An interactive control method and system based on robot control

Abstract

The application provides an interactive control method and system based on robot control, and relates to the technical field of robot control.The method comprises the following steps: performing differential processing on the visible light image and the infrared image in the early education environment, and extracting the suspicious target area in the differential image; designing a feature descriptor for the suspicious target area, and performing feature analysis on the suspicious target area according to the feature descriptor to obtain a feature analysis result; performing preliminary discrimination on the suspicious target area according to the feature analysis result, and processing the target set after discrimination through a clustering algorithm to obtain a target list.The application improves the accuracy of target recognition in the early education environment, and enables the early education robot to perform personalized interactive response according to the recognition result, thereby enhancing the educational interactivity and interest.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of robot control, in particular to an interactive control method and system based on robot control. BACKGROUND

[0002] Traditional early education robot interaction methods are often single and lack sufficient intelligence. For example, some early education robots only support simple touch interaction, and users need to interact with the robot through the touch screen. Although this interaction method is intuitive and easy to use, it limits the user's operation freedom and flexibility and cannot fully utilize the potential of artificial intelligence technology.

[0003] In addition, although some early education robots are equipped with voice recognition functions, they still have many problems in actual application. These robots often cannot accurately recognize user voice commands, especially in noisy environments or when the user's pronunciation is unclear. At the same time, they also lack the ability to intelligently adjust according to the user's specific situation. For example, when the distance between the user and the robot changes, the robot cannot automatically adjust the volume or recognition sensitivity; in environments with different light brightness, the robot's screen brightness and display content cannot be adjusted accordingly, affecting the user's experience. SUMMARY

[0004] The technical problem to be solved by the present application is to provide an interactive control method and system based on robot control, which accurately extracts suspicious target areas and improves the accuracy and efficiency of target detection.

[0005] To solve the above technical problems, the technical solution of the present application is as follows:

[0006] In a first aspect, an interactive control method based on robot control is provided, which comprises:

[0007] Difference processing is performed on the visible light image and the infrared image in the early education environment to extract suspicious target areas in the difference image;

[0008] For the suspicious target area, a feature descriptor is designed, and feature analysis is performed on the suspicious target area according to the feature descriptor to obtain a feature analysis result;

[0009] According to the feature analysis result, the suspicious target area is preliminarily identified, and the target set after identification is processed by a clustering algorithm to obtain a target list;

[0010] According to the target list, a convolutional neural network model and a random forest model are constructed, and the targets in the target list are feature extracted and recognized using the visible light image and the infrared image respectively to obtain a preliminary recognition result;

[0011] The preliminary recognition result is compared with target information in a database preset by the early education robot, and a flower pollination algorithm is used to optimize and adjust parameters of a convolutional neural network model and a random forest model to obtain a final recognition result and target position information.

[0012] According to the final recognition result and the target position information, the early education robot performs a corresponding interactive response, including playing educational content related to the identified target, having a conversation with the child, and displaying an expression animation related to the identified target.

[0013] Further, the visible light image and the infrared image in the early education environment are differentially processed to extract a suspicious target area in the differential image, including:

[0014] The visible light image and the infrared image in the early education environment are captured by a camera device built into the early education robot;

[0015] The visible light image and the infrared image are differentially processed by calculating the difference in pixel values between the visible light image and the infrared image to obtain a differentially processed image;

[0016] The differentially processed image is analyzed, including threshold segmentation, edge detection, and morphological processing, to identify a suspicious target area in the early education environment, including a dropped toy and an unreturned teaching aid.

[0017] Further, a feature descriptor is designed for the suspicious target area, and feature analysis is performed on the suspicious target area according to the feature descriptor to obtain a feature analysis result, including:

[0018] For the circularity feature, according to a preset threshold, it is determined whether the shape of the suspicious target area conforms to the typical shape of objects in the early education environment, including circular toys and spherical teaching aids; for the highlight area area feature, the size of the highlight area and the proportion of the highlight area in the suspicious target area are analyzed to identify objects with reflective properties, including metal toys and mirror teaching aids; for the infrared light before and after the gray scale change feature, the amplitude and trend of the gray scale change are analyzed to determine whether the suspicious target area includes an object with specific thermal properties, including an electronic device that is heating up and a heat-sensitive material; for the local contrast feature, according to the strength of the contrast, the degree of the suspicious target area in the image and the degree of differentiation from the surrounding environment are determined;

[0019] According to the analysis of circularity, highlight area area, infrared light before and after the gray scale change, and local contrast, the feature analysis result is obtained.

[0020] Further, according to the feature analysis result, the suspicious target area is preliminarily identified, and the identified target set is processed by a clustering algorithm to obtain a target list, including:

[0021] Set a discrimination threshold, when the feature analysis result of the suspicious target region ≥ the discrimination threshold, the corresponding feature is determined as the camera target, and the corresponding feature is added to the target set after discrimination;

[0022] According to the feature descriptor of the target in the target set after discrimination, including position, shape, size, set clustering parameters, including clustering number, distance threshold;

[0023] The target set after discrimination is divided into different clustering clusters, and the cluster center of each clustering cluster is extracted;

[0024] Arrange the cluster centers of all clustering clusters in order to form a target list.

[0025] Further, according to the target list, construct a convolutional neural network model and a random forest model, and respectively use the visible light image and the infrared image to extract and identify the target in the target list to obtain the preliminary identification result, including:

[0026] According to the target list, a convolutional neural network model is constructed, including convolutional layer, pooling layer and full connection layer; the visible light image is input into the convolutional neural network model, the high-level semantic features in the image are extracted through convolutional layer and pooling layer operation, including shape, color pattern and texture, and the high-level semantic features are input into the full connection layer of the convolutional neural network model for classification, and the visible light image preliminary identification result is output, including target type, position and confidence information;

[0027] According to the target list, a random forest model is constructed; the infrared image corresponding to the target is determined as the training data from the target list, and the image is extracted, including calculating temperature distribution and infrared reflectivity bottom texture feature; the infrared image feature is input into the random forest model, each decision tree in the random forest model makes independent judgment according to the feature, and outputs the infrared image preliminary identification result, including target type, position and confidence information.

[0028] Integrate the visible light image preliminary identification result and the infrared image preliminary identification result to obtain the integrated preliminary identification result.

