Zoom trigger recognition method and system based on optical adapter

Abstract

The present invention discloses a zoom trigger identification method and system based on an optical adapter, which relates to the field of zoom control technology. The method includes: obtaining a target zoom optical adapter; obtaining a first endoscopic image output by an external imaging device; traversing the first endoscopic image to perform convolution similarity comparison to obtain a first convolution similarity index; if the similarity is greater than a first predetermined similarity, triggering a zoom control module to perform local zoom on the first endoscopic image and obtain motion information of a moving mechanism of the target zoom optical adapter; performing mechanical error identification on the moving mechanism to obtain a first mechanical motion deviation; performing feedback optimization on a first predetermined zoom parameter, and the zoom control module triggering zoom with the first optimized zoom parameter to obtain a first zoom image. The present invention solves the technical problems of limited zoom function and inaccurate image zoom control in the prior art optical adapter, thereby achieving the technical effect of improving the quality and stability of endoscopic images.

Description

Technical Field

[0001] The present invention relates to the field of zoom control technology, and in particular to a zoom trigger recognition method and system based on an optical adapter. Background Art

[0002] With the widespread application of endoscopy technology, the requirements for image quality and recognition accuracy are becoming increasingly higher. Traditional endoscopic image acquisition and recognition methods often suffer from problems such as limited zoom capabilities and imprecise image zoom control, making it difficult to meet the needs of modern complex and sophisticated application scenarios. Summary of the Invention

[0003] The present application provides a zoom trigger recognition method and system based on an optical adapter, which is used to solve the technical problems of limited zoom function and inaccurate image zoom control in the prior art optical adapters.

[0004] In view of the above problems, the present application provides a zoom trigger recognition method and system based on an optical adapter.

[0005] In a first aspect of the present application, a zoom trigger recognition method based on an optical adapter is provided, the method comprising:

[0006] A target zoom optical adapter is obtained, wherein the target zoom optical adapter includes a plurality of lenses with a zoom function and is used to connect a target endoscope to an external imaging device; a first endoscopic image output by the external imaging device is obtained; a convolution similarity comparison is performed on the first endoscopic image using a predetermined abnormal convolution feature to obtain a first convolution similarity index, wherein the first convolution similarity index includes a first image region identifier; if the first convolution similarity index is greater than a first predetermined similarity, a zoom control module of the target zoom optical adapter is triggered to locally zoom the first endoscopic image according to a first predetermined zoom parameter based on the first image region identifier, and motion information of a moving mechanism of the target zoom optical adapter is monitored and obtained; a mechanical error of the moving mechanism is identified based on the first predetermined zoom parameter and the motion information of the moving mechanism to obtain a first mechanical motion deviation; the first predetermined zoom parameter is feedback-optimized based on the first mechanical motion deviation, and the zoom control module triggers zooming using the first optimized zoom parameter to obtain a first zoom image.

[0007] A second aspect of the present application provides a zoom trigger recognition system based on an optical adapter, the system comprising:

[0008] a target zoom optical adapter acquisition unit, the target zoom optical adapter acquisition unit acquires a target zoom optical adapter, wherein the target zoom optical adapter includes a plurality of lenses with a zoom function, and the target zoom optical adapter is used to connect a target endoscope to an external imaging device; a first endoscopic image output unit, the first endoscopic image output unit acquires a first endoscopic image output by the external imaging device; a first convolution similarity index acquisition unit, the first convolution similarity index acquisition unit traverses the first endoscopic image with a predetermined abnormal convolution feature to perform a convolution similarity comparison to obtain a first convolution similarity index, and the first convolution similarity index has a first image area identifier; a motion information acquisition unit, the motion information acquisition unit is used to obtain a first convolution similarity index if the first A convolution similarity index is greater than a first predetermined similarity, triggering a zoom control module of the target zoom optical adapter to locally zoom the first endoscopic image according to a first predetermined zoom parameter based on the first image area identifier, and monitoring and acquiring motion information of a moving mechanism of the target zoom optical adapter; a first mechanical motion deviation acquisition unit, which identifies a mechanical error of the moving mechanism based on the first predetermined zoom parameter and the motion information of the moving mechanism to obtain a first mechanical motion deviation; and a first zoom image acquisition unit, which performs feedback optimization on the first predetermined zoom parameter based on the first mechanical motion deviation. The zoom control module triggers zooming with the first optimized zoom parameter to obtain a first zoom image.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0010] The present application obtains a target zoom optical adapter, wherein the target zoom optical adapter includes multiple lenses with zoom functions, and the target zoom optical adapter is used to connect a target endoscope to an external imaging device; obtains a first endoscopic image output by the external imaging device; traverses the first endoscopic image with a predetermined abnormal convolution feature to perform a convolution similarity comparison to obtain a first convolution similarity index, and the first convolution similarity index has a first image area identifier; if the first convolution similarity index is greater than a first predetermined similarity, triggers a zoom control module of the target zoom optical adapter, performs local zooming on the first endoscopic image according to a first predetermined zoom parameter based on the first image area identifier, and monitors and obtains motion information of a moving mechanism of the target zoom optical adapter; performs mechanical error identification of the moving mechanism based on the first predetermined zoom parameter and the motion information of the moving mechanism to obtain a first mechanical motion deviation; feedback optimizes the first predetermined zoom parameter based on the first mechanical motion deviation, and the zoom control module triggers zooming with the first optimized zoom parameter to obtain a first zoom image. The present invention solves the technical problems of limited zoom function and inaccurate image zoom control in optical adapters in the prior art. By combining technologies such as zoom control, image recognition and mechanical error recognition, the present invention achieves the technical effect of improving the image quality and stability of endoscopes. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0012] Figure 1 A flowchart of a zoom trigger recognition method based on an optical adapter provided in an embodiment of the present application;

