42results about How to "The mapping relationship is accurate" patented technology

The invention discloses an image super-resolution enhancement method based on a bidirectional recursion convolution nerve network. The method comprises two parts, i.e., model training and resolution enhancement. The model training comprises: obtaining multiple groups of high and low-resolution images comprising the completely same content, extracting high and low-resolution image block sequences through a window sliding method, and accordingly, training a bidirectional recursion convolution nerve network model. The resolution enhancement comprises: segmenting low-resolution images to be processed into a group of low-resolution image blocks to be processed through the window sliding method, generating corresponding high-resolution image blocks through the well trained bidirectional recursion convolution nerve network model, and generating corresponding high-resolution images through fusion. According to the method, space position relations of adjacent image blocks are learnt by full use of the bidirectional recursion convolution nerve network, and accordingly, the high-resolution images comprising detail information such as more edge textures and the like are obtained.

The invention discloses a hybrid deep regression network-based head pose estimation method and belongs to the computer vision and machine learning field. According the method of the invention, on thebasis of a traditional regression algorithm and a deep network framework, image information of different modalities is applied to a head pose estimation problem for the first time; and sub-networks ofa plurality of modalities, which are obtained through training, are fused, so that a final hybrid deep regression network can be obtained. Compared with other head pose estimation methods, the algorithm of the invention can achieve higher estimation accuracy and have high robustness; and in addition, the method of the present invention has a certain universality, and is not only applicable to typical head pose estimation problems such as human-computer interaction, safe driving, and face recognition, but also can expand problem conditions to other deep regression problems.

The invention discloses a SAR image change detection method based on a complex neural network, which inputs two original SAR gray-scale images and uses a preliminary difference map obtained by a traditional method as a preliminary label. According to the preliminary tags, some pixel blocks in the two original images are selected as candidate training samples by confidence detection. A complex network is constructed, which consists of three layers of all connection layers and two layers of multiple batches of normalized layers. The complex training samples are selected from the candidate training samples according to the stochastic proportional method to train the complex network. The training complex network is used to test the test complex samples constructed directly from two original SAR images, and the final change detection results are obtained. The invention not only makes full use of the advantages of the traditional change detection result, but also displays the characteristicsof the original data, so that the neural network can better learn the relationship between the two images, thereby obtaining a better change detection result.

The invention provides a coordinate conversion method and device, relating to the field of a wafer scriber and solving the problems of inexact mapping relationship, influence to quality and efficiency of an image processing algorithm and influence to precision of a machine. The coordinate conversion method comprises the steps of: regarding one point as an original point of a coordinate system, and obtaining a first workbench coordinate difference vector of a workbench from the original point to a first target position and a first image coordinate difference vector of an image of a processing object on the workbench from the original point to the first target position; obtaining a coordinate conversion matrix based on the first workbench coordinate difference vector and the first image coordinate difference vector; performing conversion between the coordinate of the workbench and the coordinate of the image of the processing object based on the coordinate conversion matrix. Pixels are determined through non-single-direction movement of the workbench of the wafer scriber, so that the problem of mapping between the image pixel coordinate and the workbench coordinate is solved, and the mapping precision is improved.

The invention discloses a spatial double-arm system constraint motion planning method based on deep intensive learning. The method comprises the following steps of 1) establishing a kinematic model ofa free floating spatial double-mechanical-arm system by utilizing a generalized jacobian matrix; 2) designing a spatial double-arm system motion planning algorithm based on the kinematic model of thespatial double-arm system in combination with a DDPG algorithm; and 3) designing a reward function in the DDPG algorithm to realize the satisfaction of each constraint condition in the spatial double-arm system motion planning algorithm, including speed constraint of a mechanical arm end effector and self-collision constraint of double-mechanical-arm cooperative motion. The universality and safety of the motion planning algorithm is improved.