[0029] Further, compare the preliminary identification result with the target information in the preset database of the early education robot, and optimize and adjust the parameters of the convolutional neural network model and the random forest model by using the flower pollination algorithm to obtain the final identification result and the target position information, including:

[0030] Compare the preliminary identification result with the target information in the preset database of the early education robot to determine the difference between the preliminary identification result and the target;

[0031] According to the difference between the preliminary identification result and the target, the related parameters of the flower pollination algorithm are initialized, including the number of flowers, the pollination probability, the local search radius, and the parameters of the convolutional neural network model and the random forest model are optimized, wherein each flower performs local search and global pollination according to the evaluation function to obtain the optimized parameters output by the flower pollination algorithm;

[0032] According to the optimized parameters output by the flower pollination algorithm, the targets in the target list are re-identified to obtain new identification results;

[0033] The new identification results are compared again with the target information in the database preset by the early education robot, and the new errors are calculated;

[0034] The processes of evaluation function design, parameter optimization, re-identification and error calculation are repeated until the preset iteration number is reached, and the identification result obtained in the last iteration is taken as the final identification result and target position information.

[0035] Further, the calculation process of the evaluation function is as follows:

[0036] All target types are traversed, and for each target type, the predicted type is compared. If the target type is the same as the predicted type, the indication value of the corresponding case is set to 1, otherwise it is set to 0;

[0037] All cases with an indication value of 1 are accumulated to obtain the number of correctly identified target types, and the number of target types and the preset weight coefficient are combined to obtain the weighted value of the type recognition accuracy;

[0038] All target positions are traversed, and for each target position, the predicted position is compared. If the target position is the same as the predicted position, the indication value of the corresponding case is set to 1, otherwise it is set to 0;

[0039] All cases with an indication value of 1 are accumulated to obtain the number of correctly identified target positions, and the number of target positions and the preset weight coefficient are combined to obtain the weighted value of the position recognition accuracy;

[0040] All identification results are traversed, and for each identification result, the absolute value of the difference between the confidence and the matching probability is calculated;

[0041] According to the absolute value and the weight of the corresponding identification result, the sum of the weighted confidence error is obtained;

[0042] According to the sum of the weighted confidence error and the normalization factor, the normalized confidence error is obtained, and the normalized confidence error and the preset weight coefficient are combined to obtain the weighted value of the confidence error;

[0043] According to the type recognition accuracy weighted value and the position recognition accuracy weighted value, a preliminary evaluation value is obtained;

[0044] On the basis of the preliminary evaluation value, the weighted value of the confidence error is adjusted to obtain an unnormalized evaluation result;

[0045] The unnormalized evaluation result is proportionally adjusted with the total number of recognition targets for normalization processing, and finally the value of the evaluation function is obtained.

[0046] In a second aspect, an interactive control system based on robot control includes:

[0047] A difference processing module is configured to perform difference processing on the visible light image and the infrared image in the early education environment to extract a suspicious target region in a difference image;

[0048] A feature analysis module is configured to design a feature descriptor for the suspicious target region and perform feature analysis on the suspicious target region according to the feature descriptor to obtain a feature analysis result;

[0049] A discrimination and clustering module is configured to preliminarily discriminate the suspicious target region according to the feature analysis result to obtain a target set after discrimination, and process the target set after discrimination through a clustering algorithm to obtain a target list;

[0050] An identification module is configured to construct a convolutional neural network model and a random forest model according to the target list, and perform feature extraction and identification on the targets in the target list using the visible light image and the infrared image to obtain a preliminary identification result;

[0051] A comparison and optimization module is configured to compare the preliminary identification result with target information in a database preset by the early education robot, and optimize and adjust parameters of the convolutional neural network model and the random forest model using an optimization algorithm to obtain a final identification result and target position information;

[0052] An interactive response module is configured to cause the early education robot to perform a corresponding interactive response according to the final identification result and the target position information, including playing educational content related to the identified target, having a conversation with the child, and displaying an expression animation related to the identified target.

[0053] In a third aspect, a computing device includes:

[0054] One or more processors;

[0055] A storage device is configured to store one or more programs, when the one or more programs are executed by the one or more processors, so that the one or more processors implement the method.

[0056] In a fourth aspect, a computer readable storage medium having stored therein a program, which, when executed by a processor, implements the method.

[0057] The above scheme of the present application at least has the following beneficial effects:

[0058] By differentiating the visible light image and the infrared image in the early education environment, the difference between the two can be effectively highlighted, so that the suspicious target area can be accurately extracted, the accuracy and efficiency of target detection are improved, and the possibility of false detection and missed detection is reduced. A specific feature descriptor is designed for the suspicious target area, and feature analysis is performed to obtain detailed feature analysis results, which provides rich feature information for preliminary discrimination and clustering processing, and improves the accuracy of recognition. According to the feature analysis result, the suspicious target area is preliminarily discriminated, and the target set after discrimination is processed through a clustering algorithm to obtain a clear target list, which is helpful for classifying and organizing the targets. A convolutional neural network model and a random forest model are constructed, and the targets in the target list are feature extracted and recognized by using the visible light image and the infrared image respectively, which combines the advantages of the two different models and improves the robustness and accuracy of recognition.

[0059] The preliminary recognition result is compared with the target information in the database preset by the early education robot, and the parameters of the convolutional neural network model and the random forest model are optimized and adjusted by using the flower pollination algorithm. As an intelligent optimization algorithm, the flower pollination algorithm can effectively find the optimal solution of the model parameters, further improving the accuracy of recognition. According to the final recognition result and the target position information, the early education robot performs corresponding interactive response, including playing education content related to the recognized target, having a conversation with the child, and displaying expression animation related to the recognized target, which enhances the interactivity and interest of the early education robot, improves the learning interest and participation of the child, and also shows the intelligent and personalized service characteristics of the early education robot. BRIEF DESCRIPTION OF DRAWINGS

[0060] Figure 1 is a flowchart of an interactive control method based on robot control provided by an embodiment of the present application.

[0061] Figure 2 is a schematic diagram of an interactive control system based on robot control provided by an embodiment of the present application. DETAILED DESCRIPTION

[0062] Exemplary embodiments of the present disclosure will be described in greater detail below with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it is understood that the present disclosure can be embodied in various forms without being limited by the embodiments set forth herein. Rather, these embodiments are provided so that the present disclosure will be thoroughly and completely understood, and will fully convey the scope of the present disclosure to those skilled in the art.