[0013] Figure 2 A schematic structural diagram of a zoom trigger recognition system based on an optical adapter provided in an embodiment of the present application.

[0014] Explanation of reference numerals: target zoom optical adapter acquiring unit 11 , first endoscopic image output unit 12 , first convolution similarity index acquiring unit 13 , motion information acquiring unit 14 , first mechanical motion deviation acquiring unit 15 , first zoom image acquiring unit 16 . DETAILED DESCRIPTION

[0015] This application provides a zoom trigger recognition method and system based on an optical adapter to solve the technical problems of limited zoom function and inaccurate image zoom control in the existing optical adapter. By combining technologies such as zoom control, image recognition and mechanical error recognition, the technical effect of improving the image quality and stability of the endoscope is achieved.

[0016] The following will be combined with the accompanying drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making any creative work are within the scope of protection of this application.

[0017] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products or devices.

[0018] Example 1

[0019] like Figure 1 As shown, the present application provides a zoom trigger recognition method based on an optical adapter, the method comprising:

[0020] Step S100: obtaining a target zoom optical adapter, wherein the target zoom optical adapter comprises a plurality of lenses with a zoom function, and the target zoom optical adapter is used to connect a target endoscope with an external imaging device;

[0021] In the embodiments of the present application, the specific specifications and performance requirements of the zoom optical adapter are determined based on factors such as the type of endoscope, the application scenario, and the required zoom range and accuracy. For example, some applications may require a high-magnification zoom function to observe minute details, while other scenarios may prioritize smooth and accurate zooming. A target zoom optical adapter is then determined based on the specific specifications and performance requirements of the zoom optical adapter. The target zoom optical adapter seamlessly connects to the target endoscope and external imaging equipment.

[0022] Step S200: Acquire a first endoscopic image output by the external imaging device;

[0023] In an embodiment of the present application, the external imaging device is turned on and its settings, such as image resolution and color mode, are adjusted as needed. The output interface of the endoscope is then connected to the input interface of the external imaging device. The endoscope is controlled to enter the observation area, and the image capture function of the endoscope is activated to begin capturing images of the target area. The external imaging device receives image data from the endoscope and performs a series of processing operations on the image data, such as noise suppression, contrast enhancement, and color correction, to improve image quality.

[0024] Finally, the external imaging device generates a wide-field-of-view image covering a larger area based on the image range captured by the endoscope, and outputs the wide-field-of-view image as the first endoscopic image.

[0025] Step S300: performing convolution similarity comparison on the first endoscopic image using a predetermined abnormal convolution feature to obtain a first convolution similarity index, wherein the first convolution similarity index has a first image region identifier;

[0026] In an embodiment of the present application, the predetermined abnormal convolution feature traversal is pre-set through expert experience and represents a possible abnormality pattern. The first endoscopic image is traversed using the predetermined abnormal convolution feature. The traversal process is to slide a window of the same size as the convolution feature on the image and perform a convolution operation on each window position. During the traversal process, for each window position, the convolution operation result between the predetermined abnormal convolution feature and the image at the current window position is calculated. The convolution operation measures the degree of overlap between two signals or images and determines the similarity between them. By comparing the results of the convolution operation, the area similar to the predetermined abnormal feature is found. Based on the result of the convolution operation, the first convolution similarity index is calculated by cosine similarity or the like. The first convolution similarity index measures the degree of similarity between a certain area in the first endoscopic image and the predetermined abnormal feature.

[0027] At the same time, a first image region identifier is generated to identify the image region that is most similar to the predetermined abnormal feature. The first image region identifier is a two-dimensional matrix or image of the same size as the image, in which the region similar to the predetermined abnormal feature is marked by highlighting or assigning a specific pixel value.

