40 results about "Translation algorithm" patented technology

The invention provides a method for counting passenger flow of buses in real time, which adopts means of machine learning, multi-target detection and tracking, target behavior analysis and the like, and belongs to the technical field of pattern recognition. The method comprises the following concrete steps: detecting passenger targets by analyzing shape information and texture information of passenger heads in video images, wherein a column diagram in a gradient direction is used for representing the shape information, and a column diagram in a local binary mode is used for representing the texture information; then, accurately positioning passengers by adopting the target tracking policy combining gray level cross-correlation matching tracking and mean translation algorithm searching tracking; and finally, judging the behavior characteristics of the passengers by analyzing the moving tracks of the passengers, thereby accurately counting the passenger flow of buses. The practice shows that the method provided by the invention can provide an accurate and reliable basis for realizing intelligent scheduling of buses for bus companies, analyzing the acceptability of mobile media advertisements in buses, and the like, and the invention has very high practical value.

The invention discloses a method for generating virtual multi-viewpoint images based on depth image layering, which comprises the following steps: (1) preprocessing depth images to be processed; (2) layering the preprocessed depth images so as to obtain layered depth images; (3) selecting depth layers focused by camera arrays, and determining prospect layers and background layers; (4) layering two dimensional images to be processed so as to obtain layered two dimensional images; (5) calculating valves of parallax errors corresponding to the two dimensional image layers corresponding to various depth layers; (6) expanding the layered two dimensional images so as to obtain expanded layered two dimensional images; (7) obtaining the virtual two-dimensional images of various virtual viewpoint positions by using a weighted level translation algorithm. By using the method of the invention, virtual multi-viewpoint images required by a multi-viewpoint auto-stereoscopic display system can be generated rapidly and effectively without parameters of a virtual multi-viewpoint camera array model, and the method has good fault-tolerant capacity to the input depth images.

The invention provides a real-time bus passenger flow volume statistical method based on a target prior distribution, which adopts the methods of target detection, target tracking, target behavioral analysis and the like, and belongs to the technical field of pattern recognition. The method concretely comprises the following steps: utilizing the target prior grayscale to do statistics of the back-projection operation of a histogram on an input video image, and then performing the treatment of difference, binaryzation, wave filtering, connected domain marking and the like on a back-projection image to realize the target detection; adopting the grayscale to interrelate, match and track and a mean translation algorithm to search and track so as to realize the accurate positioning of the target; and finally analyzing the motion trail of the target to judge the get on or off of passengers, and realizing the passenger flow volume statistics. The invention can provide fine-grained reliable basis for a public transit company to realize intelligent dispatching of buses, audience analysis of the mobile media advertising of the busses and the like.

A new, more efficient memory translation algorithm facilitating the acquisition of a most appropriate translation in a target language from among those of nearly narrowed-down candidates of translation by separately applying the so-called dimension reducing functions of a template automaton and the LSI (latent semantic index) technique. Both the template automaton and the LSI principle play an important role in implementing an efficient process of narrowing down an efficient solution space from among the many example sentences of the databases in a target language by exploiting their respective unique search space reduction function. Once developed into a fully operational system, an expert editor rather than an expert translator can tune up the translation memory system, markedly widening the range of available experts who can utilize the system.

The invention discloses an absolute detection method for a surface shape error of a spherical surface. The absolute detection method comprises the steps of: establishing an equation system about Zernike multinomial coefficients of a measured spherical surface and a reference surface by utilizing measurement data of multiple rotation and concentric translation of the measured spherical surface at a con-focal position and adopting a Zernike multinomial fitting based rotation translation algorithm; and obtaining the Zernike multinomial coefficient by applying a least square method so as to obtain absolute surface shape information of the measured spherical surface and the reference surface. The absolute detection method is also available for the detection of the surface shape error of a plane. As a global optimized idea is adopted to resolve multinomial coefficients of the measured spherical surface and the reference surface at the same time, the absolute detection method for the surface shape error of the spherical surface can suppress system error and random noise for the better, and achieves stronger interference resistance and important application values.

The invention discloses a method for translating an Internet protocol version 4 (IPv4)/Internet protocol version 6 (IPv6) initiating communication based on cloud service by using the IPv4, which relates to the technical field of computer networks. In the method, an information resource database and a domain name resolution server are installed in a cloud service data center; simultaneously, a serial stateless IPv4/IPv6 translator and an IPv4/IPv6 translator configured through the information resource database are installed; an address translation algorithm of the stateless IPv4/IPv6 translator is used for translating an IPv4 address reserved for accessing a pure IPv6 service from an IPv4 network into a special IPv6 address into which the IPv4 address is embedded and an operator prefix is used; an address translation algorithm of the IPv4/IPv6 translator is used for mapping an IPv6 address which corresponds to a pure IPv6 information resource into the special IPv6 address into which the IPv4 address is embedded; and the domain name resolution server is used for returning a record A to a user in response to a domain name resolution request of the user according to a mapping relation. Therefore, an IPv4 user can initiate access to the information resource of a pure IPv6, and the aim of communicating the IPv4 with the IPv6 is fulfilled.

A home environment managing method and system to manage home environment values from a plurality of locations is disclosed. The method allows in an acquisition mode to acquire through a Universal Remote Console a plurality of reference sets of home environment values from the plurality of locations. Each of the plurality of reference sets of home environment values are converted into a normalized set using a respective translation algorithm. Each normalized set is stored with a link to the translation algorithm used during the conversion. Furthermore, in a restore mode, a normalized set of home environment values is converted into a reference set using the translation algorithm linked to it, and then the reference set is restore to the appropriate location through a Universal Remote Console.