[0063] As shown in Figure 1 An embodiment of the present disclosure proposes an interactive control method based on robot control, the method comprising the following steps:

[0064] Step 1, differentiating the visible light image and the infrared image in the early education environment to extract the suspicious target area in the difference image;

[0065] Step 2, designing a feature descriptor for the suspicious target area, and performing feature analysis on the suspicious target area according to the feature descriptor to obtain a feature analysis result;

[0066] Step 3, performing preliminary discrimination on the suspicious target area according to the feature analysis result, and processing the target set after discrimination through a clustering algorithm to obtain a target list;

[0067] Step 4, constructing a convolutional neural network model and a random forest model according to the target list, and performing feature extraction and recognition on the targets in the target list using the visible light image and the infrared image respectively to obtain a preliminary recognition result;

[0068] Step 5, comparing the preliminary recognition result with the target information in the preset database of the early education robot, and optimizing and adjusting the parameters of the convolutional neural network model and the random forest model using a flower pollination algorithm to obtain a final recognition result and target position information;

[0069] Step 6, according to the final recognition result and the target position information, the early education robot performs corresponding interactive responses, including playing education content related to the recognized target, having a conversation with the child, and displaying an expression animation related to the recognized target.

[0070] In the embodiments of the present application, the suspicious target area in the early education environment can be more accurately extracted through differential processing and feature analysis technology, and feature analysis is performed based on rich feature descriptors, thereby improving the accuracy of target recognition. Combined with the interactive response capability of the early education robot, corresponding interactive operations can be performed according to the recognition result, such as playing educational content related to the identified target, having a simple conversation with the child, etc., greatly enhancing the interactivity and educational effect of the early education robot. The intelligent level of the early education robot can be continuously improved by using machine learning models and optimization algorithms to continuously optimize and adjust the recognition model, so that it can better adapt to different scenes and user needs. This method is not only suitable for early education environments, but can also be extended to other scenes that require intelligent recognition and interaction, providing a wider application prospect for the development of intelligent robot technology.

[0071] In a preferred embodiment of the present application, step 1, the visible light image and the infrared image in the early education environment are subjected to differential processing to extract the suspicious target area in the differential image, which can include:

[0072] Step 112, using the built-in camera device of the early education robot, visible light images and infrared images in the early education environment are collected;

[0073] Step 113, the visible light image and the infrared image are subjected to differential processing by calculating the difference between the pixel values of the visible light image and the infrared image to obtain the differential processed image;

[0074] Step 114, the differential processed image is analyzed, including threshold segmentation, edge detection, and morphological processing, to identify suspicious target areas in the early education environment, including dropped toys and unreturned teaching aids.

[0075] In the embodiments of the present application, the early education robot is built-in with a high-definition visible light camera and an infrared camera to ensure that both types of images can be collected simultaneously. The visible light camera collects color images of the early education environment at a standard resolution (such as 1080p), recording visual information such as lighting and color of the environment. The infrared camera uses an infrared light source and an infrared sensor to collect infrared radiation images of the environment, mainly for identifying the thermal characteristics of objects such as body temperature and heat sources. The central processing unit (CPU) or the image processing unit (GPU) of the early education robot controls the synchronous collection of the two cameras to ensure the consistency of the two types of images in time, facilitating subsequent differential processing.

[0076] Step 113, the collected visible light image and infrared image are grayed, and the color image is converted into a gray image. The gray image is normalized to unify the pixel value range to [0, 1], reducing the influence of light changes on the difference result. The pre-processed visible light image and infrared image are calculated pixel by pixel, that is, for each pixel point (x, y), the difference between the gray values is calculated. The difference result is mapped to a new gray image to form the difference-processed image. The highlight area in the difference image represents the area with a large difference between the two images, which is the suspicious target area.

[0077] Step 114, a suitable threshold T is set to divide the pixel values in the difference image into two categories: pixel points greater than T are considered to belong to the suspicious target area, and pixel points ≤ T are considered to belong to the background area. Through binaryzation processing, the difference image is converted into a binary image. An edge detection algorithm (such as Canny edge detection) is applied to process the binary image to identify the edge profile of the suspicious target area, which provides important geometric feature information for subsequent morphological processing. Morphological opening operation (erosion first, then dilation) and closing operation (dilation first, then erosion) are applied to process the image after edge detection to remove noise and small holes, fill cracks and fine cracks. Through the image after morphological processing, the suspicious target area in the early education environment, such as the fallen toy, the unreturned teaching aid, etc., can be more clearly identified.

[0078] Suppose the early education robot is working in a child's room, and there is a red toy ball fallen on the ground and an unreturned building block teaching aid in the room. The visible light camera and infrared camera built-in the early education robot simultaneously collect images in the room. The visible light camera captures the color scene in the room, including the red toy ball and the building block teaching aid; the infrared camera captures the thermal radiation distribution in the room, and since the temperature of the toy ball and the building block teaching aid differs from the surrounding environment, they will also be reflected in the infrared image. The collected visible light image and infrared image are difference-processed. Since the positions and gray values of the toy ball and the building block teaching aid differ in the two images, they will be highlighted in the difference image. The threshold segmentation, edge detection and morphological processing are performed on the difference-processed image. By setting a suitable threshold, the difference image is converted into a binary image; then an edge detection algorithm is applied to identify the edge profile of the suspicious target area; finally, through morphological processing, noise is removed and cracks are filled to obtain a clear suspicious target area image. According to the recognition result, the early education robot can judge that there is a fallen toy ball and an unreturned building block teaching aid in the room, and remind the child or the parent to tidy up.

[0079] By differential processing and image analysis technology, suspicious target areas in early education environment such as dropped toys, unreturned teaching aids, etc. can be more accurately identified, reducing the occurrence of false positives and false negatives. Early education robots can timely discover and remind children or parents to handle potential safety hazards such as tripping or accidental ingestion caused by small objects dropped on the ground, improving the safety of early education environment. Early education robots can provide personalized interactive services according to the identification results, such as playing education content related to the identified target, guiding children to return the objects, etc., enhancing the interactive experience and education effect of users. The method provides a new technical path and solution for the intelligent development of early education robots, and helps to promote the continuous progress and application expansion of early education robot technology.

[0080] In a preferred embodiment of the present application, step 2, for the suspicious target area, a feature descriptor is designed, and feature analysis is performed on the suspicious target area according to the feature descriptor to obtain a feature analysis result, including:

[0081] Step 221, for the circularity feature, according to a preset threshold, it is judged whether the shape of the suspicious target area conforms to the typical shape of objects in the early education environment, including circular toys and spherical teaching aids; for the high-light area area feature, the size of the high-light area and the proportion in the suspicious target area are analyzed to identify objects with light reflection characteristics, including metal toys and mirror teaching aids; for the infrared light before and after the gray scale change feature, the amplitude and trend of the gray scale change are analyzed to determine whether the suspicious target area includes objects with specific thermal characteristics, including electronic devices that are heating and heat-sensitive materials; for the local contrast feature, according to the strength of the contrast, it is judged whether the degree of the suspicious target area in the image is distinguishable from the surrounding environment;

[0082] Step 222, according to the analysis of circularity, high-light area area, infrared light before and after the gray scale change, and local contrast, a feature analysis result is obtained.