[0028] Step S400: If the first convolution similarity index is greater than a first predetermined similarity, triggering a zoom control module of the target zoom optical adapter to locally zoom the first endoscopic image according to a first predetermined zoom parameter based on the first image region identifier, and monitoring and acquiring motion information of a moving mechanism of the target zoom optical adapter;

[0029] In an embodiment of the present application, when the first convolution similarity index is greater than a first predetermined similarity, it indicates that regions in the first endoscopic image are highly similar to predetermined abnormal features, and these regions may be regions of interest or lesion areas. In this case, the zoom control module of the target zoom optical adapter is triggered to perform a local zoom on the first endoscopic image. The zoom control module is one of the core components of the target zoom optical adapter and is responsible for controlling the movement of the lens group to achieve the zoom function. When the first convolution similarity index exceeds a threshold, a trigger signal is sent to the zoom control module.

[0030] When performing local zoom, based on a first predetermined zoom parameter, such as a certain zoom factor, the zoom control module calculates the movement distance of the lens on the optical axis and controls the moving mechanism to move the lens to a specified position to achieve local zoom on a specific area.

[0031] During the zoom process, sensors and other equipment monitor the movement of the moving mechanism in real time. This monitoring includes parameters such as the movement distance, speed, and position of the lens group. This ensures smooth and precise movement of the lens group to meet the requirements of local zoom. Based on this information, the zoom control module further adjusts the positions of the zoom and compensating lenses. The zoom lens group changes the image magnification, while the compensating lens group corrects for image distortion or offset that may occur during zooming. Precisely controlling the positions of these lenses enables high-quality local zoom.

[0032] Step S500: performing a mechanical error identification of the moving mechanism based on the first predetermined zoom parameter and the movement information of the moving mechanism to obtain a first mechanical movement deviation;

[0033] In this embodiment of the present application, the first predetermined zoom parameter is compared with the actual movement information of the moving mechanism. By comparing the expected lens movement distance with the actual movement distance, a preliminary determination is made as to whether there is a mechanical error. Parameters such as velocity and acceleration are also compared to obtain more comprehensive mechanical error information.

[0034] Based on the comparison results, the mechanical errors of the moving mechanism are identified. These errors are caused by factors such as wear, looseness, and deformation of mechanical components. By calculating the difference between the expected parameters and the actual parameters, the first mechanical motion deviation is obtained.

[0035] Step S600: performing feedback optimization on the first predetermined zoom parameter based on the first mechanical motion deviation, and the zoom control module triggers zooming with the first optimized zoom parameter to obtain a first zoom image.

[0036] In an embodiment of the present application, a first mechanical motion deviation is analyzed to determine its magnitude, direction, and possible causes. Based on the analysis results of the first mechanical motion deviation, a first predetermined zoom parameter is adjusted to compensate for the mechanical error by adjusting parameters such as the lens's movement distance and speed, thereby achieving a more accurate zoom operation. The adjusted first predetermined zoom parameter is referred to as a first optimized zoom parameter. A zoom control module performs a zoom operation based on the first optimized zoom parameter, and the operation of the zoom control module yields a first zoomed image based on the first optimized zoom parameter.

[0037] Furthermore, step S300 in the method provided in the application embodiment further includes:

[0038] Obtaining the convolution receptive field of the predetermined abnormal convolution feature and performing a comparison convolution image extraction on the first endoscopic image to obtain a first comparison image block;

[0039] Build a convolutional similarity analysis network;

[0040] The convolution similarity analysis network is used to perform a similarity comparison between the predetermined abnormal convolution feature and the first comparison image block to obtain the first convolution similarity index, and the first image region identifier is generated according to the position of the first comparison image block.

[0041] In the present embodiment, the convolution receptive field refers to the size of the area on the input image corresponding to a pixel point on the output feature map of a layer in a convolutional neural network. When obtaining the convolution receptive field of a predetermined abnormal convolution feature, the corresponding receptive field position on the input image is determined by tracking and calculating the network layer based on the position of the predetermined abnormal convolution feature in the convolutional neural network.

[0042] Based on the size and position of the receptive field, a window of corresponding size is slid across the first endoscopic image. This window traverses every possible position in the image, extracting an image block of the corresponding area each time it moves. For each extracted image block, it is input into the convolutional neural network, and its feature representation is calculated through forward propagation. These feature representations are then compared with the predetermined abnormal convolution features to find image blocks with high similarity. During the comparison process, a list of similarity scores is obtained, where each score corresponds to the degree of similarity between an extracted image block and the predetermined abnormal convolution feature. Based on the similarity score, the image block with the highest score is selected as the first comparison image block.