The invention relates to a highly-dynamical weak signal GPS capturing method assisted by an inertial navigation system. The method comprises five steps that 1) capturing performances of coherent, non-coherent and differential coherent accumulation are compared, accumulation manners are selected and combined, and signal energy accumulation is reinforced; 2) coherent accumulation is carried out on input intermediate-frequency signals, the inertial navigation system and satellite ephemeris are used to estimate the Doppler parameter and code phase of received signals, and the frequency and the phase searching space are compressed; 3) Kalman filtering and smoothing are carried out on the estimation values of the Doppler parameters and the code phase, and a smoothing result is used to calculate the dynamic frequency/phase offset between local signals and the reception signals; 4) the dynamic offset amount of an output matrix of coherent accumulation is compensated by a cycled translation algorithm; and 5) non-coherent accumulation is carried out on the compensated output matrix, and an output result is compared with a threshold to deciding capture. The dynamic tolerance of GPS signal capturing is improved, and high-sensitivity GPS signal capturing can be realized in the highly dynamic environment.

The invention discloses a spherical surface shape rotation and translation absolute detection method. According to the method of the invention, based on the measurement data of the rotation and translation of a detected spherical surface are utilized, an equation set of Zernike polynomial coefficients which are about the detected spherical surface and a reference surface is constructed through adopting a Zernike polynomial fitting-based rotation and translation algorithm, and the coefficients of a Zernike polynomial are solved through utilizing a least square method, and therefore, the absolute surface shapes of the detected spherical surface and the reference surface can be obtained; and adjustment errors in large-numerical value aperture spherical surface phase-shift detection are adopted as independent error terms, and the adjustment errors are applied to an absolute detection process, and the adjustment errors and interferometer system errors are removed together from test wave surface data, and therefore, computation accuracy can be improved, and the method has an important application value.

The invention discloses a real-time translation method and apparatus. The method comprises the following steps: step 1, obtaining a text input by a user and a corresponding pinyin string in real time; step 2, segmenting the pinyin string according to all possibilities by taking pinyin of a single Chinese character as a unit, taking candidate Chinese characters corresponding to the pinyin as nodes, forming a grid structure by connecting lines of the candidate Chinese characters, scoring a Chinese character string composed of any segmentation path in the grid by using a language model to obtain candidate results, and selecting an optimal Chinese character string from the candidate results; and step 3, calling a translation algorithm to translate the optimal Chinese character string, and outputting a translation result. Through the real-time translation method and apparatus, the technical problem that the user needs to occasionally perform switching between the text and an online dictionary during real-time translation of the input text is solved.

The invention relates to a method for realizing the functions of character extraction and automatic translation of a digital camera. The method generates a picture by taking a picture of characters, extracts the characters through image processing technology, generates a text file, uses a translation algorithm to translate the section of the characters into Chinese, and displays the translated result on an LCD. The method comprises the following steps: firstly, setting the camera into a translation mode according a corresponding item in a menu; secondly, taking the photo of the characters; thirdly, automatically using a character extraction algorithm to extract the characters and generating the text file by the camera; fourthly, automatically utilizing the prior translation algorithm to translate the characters by the camera; and fifthly, displaying the translated result on the LCD. The method can directly utilize the prior hardware resource of the digital camera, does not need to add any hardware cost, makes the digital camera have the functions of character extraction and automatic translation, and can bring convenience for users.

The invention discloses an automatic white balance correcting method which comprises the following steps of 1, respectively acquiring histogram information of each RGB channel of a current image; 2, selecting an appropriate algorithm mode through the characteristic of the image RGB channel histogram, wherein the algorithm mode comprises a histogram matching algorithm and a histogram translation algorithm; 3, performing histogram adjustment, wherein calculation of a new pixel value of the RGB channel is required in a histogram matching mode, and calculation of a translation coefficient is required in a histogram translation mode; and 4, converting each pixel value of the RGB channel, thereby obtaining a corrected image. According to the automatic white balance correcting method, effectiveness determining is performed on the white balance algorithm based on histogram matching through using the distribution characteristic of the RGB channel histogram; according to the characteristics in histogram matching and histogram translation, defects of the existing histogram-based white balance algorithm are compensated; a high correction effect is insured; and furthermore an application rangeof the algorithm is greatly expanded.

The invention relates to the technical field of image processing recognition and automatic translation, and provides an object recognition and real-time translation method which comprises the following steps: acquiring an image shot by a front object in real time through a camera; inputting the image into a convolutional neural network model, and extracting depth feature information of the image;inputting the extracted depth feature information into an image recognition model to recognize the category of the object, and outputting the recognized category of the object; and translating the object category into a target language through a translation algorithm and outputting the target language. The invention further provides a device applying the method which comprises a central processingunit, an image acquisition unit, a display screen, a camera and a device shell, the camera is arranged on one side surface of the device shell, the display screen is arranged on the other side surface of the device shell, and the central processing unit and the image acquisition unit are integrally arranged in the device shell. According to the invention, the current object can be recognized andtranslated, and the learning experience of a user is improved.

The invention discloses an IPv4/IPv6 stateless translation algorithm. By using information included in the head of an application layer protocol, packets with the same IPv4 address and the same port are mapped to different IPv6 addresses through an application layer, and IPv4/IPv6 translation from the same IPv4 address and the same service port to different IPv6 addresses and the same service port is achieved. According to the IPv4/IPv6 stateless translation algorithm, different pure IPv6 servers can be accessed by any IPv4 address in the IPv4 Internet, the pure IPv6 servers can provide services for IPv6 clients, and can also provide services for IPv4 clients by using the same port, the problem of the shortage of a public IPv4 address in a data center is solved effectively, and transition of Internet information resources from IPv4 to IPv6 is promoted.