[0083] In the embodiment of the present application, edge detection and contour extraction are performed on the suspicious target area image after differential processing. For each contour, its circularity (such as using shape factor, circularity index, etc.) is calculated. The circularity is an index that measures the degree to which the contour approximates a circle. The calculated circularity is compared with a preset threshold. If the circularity is higher than the threshold, it is considered that the shape of the suspicious target area conforms to the typical shape of objects in the early education environment (such as circular toys and spherical teaching aids), and the judgment result is recorded as part of the feature analysis result. Brightness analysis is performed on the suspicious target area image to identify high-light areas. The size of the high-light area and its proportion in the suspicious target area are calculated. According to the size and proportion of the high-light area, it is judged whether the suspicious target area has light reflection characteristics (such as metal toys and mirror teaching aids), and the judgment result is recorded as part of the feature analysis result.

[0084] The gray value of the visible light image before difference processing and the infrared image is compared, and the gray scale change amplitude and trend of the suspicious target area before and after infrared light are calculated. According to the amplitude and trend of the gray scale change, it is judged whether the suspicious target area includes objects with specific thermal characteristics (such as electronic devices that are heating up, heat-sensitive materials). The judgment result is recorded as part of the feature analysis result. Contrast analysis is performed on the suspicious target area and its surrounding environment, and the strength of the local contrast is calculated. According to the strength of the contrast, the degree of differentiation of the suspicious target area in the image from the surrounding environment is judged, and the judgment result is recorded as part of the feature analysis result.

[0085] Step 222, the analysis results of circularity, high light area, gray scale change before and after infrared light, and local contrast are integrated together to form a complete feature analysis result.

[0086] Suppose the early education robot is working in a child's room and detects a suspicious target area. The robot identifies the contour of the suspicious target area through edge detection and contour extraction, calculates the circularity of the contour, and finds that the value is higher than the preset threshold, judging that the shape of the suspicious target area conforms to the typical shape of circular toys in the early education environment. The robot performs brightness analysis on the suspicious target area, identifies the high light area, calculates the area and proportion of the high light area, and finds that the proportion is large, judging that the suspicious target area may have a reflective property, such as a metal toy. The robot compares the images before and after difference processing, calculates the gray scale change amplitude and trend of the suspicious target area, finds that the gray scale change amplitude is small but the trend is steadily rising, and judges that the suspicious target area may include objects with specific thermal characteristics, such as electronic devices that are heating up slowly. The robot performs contrast analysis on the suspicious target area and its surrounding environment, finds that the contrast of the suspicious target area and the surrounding environment is strong, judges that the suspicious target area is prominent in the image and easy to identify, and the robot integrates the above analysis results together to form a complete feature analysis result.

[0087] By designing multiple feature descriptors and performing comprehensive analysis, the suspicious target area in the early education environment can be more accurately identified, reducing the occurrence of misjudgment and missed judgment. This method can identify suspicious target areas with different characteristics, such as shape, reflective property, thermal characteristic, and contrast, enhancing the recognition ability of the early education robot. This method provides new technical means and solutions for the intelligent application of early education robots, helping to promote the continuous progress and application expansion of early education robot technology. By accurately identifying and reminding users to pay attention to suspicious target areas, the user experience and education effect of the early education robot can be improved, enhancing user trust and satisfaction with the early education robot.

[0088] In a preferred embodiment of the present application, step 3, according to the feature analysis result, the suspicious target region is preliminarily distinguished, and the target set after the distinction is processed by a clustering algorithm to obtain a target list, which can include:

[0089] Step 331, a distinction threshold is set, when the feature analysis result of the suspicious target region is greater than or equal to the distinction threshold, the corresponding feature is determined as a camera target, and the corresponding feature is added to the target set after the distinction;

[0090] Step 332, according to the feature descriptor of the target in the target set after the distinction, including position, shape, and size, clustering parameters are set, including the number of clusters and the distance threshold;

[0091] Step 333, the target set after the distinction is divided into different clustering clusters, and the cluster center of each clustering cluster is extracted;

[0092] Step 334, the cluster centers of all clustering clusters are arranged in order to form a target list.

[0093] In an embodiment of the present application, the early education robot sets a distinction threshold according to its preset safety standard and monitoring requirement, which is used to determine whether the suspicious target region has the feature of the camera target. The early education robot captures the environment image through the camera, and pre-processes and analyzes the feature of the image. For each suspicious target region, the robot calculates its feature value (such as edge density, reflection intensity, infrared thermal imaging feature, etc.). The calculated feature value is compared with the distinction threshold, if the feature value is greater than or equal to the threshold, it is determined that the suspicious target region is a camera target. The feature (such as position coordinates, shape description, size, etc.) corresponding to the region determined as the camera target is added to the target set after the distinction.

[0094] Step 332, the early education robot analyzes the feature descriptor of the target set after the distinction, to understand the distribution of the camera targets in terms of position, shape, size, etc. According to the analysis result, the robot can determine whether the camera targets present aggregation and the degree of aggregation. According to the feature distribution of the target set and the application requirement of the early education robot, the number of clusters (i.e. how many clustering clusters the camera targets are expected to be divided into) and the distance threshold (i.e. the distance standard used to determine whether two camera targets belong to the same clustering cluster) are set. The number of clusters can be determined based on the number of camera targets, distribution density, and recognition accuracy. The distance threshold can be determined based on the position distance between the camera targets, shape similarity, or size difference.

[0095] Step 333, the early education robot uses a clustering algorithm (such as K-means, DBSCAN, etc.) to cluster the target set after discrimination. According to the set number of clusters and distance threshold, the camera targets are divided into different cluster clusters, and each cluster cluster represents a group of camera targets similar in position, shape, size, etc. For each cluster cluster, the early education robot calculates its cluster center. The cluster center can be the average of the position coordinates of all camera targets in the cluster cluster, the center point of the shape description, or the median of the size, etc. The cluster center serves as the representative of the cluster cluster.

[0096] Step 334, the early education robot arranges the cluster centers of all cluster clusters in order. The arrangement order can be based on the position coordinates of the cluster center, the priority of the shape description, or the size of the size, etc. The arranged cluster centers form an ordered target list, which contains the cluster center information of the target set of all camera targets in the early education environment. The robot can display the target list on the user interface for the user to view and process.