[0043] The convolutional similarity analysis network is a neural network used to compare the similarity between two input images. When building a convolutional similarity analysis network, a suitable network architecture is designed to effectively extract features from the input images and compare their similarity. The convolutional similarity analysis network consists of an input layer, a convolutional layer, a pooling layer, a fully connected layer, and a similarity calculation layer. The similarity calculation layer compares the feature vectors of the two input images using methods such as cosine similarity and Euclidean distance, and outputs a similarity score.

[0044] After building the network architecture, the network parameters are initialized using methods such as random initialization and pre-trained model initialization. The convolutional similarity analysis network is then trained, using methods such as mean squared error loss and cross-entropy loss to measure the difference between the similarity scores output by the network and the true labels. The convolutional similarity analysis network is trained using a labeled dataset. Using the backpropagation algorithm and gradient descent optimizer, the network parameters are adjusted to minimize the loss function. During training, regularization and learning rate adjustment are used to prevent overfitting and improve training efficiency. After training, the convolutional similarity analysis network is evaluated using a test set. Network performance is assessed by calculating metrics such as precision and recall. Based on the evaluation results, the network architecture and parameter settings are fine-tuned.

[0045] Through the above process, the construction of the convolutional similarity analysis network is completed.

[0046] The predetermined abnormal convolutional features and the first comparison image block extracted from the first endoscopic image are input into a trained convolutional similarity analysis network. The convolutional similarity analysis network extracts features from the input images, compares the similarities of these features, and ultimately outputs a similarity index, namely a first convolutional similarity index. A first image region identifier is generated based on the first convolutional similarity index and the position of the first comparison image block in the first endoscopic image.

[0047] Furthermore, the method further comprises:

[0048] The convolutional similarity analysis network includes a first convolutional neural network, a second convolutional neural network and a feature loss analysis network, and the first convolutional neural network and the second convolutional neural network share weights;

[0049] The feature loss analysis network is used to perform deviation comparison on the image features output by the first convolutional neural network and the second convolutional neural network, and generate a similarity index according to the deviation comparison result;

[0050] The first convolutional neural network, the second convolutional neural network and the feature loss analysis network are obtained by training with identification information of multiple groups of image sample combinations and identification similarities.

[0051] In an embodiment of the present application, the convolutional similarity analysis network includes a first convolutional neural network, a second convolutional neural network, and a feature loss analysis network. The first convolutional neural network and the second convolutional neural network have the same structure and weights. The first convolutional neural network and the second convolutional neural network are used to extract features of the input image. Since they share weights, no matter which image is input, the two networks will extract features in the same way, ensuring consistency of comparison.

[0052] The feature loss analysis network receives the features output by the first and second convolutional neural networks and compares their deviations. The feature loss analysis network includes fully connected layers that integrate the features and convert them into comparable vectors. It calculates the similarity between the two feature vectors using methods such as cosine similarity and Euclidean distance to generate a similarity metric.

[0053] When training a convolutional similarity analysis network, it uses multiple sets of image samples and similarity markers. Each set of image samples contains two images, which can be similar or dissimilar. Each set of images has similarity markers, indicating whether the two images are similar. This marker can be a binary or continuous similarity score. This marker information serves as a supervisory signal, guiding the network to learn how to accurately compare image similarities.

[0054] During training, image samples are fed into the convolutional similarity analysis network. The first and second convolutional neural networks extract features for each image. The feature loss analysis network compares these features and outputs a similarity metric. Methods such as mean squared error loss and cross-entropy loss are used to calculate the difference between the similarity metric output by the network and the true identification information. The network weights are updated using a backpropagation algorithm and a gradient descent optimizer to minimize the loss function. These steps are repeated until the network achieves satisfactory performance on the validation set.

[0055] The first convolutional neural network, the second convolutional neural network and the feature loss analysis network are obtained through training.

[0056] Furthermore, based on the first predetermined zoom parameter and the movement information of the movement mechanism, a mechanical error of the movement mechanism is identified to obtain a first mechanical movement deviation. The method further includes:

[0057] performing optical path deviation identification based on the first predetermined zoom parameter and the movement information of the moving mechanism;

[0058] Performing zoom imaging impact analysis based on the optical path deviation to obtain a first abnormal impact index;

[0059] If the first abnormal impact index is less than or equal to a predetermined impact threshold, the motion deviation of the moving mechanism is inferred from the optical path deviation to obtain the first mechanical motion deviation;

[0060] If the first abnormal impact indicator is greater than a predetermined impact threshold, first moving mechanism wear alarm information is generated.

