Intelligent robot with high service level

[0012] The present invention will be further described in conjunction with the following examples.

[0013] See figure 1 , The intelligent robot with high service level of this embodiment includes a central processing unit 1, a vision acquisition system 2, an image processing system 3, an image display system 4, and a voice system 5. The central processing unit 1 and the visual The acquisition system 2, the image processing system 3, the image display system 4, and the voice system are electrically connected to 5. The central processor 1 is used to manage the visual acquisition system 2, the image processing system 3, the image display system 4, and the voice system 5. The visual acquisition system 2 is used to collect images, the image processing system 3 is used to recognize the collected images, the image display system 4 is used to display recognition results, and the voice system 5 is used to broadcast the recognition results .

[0014] The robot recognition performance of this embodiment is good, and the service level of the robot is improved.

[0015] Preferably, the image display system 4 includes a high-definition display screen.

[0016] The display of this preferred embodiment is clear.

[0017] Preferably, the vision acquisition system 2 includes lighting equipment and a high-definition camera.

[0018] The preferred embodiment can also work at night, which improves the night service level of the robot.

[0019] Preferably, the image processing system 3 includes a primary processing module, a secondary processing module, and a tertiary processing module. The primary processing module is used to retrieve an image similar to the image to be recognized from the image data set, and obtain the retrieval result. The secondary processing module is used for sorting the retrieval results according to the similarity between the retrieval results and the image to be recognized, and obtaining a ranking list, and the third processing module takes the retrieval result with the highest similarity as the image recognition result.

[0020] The primary processing module includes a primary processing unit and a secondary processing unit, the primary processing unit is used for mining similar relationships between images in the image data set, and the secondary processing unit is used for retrieving images from the image data set. The mining of the similarity relationship between the images is carried out using the following steps: Step 1. A given image data set SY={x 1 ,x 2 ,...,X n } And several distance measures {FN 1 ,FN 2 ,...,FN m }, let any two images x in SY i And x j Measuring FN l The distance down is s l (x i ,x j ), where l∈[1,m]; then for any metric FN l , There exists a directed graph G l (AY,EU,w l ), where AY=SY is the set of vertices, Is the set of directed edges, w l Used to calculate the weight of any edge, and w l (x i ,x j ) Is abbreviated as w l (i,j); In the formula, β ij Is the scale factor, Among them, av means average, x i (N) and x j (N) respectively represent x i And x j The sum of the distance to the first N data with the smallest distance;

[0021] Step 2. Use kN l (x i ) Means x i In the directed graph G l K nearest neighbors below, get the directed graph G 1 ,G 2 ,...,G m K nearest neighbor graph G k1 ,G k2 ,...,G km , For any k nearest neighbor graph G kl , Where l∈[1,m], only when x j ∈kN l (x i ), the two have edge b ij , The weight is w kl (i,j)=w l (i,j), the weight of the edge in other cases is 0; combine k nearest neighbor graphs into graph G k (AY,EU,w k ), if the sample x i And x j In any graph G kl For edges whose weight is not 0, then G k Edge b ij , Weight w k for: In the formula, q l Indicates the importance index of each metric, If w l 0, then c l =1, otherwise 0, only x j Belongs to x under at least one measurement method i K nearest neighbors, w k (i,j) is not 0;

[0022] Step 3. Directed graph G k Corresponding to a Markov chain on the data set SY, its transition probability matrix HX k =[a kij ] n×n ,among them, In the formula, a kij Represents the Markov system from x i To x j The transition probability;

[0023] Step 4. Establish the propagation process:

[0024] In the formula, kN(x i ) Means x i In Figure G k K nearest neighbors, kN(x j ) Means x j In Figure G k K nearest neighbors, Represents the similarity matrix after the tth iteration, the initial matrix HX k (x i ,x p ) Means from x i To x p The transition probability matrix, HX k (x q ,x j ) Means from x q To x j The transition probability matrix;

[0025] Step 5. Run the propagation process in step 4, and the pairwise similarity relationship between the samples is propagated to a distance. After several steps of iteration, the intrinsic similarity relationship between the samples is obtained.

[0026] In the preferred embodiment, the one-time processing module sets the one-time processing unit to mine the similar relationship between the images in the image data set. Since most irrelevant samples are excluded from the propagation process, the important information in each measurement method is highlighted, so The formed sparse matrix greatly reduces the computational requirements in the iterative process; G k There are at most mkn edges in the middle, which eliminates the background cumulative effect of low-weight relationships, and makes high-weight similar relationships more prominent; one processing unit fuses multiple distance metrics on the data set into a sparse graph, and then uses locally constrained The propagation method evolves and spreads, which helps to mine similar relationships in the graph.

[0027] Preferably, the distance metric includes a first distance metric FN 1 And the second distance metric FN 2 , The first distance metric FN is determined in the following way 1 : In the formula, y K And z K Respectively represent the gray value of the Kth pixel of the two images y and z, W and H are the width and height of the image respectively, |W×H| represents the number of image pixels;

[0028] Use the following method to determine the second distance metric FN 2 :

[0029] In the preferred embodiment, the primary processing unit introduces a new distance measurement method, the first distance measurement and the second distance measurement, to obtain a more accurate image distance and improve the calculation level of image similarity.

[0030] Preferably, the retrieval of images from the image data set is carried out using the following steps: Step 1. Input data set SY={x 1 ,x 2 ,...,X n }, distance measure {FN 1 ,FN 2 ,...,FN m } And its corresponding Neighborhood size k and iteration number T, the output result is the optimized similarity matrix matrix The i-th row corresponds to x i Similarity with all data in the data set;

[0031] Step 2. Select the image with the largest RL before the similarity, and get x i The search result when it is the target, where RL∈[3,7];

[0032] Step 3. Add the image to be recognized into the data set, and obtain an image with high similarity to the image to be recognized according to the method of step 1.

[0033] In the preferred embodiment, the primary processing module is provided with a secondary processing unit, and the similarity matrix of the entire data set can be obtained by calculating the similarity between all samples in the data set. In the image retrieval task, the image to be recognized is added to the data set, the similarity matrix containing the image to be recognized is obtained, and the image similar to the image to be recognized is obtained, and the image retrieval is completed, thereby improving the recognition level of the robot and thus The service level of the robot.

[0034] The intelligent robot with high service level of the present invention is used to provide services. When the RL takes different values, the service efficiency and user satisfaction are counted. Compared with the current robot, the beneficial effects produced are shown in the following table:

[0035] RL

[0036] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the scope of protection of the present invention. Although the present invention has been described in detail with reference to preferred embodiments, those of ordinary skill in the art should understand The technical solution of the present invention can be modified or equivalently replaced without departing from the essence and scope of the technical solution of the present invention.