[0097] Suppose the early education robot detects multiple suspicious target areas in a child's room and performs feature analysis.

[0098] The robot sets the discrimination threshold to 0.7 (based on historical data analysis), captures the room image, and performs preprocessing and feature analysis on the image. It is found that the feature values of three regions are greater than or equal to 0.7, and they are determined as camera targets. The features of the three camera targets are added to the target set after discrimination. The robot performs feature descriptor analysis on the target set after discrimination and finds that the camera targets are clustered at different positions in the room. The number of clusters is set to 2 (expecting to divide the camera targets into two cluster clusters), and the distance threshold is set to 0.2 meters (based on the distance between the camera targets). The K-means clustering algorithm is used to cluster the target set after discrimination. The camera targets are divided into two cluster clusters, located in the upper left corner and the lower right corner of the room, respectively. The cluster centers of each cluster cluster are calculated, and the position coordinates of the two cluster centers are obtained. The cluster centers of the two cluster clusters are arranged in order (such as arranging the cluster centers from left to right according to their positions). Form a target list that contains the position coordinates of the cluster centers of the two cluster clusters. Display the target list on the user interface to remind the user to pay attention to the camera targets in the room.

[0099] Through preliminary identification and clustering processing, the early education robot can quickly and accurately identify the camera targets in the room, improving the monitoring efficiency. The early education robot can timely discover and remind the user to pay attention to the camera targets in the room, enhancing the safety guarantee of children. Through the display of the target list, the early education robot can make the user more intuitively understand the camera distribution in the room, optimizing the user experience. This method provides new technical means and solutions for the intelligent application of early education robots, which helps to promote the continuous progress and application expansion of early education robot technology. For example, the early education robot can perform more refined monitoring and management according to the camera target information in the target list.

[0100] In a preferred embodiment of the present application, step 4, constructing a convolutional neural network model and a random forest model according to the target list, and respectively using visible light images and infrared images to extract features and identify targets in the target list to obtain preliminary identification results, can include:

[0101] Step 441, according to the target list, a convolutional neural network model is constructed, including convolutional layers, pooling layers, and fully connected layers; the visible light image is input into the convolutional neural network model, the high-level semantic features in the image are extracted through convolutional layer and pooling layer operations, including shape, color pattern, and texture, and the high-level semantic features are input into the fully connected layer of the convolutional neural network model for classification, outputting the visible light image preliminary identification result, including target type, position, and confidence information;

[0102] Step 442, according to the target list, a random forest model is constructed; the infrared image corresponding to the target is determined from the target list as training data, and the image is feature extracted, including calculating temperature distribution and infrared reflectivity bottom texture features; the infrared image features are input into the random forest model, each decision tree in the random forest model makes independent judgment according to the features, and outputs the infrared image preliminary identification result, including target type, position, and confidence information.

[0103] Step 443, integrating the visible light image preliminary identification result and the infrared image preliminary identification result to obtain the integrated preliminary identification result.

[0104] In the embodiment of the present application, a convolutional neural network (CNN) model suitable for processing visible light images is designed according to the characteristics of the target list in the early education robot monitoring environment. The input layer (for receiving visible light images), multiple convolutional layers (for extracting spatial features of images), pooling layers (for reducing the dimensionality of feature maps and preserving important features), fully connected layers (for mapping features to target categories), and output layers (for giving recognition results) are included. The early education robot captures visible light images through the camera and passes them as input data to the CNN model. In the CNN model, the convolutional layer extracts low-level features such as shape and color pattern of the image by convolution operation with the input image. The pooling layer reduces the dimensionality of the feature map output by the convolutional layer, preserving important high-level semantic features. The high-level semantic features output by the pooling layer are input into the fully connected layer for classification of target categories. The output layer gives the target type (such as camera, toy, person, etc.), the position of the target in the image, and the confidence information (indicating the degree of certainty of the model's recognition result).

[0105] Step 442, according to the characteristics of the target list in the early education robot monitoring environment, a random forest model suitable for processing infrared images is constructed. The random forest model is composed of multiple decision trees, each tree makes independent judgments, and finally gives the recognition result through the voting mechanism. From the target list, the corresponding infrared image of the target is determined as the training data, and the features of the infrared image are extracted, including calculating the temperature distribution, infrared reflectivity and underlying texture features of the target in the image. These features can reflect the unique properties of the target in the infrared band, which helps to distinguish different types of targets. The extracted infrared image features are used to train the random forest model. In the recognition stage, the new infrared image features are input into the random forest model, each decision tree makes independent judgments according to the features, and finally the random forest model integrates the outputs of all decision trees to give the target type, position and confidence information.

[0106] Step 443, the early education robot adopts a suitable strategy (such as weighted average, voting mechanism, etc.) to integrate the preliminary recognition results of visible light images and infrared images. This strategy should consider the complementarity of the two kinds of image information to improve the accuracy and reliability of recognition. Through the integration strategy, the early education robot gets the integrated preliminary recognition result, which integrates the advantages of visible light and infrared image information, and can more accurately reflect the true situation of the target in the monitoring environment. The integrated result includes target type, position and confidence information, which provides a basis for the subsequent decision and processing of the early education robot.

[0107] Suppose the early education robot is monitoring a child activity area, and the target list contains a camera target. A visible light image containing the camera is captured. The image is input into the convolutional neural network model, and the model extracts features such as the shape and color pattern of the camera through convolutional layers and pooling layers. The fully connected layer maps these features to the "camera" category and outputs the recognition result, including the target type (camera), location (coordinates in the image), and confidence (e.g., 95%). The early education robot also captures an infrared image containing the camera. The infrared image is processed to extract features such as the temperature distribution and infrared reflectivity of the camera. These features are input into the random forest model, and each decision tree makes an independent judgment based on the features. The final output includes the target type, location, and confidence. The visible light and infrared image recognition results are integrated through weighted averaging or voting mechanisms to obtain the final recognition result: the target is a camera, the location is accurate, and the confidence is as high as 98%.

[0108] By combining visible light and infrared image information, the early education robot can more accurately identify targets in the monitoring area, which helps to timely discover and handle potential safety hazards such as unauthorized cameras. Visible light and infrared images have different environmental adaptability. In low light or blocked conditions, infrared images can still provide effective target information. By integrating both image information, the early education robot can maintain stable monitoring performance in various environmental conditions. Accurate identification and monitoring can provide a safer and more comfortable learning environment for children, and parents and educators can be more confident in allowing children to freely move under the supervision of the early education robot.