[0061] In this embodiment of the present application, the first predetermined zoom parameter is compared with the motion information of the moving mechanism, and optical path deviation is identified by calculating the difference between the theoretical optical path and the actual optical path and analyzing abnormal fluctuations in the motion information. Optical path deviation refers to the phenomenon in which the optical path of the optical system changes due to inaccurate or unstable motion of the moving mechanism, thereby affecting image quality.

[0062] Zoom imaging impact analysis is performed based on optical path deviation, which includes imaging quality assessment, distortion analysis, and distortion assessment.

[0063] Image quality assessment compares the image quality degradation when optical path deviation exists with that under ideal conditions by calculating parameters such as resolution, contrast, and signal-to-noise ratio. Distortion analysis identifies image distortions such as barrel distortion and pincushion distortion by analyzing the image's geometry and edge information. Distortion assessment assesses the degree and impact of distortion by analyzing the image's color space and brightness distribution. Finally, quantitative analysis yields the first abnormal impact index, which measures the negative impact of optical path deviation on zoom imaging.

[0064] When the first abnormal impact index is less than or equal to the predetermined impact threshold, the motion deviation of the mobile mechanism is reversed through the optical path deviation. In order to obtain the first mechanical motion deviation, a mathematical model is established to describe the relationship between the optical path deviation and the motion deviation of the mobile mechanism. The model can reflect how different motion states of the mobile mechanism cause changes in the optical path. Collect data related to the optical path deviation, including zoom parameters, motion trajectory of the mobile mechanism, speed curve, etc. Quantitative analysis is performed on the collected optical path deviation data to extract key characteristic parameters, such as the size, direction, and rate of change of the deviation. Back-calculation is performed using the established correlation model and the optical path deviation parameters obtained by quantitative analysis. After the above steps, the first mechanical motion deviation is obtained, that is, the specific motion error of the mobile mechanism that causes the optical path deviation.

[0065] When the first abnormal impact index exceeds a predetermined threshold, that is, due to severe wear or failure of the moving mechanism, the optical path deviation has caused a significant negative impact on zoom imaging, and first moving mechanism wear alarm information is generated.

[0066] Furthermore, after obtaining the first mechanical motion deviation, the method further includes:

[0067] Collecting lens temperature information of a target zoom optical adapter;

[0068] Identifying the impact of lens performance using the lens temperature information to obtain a lens performance impact index;

[0069] Identify a zoom error according to the lens performance influencing index, and if the zoom error exceeds a predetermined error threshold, obtain a predetermined lens temperature;

[0070] Starting a cooling device to control the temperature of the lens of the target zoom optical adapter with the predetermined lens temperature as the target;

[0071] The first zoom image is acquired by using the target zoom optical adapter after temperature reduction control.

[0072] In the embodiment of the present application, a temperature sensor is first used to collect the lens temperature information of the target zoom optical adapter.

[0073] Based on the collected lens temperature information, combined with the lens performance parameters and their relationship with temperature, the specific impact of temperature on lens performance is analyzed. After analyzing the impact of temperature on lens performance, the lens performance impact index is calculated based on the focal length change and aberration change caused by temperature changes.

[0074] Due to variations in lens performance, a zoom optical adapter may not achieve the expected zoom effect, resulting in zoom error. The magnitude of the zoom error is determined by comparing lens performance influencing indicators with preset zoom performance standards. If the zoom error exceeds a predetermined error threshold, a predetermined lens temperature is acquired. The predetermined lens temperature is set based on optimal lens performance and is within the temperature range where lens performance is most stable and zoom error is minimized.

[0075] After obtaining the desired lens temperature, the cooling device is activated to control the lens temperature. The cooling device is a cooling plate or other device that reduces the lens temperature through refrigeration or other means. During the cooling process, the lens temperature is monitored in real time, and the operating state of the cooling device is adjusted according to temperature changes to ensure that the lens temperature gradually approaches and stabilizes at the desired lens temperature.

[0076] When the lens temperature reaches a predetermined lens temperature and remains stable, a zoom operation is performed using the target zoom optical adapter to acquire a first zoom image.

[0077] Furthermore, the lens performance impact is identified using the lens temperature information to obtain a lens performance impact index, and the method further includes:

[0078] Obtaining lens performance influencing factors, wherein the lens performance influencing factors include at least a refractive index and an expansion coefficient;

[0079] The lens performance impact factor is used to identify the impact of lens zoom, and the lens performance impact index is generated.

[0080] In the embodiments of this application, lens performance influencing factors refer to physical parameters that can affect lens performance. During zooming, these factors change with temperature, affecting the lens's imaging quality. These factors include refractive index and thermal expansion coefficient.

[0081] The refractive index describes the change in the speed of light as it propagates through a lens. It depends on the properties of the lens material and changes with temperature. You can obtain the refractive index data of the lens at different temperatures by consulting the technical manual of the lens material.

