After the liquid surface is imaged by the camera lens, image acquisition and A / D (Analog Digital) conversion are completed on the image sensor. Subsequently, DSP (Digital Signal Process) data processing is performed to obtain evaluation results, and the processing results are used to control the drive circuit to adjust the camera lens position. The principle block diagram is shown below. Figure 3shown.

[0078] In addition, the functions implemented by the autofocus module, the control system, the internally embedded functional algorithms and the user interface (UI) may correspond to the processing terminal provided above.

[0079] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 6 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store sealed fermentation tank image acquisition data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a sealed fermentation tank image acquisition method is implemented.

[0080] Those skilled in the art will understand that Figure 6 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components. In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0081] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0082] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0083] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0084] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0085] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0086] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0087] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for collecting images of a sealed fermentation tank, characterized in that: The sealed fermentation tank image acquisition method comprises: An initial image of the interior of the sealed fermentation tank is collected according to collection parameters; the collection parameters include: image collection parameters, light source brightness and light source irradiation angle; the image collection parameters include collection frequency and shooting parameters; the shooting parameters include: shooting angle, exposure time, white balance and focal length; Determining whether the initial image meets set image requirements; When the initial image does not meet the set image requirements, adjusting the acquisition parameters and reacquiring the initial image; When the initial image meets the set image requirements, a removal model is used to remove reflection and light spot artifacts in the initial image to obtain a final sealed fermentation tank image; the removal model is constructed based on a dynamic reversible double residual network; Determine whether the liquid level in the sealed fermentation tank changes based on the image of the interior of the sealed fermentation tank acquired at the current acquisition frequency and the image of the interior of the sealed fermentation tank acquired at the previous acquisition frequency; When the liquid level changes, the grayscale entropy function is used to determine the position with the maximum image clarity evaluation value to achieve automatic focusing.

2. The sealed fermentation tank image acquisition method according to claim 1, characterized in that: The grayscale entropy function is expressed as: ； In the formula, represents the image clarity evaluation value, represents the gray level of the image, Indicates that the gray value of the image in the histogram is The probability of Represents binary bit entropy.

3. The sealed fermentation tank image acquisition method according to claim 1, characterized in that: The dynamic reversible dual residual network includes multiple residual modules; reversible residual connections are performed between the multiple residual modules.

4. The sealed fermentation tank image acquisition method according to claim 3, characterized in that: Each of the residual modules comprises a weight layer, an activation function layer and a shortcut connection structure; The weight layer is used to perform weighted summation operation on the input data; the activation function layer is used to introduce nonlinear characteristics; the shortcut connection structure is used to directly involve the input data of a certain layer structure in the dynamic reversible dual residual network in the output calculation of another layer structure.

5. A sealed fermentation tank image acquisition system, characterized in that: A perspective window is provided on the sealed fermentation tank; the sealed fermentation tank image acquisition system comprises: an image acquisition module, a lighting module and a processing terminal; The image acquisition module and the lighting module are both arranged on the sealed fermentation tank; the light emitted by the lighting module is emitted into the interior of the sealed fermentation tank through the perspective window; the image acquisition module acquires an image of the interior of the sealed fermentation tank through the perspective window; The image acquisition module and the lighting module are both connected to the processing terminal; The processing terminal is used to implement the sealed fermentation tank image acquisition method as described in any one of claims 1 to 4, and based on the image of the inside of the sealed fermentation tank acquired by the image acquisition module, generate adjustment instructions and obtain a final sealed fermentation tank image; the adjustment instructions are used to adjust the image acquisition parameters of the image acquisition module and the brightness of the lighting module and the illumination angle of the lighting module; the image acquisition parameters include acquisition frequency and shooting parameters; the shooting parameters include: shooting angle, exposure time, white balance and focal length.

6. The sealed fermentation tank image acquisition system according to claim 5, characterized in that: The image acquisition module includes multiple imaging devices and adjustment units corresponding to the multiple imaging devices one by one; the multiple imaging devices are arranged in an array; the imaging devices are installed on the adjustment units; the imaging devices and the adjustment units are both connected to the processing terminal; the processing terminal adjusts the shooting angle of the imaging device by adjusting the adjustment units.

7. The sealed fermentation tank image acquisition system according to claim 5, characterized in that: The lighting module includes: a plurality of light sources and adjustment units corresponding to the plurality of light sources one by one; the plurality of light sources are arranged in an array; the light sources and the adjustment units are both connected to the processing terminal; the processing terminal adjusts the illumination angle of the lighting module by adjusting the adjustment units.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the sealed fermentation tank image acquisition method according to any one of claims 1 to 4.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the sealed fermentation tank image acquisition method according to any one of claims 1 to 4 is implemented.

Images