[0109] In a preferred embodiment of the present application, step 5 above compares the preliminary recognition result with the target information in the early education robot's preset database, and uses the flower pollination algorithm to optimize and adjust the parameters of the convolutional neural network model and the random forest model to obtain the final recognition result and target location information, which can include:

[0110] Step 551 compares the preliminary recognition result with the target information in the early education robot's preset database to determine the difference between the preliminary recognition result and the target.

[0111] Step 552 initializes the relevant parameters of the flower pollination algorithm, including the number of flowers, pollination probability, and local search radius, based on the difference between the preliminary recognition result and the target, and optimizes the parameters of the convolutional neural network model and the random forest model. Each flower performs local search and global pollination according to the evaluation function to obtain the optimized parameters output by the flower pollination algorithm.

[0112] Step 553, according to the optimized parameters output by the flower pollination algorithm, the target in the target list is re-identified, and a new recognition result is obtained;

[0113] Step 554, the new recognition result is compared with the target information in the early education robot preset database again, and a new error is calculated;

[0114] Step 555, repeat the process of evaluation function design, parameter optimization, re-identification and error calculation until the preset iteration number is reached, and the recognition result obtained by the last iteration is taken as the final recognition result and target position information.

[0115] In the embodiment of the application, the early education robot loads the preset target information database from its internal storage, which contains the feature description, type, position and other information of known targets. The integrated preliminary recognition result obtained in step 443 (including target type, position and confidence information) is compared with the target information in the preset database one by one, and the difference between the preliminary recognition result and the target is calculated, such as type matching degree, position deviation, etc.

[0116] Step 552, according to the difference between the preliminary recognition result and the target, set the related parameters of the flower pollination algorithm, including the number of flowers (representing different solutions or parameter combinations), pollination probability (controlling the proportion of global search and local search), local search radius (determining the range of local search). For each flower, use the evaluation function to calculate the recognition performance under its current parameter combination, according to the pollination probability, carry out global pollination (randomly select another flower's parameters for crossover) or local search (fine-tune in the vicinity of the current parameter combination), update the parameter combination of the flower, and repeat the above process until the preset iteration number or convergence condition is reached. Finally, the flower pollination algorithm outputs the optimized parameters of the convolutional neural network model and the random forest model.

[0117] Step 553, apply the optimized parameters output by the flower pollination algorithm to the convolutional neural network model and the random forest model. Use the optimized model to re-identify the target in the target list, and obtain a new recognition result (including target type, position and confidence information).

[0118] Step 554, compare the new recognition result with the target information in the preset database again, and calculate the new recognition error according to the comparison result, such as type misrecognition rate, position error, etc.

[0119] Step 555, repeat the process of steps 552 to 554, each iteration uses the last iteration of the optimized parameters for model training and identification. After each iteration, update the design of the evaluation function to more accurately reflect the identification performance of the model, when the preset number of iterations is reached, stop the optimization process, and the identification result of the last iteration is taken as the final identification result and target position information, output to the early education robot for subsequent processing or decision-making.

[0120] Suppose the early education robot is monitoring a child activity area and has preliminarily identified an unknown object (such as a new toy).

[0121] Compare the preliminary identification result with the target information in the preset database, find that the object is not in the database, initialize the flower pollination algorithm, set the number of flowers, pollination probability and local search radius and other parameters. Through the flower pollination algorithm, the early education robot continuously optimizes the parameters of the convolutional neural network model and the random forest model to improve the identification performance. Using the optimized model, the early education robot re-identifies the unknown object and gets a more accurate identification result (such as determining that the object is a new toy). The early education robot compares the new identification result with the preset database and calculates the identification error. After multiple iterations of optimization, the early education robot gets the final identification result and target position information, and outputs it for subsequent processing or decision-making.

[0122] Through the flower pollination algorithm to optimize the model parameters, the early education robot can more accurately identify the target, reduce the misidentification and missed identification. The global search and local search ability of the flower pollination algorithm makes the early education robot adapt to different environments and target changes, improves the robustness of identification. Accurate identification result and target position information provide more reliable data support for the early education robot, help to improve user experience and satisfaction. Through continuous optimization of model parameters and identification algorithm, the early education robot can better adapt to complex and variable environment, promote its development to more intelligent and more autonomous direction.

[0123] In a preferred embodiment of the present application, the calculation process of the evaluation function is as follows:

[0124] Traverse all target types, for each target type, compare with the predicted type, if the target type is the same as the predicted type, set the indication value of the corresponding case to 1, otherwise set to 0;

[0125] All cases with indication value 1 are accumulated to get the number of correctly identified target types, and the number of target types and the preset weight coefficient are combined to get the weighted value of type identification accuracy;

[0126] Traverse all target positions, for each target position, compare with the predicted position, if the target position is same as the predicted position, set the indication value of the corresponding case to 1, otherwise set to 0;

[0127] Accumulate all cases with indication value 1, obtain the number of correctly identified target positions, and combine the number of target positions and the preset weight coefficient to obtain the weighted value of the position recognition accuracy;

[0128] Traverse all recognition results, for each recognition result, calculate the absolute value of the difference between the confidence and the matching probability;

[0129] According to the absolute value and the weight of the corresponding recognition result, obtain the sum of the weighted confidence error;

[0130] According to the sum of the weighted confidence error and the normalization factor, obtain the normalized confidence error, and combine the normalized confidence error with the preset weight coefficient to obtain the weighted value of the confidence error;

[0131] According to the weighted value of the type recognition accuracy and the weighted value of the position recognition accuracy, obtain the preliminary evaluation value;

[0132] On the basis of the preliminary evaluation value, adjust the weighted value of the confidence error to obtain the unnormalized evaluation result;

[0133] Proportionally adjust the unnormalized evaluation result and the total number of identified targets for normalization processing, and finally obtain the value of the evaluation function.