[0082] The coefficient of expansion describes the rate of change in volume or size of a lens material when the temperature changes, affecting the shape and focal length of the lens. The coefficient of expansion of a lens material can be obtained by looking in the material data sheet.

[0083] Based on the obtained lens performance influencing factors, a mathematical model describing the relationship between these factors and zoom performance was constructed. Using methods such as linear regression, changes in these factors were linked to zoom performance parameters such as focal length changes. Using this established performance influencing model, the effects of lens zoom were identified, and changes in lens zoom performance at different temperatures were analyzed, along with the correlation between these changes and the performance influencing factors.

[0084] Finally, based on the results of zoom impact identification, the lens performance impact index is generated.

[0085] Furthermore, after step S600 in the method provided in the embodiment of the application, the method further includes:

[0086] Acquire a first original image block based on the first image region identifier;

[0087] performing image color difference recognition on the first zoomed image using the first original image block;

[0088] Performing chromatic aberration correction on the first zoomed image based on the image chromatic aberration recognition result.

[0089] In the embodiment of the present application, the first image region identifier is parsed to determine specific information about the image region, such as its location, size, and shape. Based on the parsed first image region identifier, the pixel matrix of the original image is traversed to find pixels within the coordinate range specified by the identifier, and the corresponding region is located in the original image. Once the target region is located in the original image, all pixels in the region are extracted to form a separate image block, namely the first original image block.

[0090] Color features, including color histograms, color moments, and color sets, are extracted from the first original image block. Simultaneously, the same feature extraction is performed on the corresponding area in the first zoomed image. Similarity or difference metrics between the features are calculated using methods such as Euclidean distance, Manhattan distance, or cosine similarity. Based on the feature matching results, a color difference metric is calculated using methods such as mean squared error and mean absolute error to assess the degree of color difference between the first zoomed image and the first original image block.

[0091] Based on the results of color difference identification, a color difference correction strategy is determined for the first zoomed image. This correction strategy includes adjusting color balance, correcting color shift, and applying color mapping. Based on the determined correction strategy, the image's color space is adjusted using a color space conversion algorithm, and the image's brightness distribution is adjusted using a histogram equalization algorithm to correct the color difference of the first zoomed image. After correction is complete, the image is compared again with the first original image block to ensure that the color difference has been effectively reduced or eliminated. Finally, the corrected first zoomed image is output.

[0092] In the embodiments of the present application, in summary, the embodiments of the present application have at least the following technical effects:

[0093] The present application obtains a target zoom optical adapter, wherein the target zoom optical adapter includes multiple lenses with zoom functions, and the target zoom optical adapter is used to connect a target endoscope to an external imaging device; obtains a first endoscopic image output by the external imaging device; traverses the first endoscopic image with a predetermined abnormal convolution feature to perform a convolution similarity comparison to obtain a first convolution similarity index, and the first convolution similarity index has a first image area identifier; if the first convolution similarity index is greater than a first predetermined similarity, triggers a zoom control module of the target zoom optical adapter, performs local zooming on the first endoscopic image according to a first predetermined zoom parameter based on the first image area identifier, and monitors and obtains motion information of a moving mechanism of the target zoom optical adapter; performs mechanical error identification of the moving mechanism based on the first predetermined zoom parameter and the motion information of the moving mechanism to obtain a first mechanical motion deviation; feedback optimizes the first predetermined zoom parameter based on the first mechanical motion deviation, and the zoom control module triggers zooming with the first optimized zoom parameter to obtain a first zoom image. The present invention solves the technical problems of limited zoom function and inaccurate image zoom control in optical adapters in the prior art. By combining technologies such as zoom control, image recognition and mechanical error recognition, the present invention achieves the technical effect of improving the image quality and stability of endoscopes.

[0094] Example 2

[0095] Based on the same inventive concept as the zoom trigger recognition method based on the optical adapter in the aforementioned embodiment, Figure 2As shown, the present application provides a zoom trigger recognition system based on an optical adapter. The system and method embodiments in the present application are based on the same inventive concept. The system includes:

[0096] a target zoom optical adapter acquiring unit 11, wherein the target zoom optical adapter acquiring unit 11 acquires a target zoom optical adapter, wherein the target zoom optical adapter comprises a plurality of lenses with a zoom function, and the target zoom optical adapter is used to connect a target endoscope with an external imaging device;

[0097] a first endoscopic image output unit 12, which acquires a first endoscopic image output by the external imaging device;

[0098] a first convolution similarity index obtaining unit 13, wherein the first convolution similarity index obtaining unit 13 traverses the first endoscopic image to perform a convolution similarity comparison using a predetermined abnormal convolution feature to obtain a first convolution similarity index, wherein the first convolution similarity index has a first image region identifier;