The present invention discloses an image-based texture generation method. The method comprises the following steps that: S1, the three-dimensional model and color coding index table of a target to be generated are established, and the index table are mapped onto the surface of the three-dimensional model as textures; S2, at least two images of the target to be generated at different angles are obtained; S3, outer contour matching is performed on the images and the three-dimensional model; S4, the index table is utilized to index the images which have been subjected to the outer contour matching, and the texture features of each of the images of the target to be generated are extracted; S5, different texture features are fused; and S6, the fused textures are mapped onto to the surface of the three-dimensional model as surface textures. According to the texture generation performed according to the steps of the method of the present invention, the textures can be generated with manual texture drawing not required; and accurate mapping relations can be established with manual adjustment of the locations of geometric model vertexes and texture details not required. The method has the advantages of high efficiency, simplicity and high automation degree.

The invention provides a three-dimensional image reconstruction method, comprising steps of extracting a two-dimensional image feature point of an input image; estimating the two-dimensional image feature point by using a first mapping relationship to obtain a three-dimensional image feature point corresponding to the two-dimensional image feature point, wherein the first mapping relationship is obtained by learning a mapping relationship between an original two-dimensional image feature point and an original three-dimensional image feature point in a training sample; expanding the training sample based on a second mapping relationship, wherein the second mapping relationship is obtained by learning a mapping relationship between the two-dimensional image feature point and the three-dimensional image feature point; and acquiring a full-pixel image based on the extended training sample and the three-dimensional image feature point. The three-dimensional image reconstruction method provided by the invention can expand the training sample, and effectively solves the problem of insufficient training samples in the full-pixel image reconstruction process.

The invention provides a controller area network (CAN) signal transmission verifying method for an automobile controller. The CAN signal transmission verifying method for the automobile controller comprises the following steps: according to the CAN signal, obtaining the first value range of the transmission of the CAN signal between a CAN bus and the bottom procedure of the controller, the second value range of the transmission between the bottom procedure and the intermediate port of the controller and the third value range of the transmission between the intermediate port and the application procedure of the controller; according to each value range, generating a test valve sequence; sending the corresponding the test value from the test valve sequence to the controller; according to the test value, generating the corresponding feedback value by the controller; and according to the test value and the feedback value, judging whether the CAN signal transmission of the controller is normal. The embodiment is capable of roundly and systemically detecting the CAN signal transmission of the controller, and avoiding the phenomenon of missed sanction. The invention further provides a CAN signal transmission verifying system for the automobile controller.

The invention provides a steel support axial force intelligent adjustment method based on a bp neural network. The method comprises the following steps that an initial axial force is set according toa design value; data of loading days of axial force of a steel support, an axial force mean value of the steel support every day and accumulated deformation of a foundation pit every day are acquired,training is conducted by taking the three data as samples through a bp neural network model, and parameters of the bp neural network model are stored after the training result meets requirements; theaxial force mean value of the same day and the loading days of the next day in the sample data are input into the bp neural network model for calculation to obtain the predicted deformation of the foundation pit of the next day; if the predicted deformation of the foundation pit is smaller than a threshold value, the initial axial force is kept; and if the predicted deformation amount is larger than the threshold value, the axial force is adjusted and continuously input into the bp neural network model for calculation until the predicted deformation amount of the foundation pit meets the requirement, and the finally adjusted axial force is taken as a set axial force. The purpose of actively adjusting the axial force in advance to control the deformation of the foundation pit before the deformation of the foundation pit occurs is achieved.

Embodiments of the invention disclose an upgrading processing method and a related device. The upgrading processing method comprises the steps that the upgrading processing device sends an interface conversion description file Dv2-v1 to a device in which a business software module A is located, wherein the interface conversion description file Dv2-v1 is used for describing an interface conversion algorithm for conversion from a business software module B with a version v2 to the business software module A with a version v1; and upgrading data UB_v1-v2 and an interface conversion description file Dv1-v2 are sent to a device in which the business software module B is located, wherein the upgrading data UB_v1-v2 is used for upgrading the version of the business software module B from the version v1 to the version v2, and the interface conversion description file Dv1-v2 is used for describing an interface conversion algorithm for conversion from the business software module A with the version v1 to the business software module B with the version v2. According to the technical scheme provided by the embodiments of the invention, the business continuity in a gray-level upgrading process can be ensured as far as possible.