[0134] In the embodiment of the application, traverse all target types j (j = 1, 2,..., M), for each target type T j , compare with the predicted type given by the user. If , the indication value , otherwise, Accumulate all cases with indication value 1 to obtain the number of correctly identified target types Combine the number with the preset weight coefficient a to obtain the weighted value of the type recognition accuracy Traverse all target positions k (k = 1, 2,..., L), for each target position P k , compare with the predicted position given by the user. If , the indication value , otherwise, Accumulate all cases with indication value 1 to obtain the number of correctly identified target positions Combine the number with the preset weight coefficient b to obtain the weighted value of the position recognition accuracy

[0135] Iterate through all recognition results i (i = 1, 2,..., N), for each recognition result, calculate the confidence C i and the difference between the matching probability P i , the absolute value of the difference |C i -P i | is obtained. According to the absolute value and the weight w(i) of the corresponding recognition result, the sum of the weighted confidence error is obtained Adjust the sum by a normalization factor Z to obtain the normalized confidence error. Fuse the normalized confidence error with a preset weight coefficient c to obtain the weighted value of the confidence error According to the weighted value of the type recognition accuracy and the weighted value of the position recognition accuracy, a preliminary evaluation value is obtained On the basis of the preliminary evaluation value, the weighted value of the confidence error is adjusted (i.e. subtracted) to obtain the unnormalized evaluation result. Divide the unnormalized evaluation result by the total number of recognition targets N to perform normalization processing, and finally obtain the value of the evaluation function

[0136] Through the evaluation function, the early education robot can more comprehensively evaluate its recognition performance, including the accuracy of type recognition and position recognition, and the consistency between confidence and matching probability. The evaluation function considers the recognition weights of different types and positions, as well as the influence of confidence error, making the evaluation result more objective and accurate. Through normalization processing, the evaluation result can be compared and analyzed between different recognition tasks, which helps the early education robot to continuously optimize its recognition algorithm and performance. In addition, the evaluation function can also provide feedback and reference for the users of the early education robot, helping them understand the recognition ability and effect of the robot, so as to better utilize the robot for early education activities.

[0137] In a preferred embodiment of the present application, step 6, according to the final recognition result and the target position information, the early education robot performs the corresponding interactive response, including playing education content related to the recognition target, having a conversation with the child, and displaying expression animation related to the recognition target, which can include:

[0138] In the embodiments of the present application, the recognition result includes the recognized target type (such as animals, plants, articles, etc.) and the detailed features (such as color, shape, size, etc.) of the target, and the target position information indicates the specific position of the target in the camera field of view. The robot analyzes the received recognition result and target position information, and extracts key information such as target type, features, and position coordinates. According to these information, it can judge the educational opportunities existing in the current environment and prepare the corresponding interactive content. According to the recognized target type, the early education robot selects the educational content related to the recognized target from its built-in educational content library. The robot plays the educational content in the form of voice to the children through the built-in audio playback system. For example, if a cat is recognized, the robot may play educational content about the living habits and interesting things of the cat. The early education robot can also have real-time dialogue with children according to the recognition result and target position information. The robot can understand the children's questions and give corresponding answers through speech recognition and synthesis technology. For example, if the child asks "What is this animal?", the robot can answer "It's a cat!"

[0139] The early education robot can also display expression animations related to the recognized target on its display screen to increase the interest of interaction. According to the recognized target type and features, the robot can select corresponding expressions and actions for display. For example, if a happy cat is recognized, the robot can display a smiling cat expression animation with cheerful sound effects. During the execution of the interactive response, the early education robot will continuously monitor the changes in the camera field of view. If the recognition result or target position information changes, the robot will update the interactive response content in time to maintain the coherence and interest of the interaction with the children.

[0140] As shown in Figure 2 The embodiments of the present application also provide an interactive control system based on robot control, which comprises:

[0141] A difference processing module is configured to perform difference processing on the visible light image and the infrared image in the early education environment, and extract a suspicious target area in the difference image;

[0142] A feature analysis module is configured to design a feature descriptor for the suspicious target area, and perform feature analysis on the suspicious target area according to the feature descriptor to obtain a feature analysis result;

[0143] A discrimination and clustering module is configured to preliminarily discriminate the suspicious target area according to the feature analysis result to obtain a target set after discrimination, and process the target set after discrimination through a clustering algorithm to obtain a target list;

[0144] The recognition module is configured to construct a convolutional neural network model and a random forest model according to the target list, and perform feature extraction and recognition on the targets in the target list by using visible light images and infrared images respectively to obtain preliminary recognition results.

[0145] The comparison and optimization module is configured to compare the preliminary recognition results with target information in a database preset by the early education robot, and optimize and adjust parameters of the convolutional neural network model and the random forest model by using an optimization algorithm to obtain final recognition results and target position information.

[0146] The interaction response module is configured to cause the early education robot to perform corresponding interaction responses according to the final recognition results and the target position information, including playing education content related to the recognized target, having a conversation with the child, and displaying an expression animation related to the recognized target.

[0147] It should be noted that the system corresponds to the above method, and all implementation manners in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0148] Embodiments of the present application also provide a computing device, comprising a processor and a memory storing a computer program, wherein the computer program is executed by the processor to perform the method described above. All implementation manners in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0149] Embodiments of the present application also provide a computer readable storage medium storing instructions, which, when executed on a computer, cause the computer to perform the method described above. All implementation manners in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0150] The above is the preferred embodiment of the present application, and it should be noted that for those skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, which should also be considered as the protection scope of the present application.

Claims

1. A method of interaction control based on robot control, characterized in that, The method comprises: Difference processing of visible light images and infrared images in an early education environment to extract suspicious target regions in the difference images; Designing a feature descriptor for the suspicious target regions and performing feature analysis on the suspicious target regions according to the feature descriptor to obtain feature analysis results; Preliminary discrimination of the suspicious target regions according to the feature analysis results, and processing of the target set after discrimination through a clustering algorithm to obtain a target list; According to the target list, a convolutional neural network model and a random forest model are constructed, and features of targets in the target list are extracted and recognized using visible light images and infrared images respectively to obtain preliminary recognition results; Comparing the preliminary recognition results with target information in a database preset by an early education robot, and optimizing and adjusting parameters of the convolutional neural network model and the random forest model using a flower pollination algorithm to obtain final recognition results and target position information; According to the final recognition results and the target position information, the early education robot performs corresponding interactive responses, including playing educational content related to the recognized target, having a conversation with the child, and displaying an expression animation related to the recognized target.

2. The robot control-based interaction control method according to claim 1, characterized by, Difference processing of visible light images and infrared images in an early education environment to extract suspicious target regions in the difference images, comprising: Using an in-built camera device of the early education robot to collect visible light images and infrared images in the early education environment; Difference processing of the visible light images and the infrared images by calculating the difference between pixel values to obtain difference-processed images; Analyzing the difference-processed images, including threshold segmentation, edge detection, and morphological processing, to identify suspicious target regions in the early education environment, including dropped toys and un-homed teaching aids.