[0099] a motion information acquisition unit 14 configured to trigger a zoom control module of the target zoom optical adapter if the first convolution similarity index is greater than a first predetermined similarity, perform a local zoom on the first endoscopic image according to a first predetermined zoom parameter based on the first image region identifier, and monitor and acquire motion information of a moving mechanism of the target zoom optical adapter;

[0100] a first mechanical motion deviation acquiring unit 15 , which performs a mechanical error identification of the moving mechanism based on the first predetermined zoom parameter and the movement information of the moving mechanism to obtain a first mechanical motion deviation;

[0101] The first zoom image acquisition unit 16 performs feedback optimization on the first predetermined zoom parameter based on the first mechanical motion deviation, and the zoom control module triggers zooming with the first optimized zoom parameter to obtain a first zoom image.

[0102] Furthermore, the system further comprises:

[0103] Obtaining the convolution receptive field of the predetermined abnormal convolution feature and performing a comparison convolution image extraction on the first endoscopic image to obtain a first comparison image block;

[0104] Build a convolutional similarity analysis network;

[0105] The convolution similarity analysis network is used to perform a similarity comparison between the predetermined abnormal convolution feature and the first comparison image block to obtain the first convolution similarity index, and the first image region identifier is generated according to the position of the first comparison image block.

[0106] Furthermore, the system further comprises:

[0107] The feature loss analysis network is used to perform deviation comparison on the image features output by the first convolutional neural network and the second convolutional neural network, and generate a similarity index according to the deviation comparison result;

[0108] The first convolutional neural network, the second convolutional neural network and the feature loss analysis network are obtained by training with identification information of multiple groups of image sample combinations and identification similarities.

[0109] Furthermore, the system further comprises:

[0110] performing optical path deviation identification based on the first predetermined zoom parameter and the movement information of the moving mechanism;

[0111] Performing zoom imaging impact analysis based on the optical path deviation to obtain a first abnormal impact index;

[0112] If the first abnormal impact index is less than or equal to a predetermined impact threshold, the motion deviation of the moving mechanism is inferred from the optical path deviation to obtain the first mechanical motion deviation;

[0113] If the first abnormal impact indicator is greater than a predetermined impact threshold, a first moving mechanism wear alarm message is generated.

[0114] Furthermore, the system further comprises:

[0115] Collecting lens temperature information of a target zoom optical adapter;

[0116] Identifying the impact of lens performance using the lens temperature information to obtain a lens performance impact index;

[0117] Identify a zoom error according to the lens performance influencing index, and if the zoom error exceeds a predetermined error threshold, obtain a predetermined lens temperature;

[0118] Starting a cooling device to control the temperature of the lens of the target zoom optical adapter with the predetermined lens temperature as the target;

[0119] The first zoom image is acquired by using the target zoom optical adapter after temperature reduction control.

[0120] Furthermore, the system further comprises:

[0121] Obtaining lens performance influencing factors, wherein the lens performance influencing factors include at least a refractive index and an expansion coefficient;

[0122] The lens performance impact factor is used to identify the impact of lens zoom, and the lens performance impact index is generated.

[0123] Furthermore, the system further comprises:

[0124] Acquire a first original image block based on the first image region identifier;

[0125] performing image color difference recognition on the first zoomed image using the first original image block;

[0126] Performing chromatic aberration correction on the first zoomed image based on the image chromatic aberration recognition result.

[0127] It should be noted that the above-mentioned order of the embodiments of the present application is for descriptive purposes only and does not represent the superiority or inferiority of the embodiments. The above description is of specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps described in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0128] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application.

[0129] This specification and drawings are merely illustrative of the present application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Obviously, those skilled in the art may make various modifications and variations to this application without departing from the scope of this application. Thus, this application is intended to include such modifications and variations as fall within the scope of this application and its equivalents.