3. The robot control-based interaction control method according to claim 2, wherein Designing a feature descriptor for the suspicious target regions and performing feature analysis on the suspicious target regions according to the feature descriptor to obtain feature analysis results, comprising: For the circularity feature, according to a preset threshold, judging whether the shape of the suspicious target region conforms to the typical shape of objects in the early education environment, including circular toys and spherical teaching aids; for the high-light area area feature, analyzing the size of the high-light area and its proportion in the suspicious target region to identify objects with reflective properties, including metal toys and mirror teaching aids; for the infrared light before and after the gray scale change feature, analyzing the amplitude and trend of the gray scale change to determine whether the suspicious target region includes objects with specific thermal properties, including electronic devices that are heating and heat-sensitive materials; for the local contrast feature, according to the strength of the contrast, judging the degree of differentiation of the suspicious target region from the surrounding environment in the image; According to the analysis of circularity, high-light area area, infrared light before and after the gray scale change, and local contrast, the feature analysis results are obtained.

4. The robot control-based interaction control method according to claim 3, characterized by, According to the feature analysis results, the suspicious target regions are preliminarily discriminated, and the target set after discrimination is processed through a clustering algorithm to obtain a target list, comprising: Setting a discrimination threshold, when the feature analysis result of the suspicious target region is greater than or equal to the discrimination threshold, the corresponding feature is determined as a camera target, and the corresponding feature is added to the target set after discrimination. According to the feature descriptors of the targets in the target set after discrimination, including position, shape, size, set clustering parameters, including the number of clusters, distance threshold; Divide the target set after discrimination into different clustering clusters, and for each clustering cluster, extract the cluster center; Arrange the cluster centers of all clustering clusters in order to form a target list.

5. The robot control-based interaction control method according to claim 4, wherein According to the target list, construct a convolutional neural network model and a random forest model, and respectively use visible light images and infrared images to extract features and identify targets in the target list to obtain preliminary identification results, including: According to the target list, a convolutional neural network model is constructed, including convolutional layers, pooling layers, and fully connected layers; the visible light image is input into the convolutional neural network model, the high-level semantic features in the image are extracted through convolutional layer and pooling layer operations, including shape, color pattern, and texture, and the high-level semantic features are input into the fully connected layer of the convolutional neural network model for classification, outputting visible light image preliminary identification results, including target type, position, and confidence information; According to the target list, a random forest model is constructed; the infrared image corresponding to the target is determined from the target list as training data, and the image is feature-extracted, including calculating temperature distribution and infrared reflectivity bottom texture features; the infrared image features are input into the random forest model, each decision tree in the random forest model makes independent judgments according to the features, and outputs infrared image preliminary identification results, including target type, position, and confidence information; Integrate the visible light image preliminary identification results and the infrared image preliminary identification results to obtain integrated preliminary identification results.

6. The robot control-based interaction control method according to claim 5, wherein Compare the preliminary identification results with the target information in the early education robot preset database, and optimize and adjust the parameters of the convolutional neural network model and the random forest model using the flower pollination algorithm to obtain the final identification results and target position information, including: Compare the preliminary identification results with the target information in the early education robot preset database to determine the differences between the preliminary identification results and the target; According to the differences between the preliminary identification results and the target, initialize the related parameters of the flower pollination algorithm, including the number of flowers, pollination probability, and local search radius, and optimize the parameters of the convolutional neural network model and the random forest model, wherein each flower performs local search and global pollination according to the evaluation function to obtain the optimized parameters output by the flower pollination algorithm; According to the optimized parameters output by the flower pollination algorithm, re-identify the targets in the target list to obtain new identification results; Compare the new identification results with the target information in the early education robot preset database again to calculate the new error; Repeat the process of evaluation function design, parameter optimization, re-identification, and error calculation until the preset iteration number is reached, and the identification result obtained in the last iteration is taken as the final identification result and target position information.

7. The robot control based interaction control method according to claim 6, wherein, The calculation process of the evaluation function is as follows: Traverse all target types, and for each target type, compare it with the predicted type. If the target type is the same as the predicted type, set the indicator value of the corresponding case to 1, otherwise set it to 0. Accumulate all cases where the indication value is 1 to obtain the number of correctly identified target types, and combine the number of target types with the preset weight coefficient to obtain the weighted value of the type recognition accuracy; Iterate through all target positions, and for each target position, compare it with the predicted position. If the target position is the same as the predicted position, set the indication value of the corresponding case to 1, otherwise set it to 0; Accumulate all cases where the indication value is 1 to obtain the number of correctly identified target positions, and combine the number of target positions with the preset weight coefficient to obtain the weighted value of the position recognition accuracy; Iterate through all recognition results, and for each recognition result, calculate the absolute value of the difference between the confidence and the matching probability; According to the absolute value and the weight of the corresponding recognition result, obtain the sum of the weighted confidence error; According to the sum of the weighted confidence error and the normalization factor, obtain the normalized confidence error, and combine the normalized confidence error with the preset weight coefficient to obtain the weighted value of the confidence error; According to the weighted value of the type recognition accuracy and the weighted value of the position recognition accuracy, obtain the preliminary evaluation value; On the basis of the preliminary evaluation value, adjust the weighted value of the confidence error to obtain the unnormalized evaluation result; Proportionally adjust the unnormalized evaluation result with the total number of recognized targets for normalization processing, and finally obtain the value of the evaluation function.

8. A robotically controlled interaction control system implementing the method of any one of claims 1 to 7, characterized by Comprise: A difference processing module for performing difference processing on the visible light image and the infrared image in the early education environment to extract suspicious target regions in the difference image; A feature analysis module for designing a feature descriptor for the suspicious target region and performing feature analysis on the suspicious target region based on the feature descriptor to obtain a feature analysis result; A discrimination and clustering module for preliminarily discriminating the suspicious target region based on the feature analysis result to obtain a target set after discrimination, and processing the target set after discrimination through a clustering algorithm to obtain a target list; An identification module for constructing a convolutional neural network model and a random forest model based on the target list, and performing feature extraction and identification on the targets in the target list using the visible light image and the infrared image to obtain a preliminary identification result; A comparison and optimization module for comparing the preliminary identification result with target information in a database preset by the early education robot, and optimizing and adjusting parameters of the convolutional neural network model and the random forest model using an optimization algorithm to obtain a final identification result and target position information; An interactive response module for causing the early education robot to perform corresponding interactive responses based on the final identification result and the target position information, including playing educational content related to the identified target, having a conversation with the child, and displaying an expression animation related to the identified target.

9. A computing device, comprising: Comprise: One or more processors; A storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, so that the one or more processors implement the method of any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a program which is executed by the processor to implement the method of any one of claims 1 to 7.

Images