Claims

1. A zoom trigger recognition method based on an optical adapter, characterized in that: include: Obtaining a target zoom optical adapter, wherein the target zoom optical adapter comprises a plurality of lenses with a zoom function, and the target zoom optical adapter is used to connect a target endoscope with an external imaging device; acquiring a first endoscopic image output by the external imaging device; Traversing the first endoscopic image with a predetermined abnormal convolution feature to perform a convolution similarity comparison to obtain a first convolution similarity index, wherein the first convolution similarity index has a first image region identifier; If the first convolution similarity index is greater than a first predetermined similarity, triggering a zoom control module of the target zoom optical adapter to locally zoom the first endoscopic image according to a first predetermined zoom parameter based on the first image region identifier, and monitoring and acquiring motion information of a moving mechanism of the target zoom optical adapter; performing a mechanical error identification of the moving mechanism based on the first predetermined zoom parameter and the movement information of the moving mechanism to obtain a first mechanical movement deviation; performing feedback optimization on the first predetermined zoom parameter based on the first mechanical motion deviation, and triggering zooming with the first optimized zoom parameter by a zoom control module to obtain a first zoom image; Performing a moving mechanism mechanical error identification based on the first predetermined zoom parameter and the moving mechanism motion information to obtain a first mechanical motion deviation includes: performing optical path deviation identification based on the first predetermined zoom parameter and the movement information of the moving mechanism; Performing zoom imaging impact analysis based on the optical path deviation to obtain a first abnormal impact index; If the first abnormal impact index is less than or equal to a predetermined impact threshold, the motion deviation of the moving mechanism is inferred from the optical path deviation to obtain the first mechanical motion deviation; If the first abnormal impact indicator is greater than a predetermined impact threshold, generating a first moving mechanism wear alarm message; After obtaining the first mechanical motion deviation, the method further includes: Collecting lens temperature information of a target zoom optical adapter; Identifying the impact of lens performance using the lens temperature information to obtain a lens performance impact index; Identify a zoom error based on the lens performance influencing index, and if the zoom error exceeds a predetermined error threshold, obtain a predetermined lens temperature; Starting a cooling device to control the temperature of the lens of the target zoom optical adapter with the predetermined lens temperature as a target; The first zoom image is acquired by using the target zoom optical adapter after temperature reduction control.

2. The method according to claim 1, wherein Traversing the first endoscopic image with a predetermined abnormal convolution feature to perform convolution similarity comparison to obtain a first convolution similarity index, including: Obtaining the convolution receptive field of the predetermined abnormal convolution feature and performing a comparison convolution image extraction on the first endoscopic image to obtain a first comparison image block; Build a convolutional similarity analysis network; The convolution similarity analysis network is used to perform a similarity comparison between the predetermined abnormal convolution feature and the first comparison image block to obtain the first convolution similarity index, and the first image region identifier is generated according to the position of the first comparison image block.

3. The method according to claim 2, wherein The convolutional similarity analysis network includes a first convolutional neural network, a second convolutional neural network and a feature loss analysis network, and the first convolutional neural network and the second convolutional neural network share weights; The feature loss analysis network is used to perform deviation comparison on the image features output by the first convolutional neural network and the second convolutional neural network, and generate a similarity index according to the deviation comparison result; The first convolutional neural network, the second convolutional neural network and the feature loss analysis network are obtained by training with identification information of multiple groups of image sample combinations and identification similarities.

4. The method according to claim 1, wherein The lens temperature information is used to identify the impact of lens performance, and a lens performance impact index is obtained, including: Obtaining lens performance influencing factors, wherein the lens performance influencing factors include at least a refractive index and an expansion coefficient; The lens performance impact factor is used to identify the impact of lens zoom, and the lens performance impact index is generated.

5. The method according to claim 1, wherein After obtaining the first zoom image, the method further includes: Acquire a first original image block based on the first image region identifier; performing image color difference recognition on the first zoomed image using the first original image block; Performing chromatic aberration correction on the first zoomed image based on the image chromatic aberration recognition result.

6. A zoom trigger recognition system based on an optical adapter, characterized in that: The system is used to perform the method according to any one of claims 1 to 5, and the system includes: a target zoom optical adapter acquiring unit, wherein the target zoom optical adapter acquires a target zoom optical adapter, wherein the target zoom optical adapter comprises a plurality of lenses with a zoom function, and the target zoom optical adapter is used to connect the target endoscope with an external imaging device; a first endoscopic image output unit, configured to acquire a first endoscopic image output by the external imaging device; a first convolution similarity index obtaining unit, wherein the first convolution similarity index obtaining unit traverses the first endoscopic image to perform a convolution similarity comparison using a predetermined abnormal convolution feature to obtain a first convolution similarity index, wherein the first convolution similarity index has a first image region identifier; a motion information acquisition unit, configured to trigger a zoom control module of the target zoom optical adapter if the first convolution similarity index is greater than a first predetermined similarity, perform a local zoom on the first endoscopic image according to a first predetermined zoom parameter based on the first image region identifier, and monitor and acquire motion information of a moving mechanism of the target zoom optical adapter; a first mechanical motion deviation acquiring unit, configured to identify a mechanical error of the moving mechanism based on the first predetermined zoom parameter and the motion information of the moving mechanism to obtain a first mechanical motion deviation; The first zoom image acquisition unit performs feedback optimization on the first predetermined zoom parameter based on the first mechanical motion deviation, and the zoom control module triggers zooming with the first optimized zoom parameter to obtain a first zoom image.

Images