37 results about "Background distribution" patented technology

An “Image Segmenter” provides a variational energy formulation for segmentation of natural objects from images. In general, the Image Segmenter operates by adopting Gaussian mixture models (GMM) to capture the appearance variation of objects in one or more images. A global image data likelihood potential is then computed and combined with local region potentials to obtain a robust and accurate estimation of pixel foreground and background distributions. Iterative minimization of a “global-local energy function” is then accomplished by evolution of a foreground/background boundary curve by level set, and estimation of a foreground/background model by fixed-point iteration, termed “quasi-semi-supervised EM.” In various embodiments, this process is further improved by providing general object shape information for use in rectifying objects segmented from the image.

The present invention utilizes depth images captured by a depth camera to detect foreground/background. In one embodiment, the method comprises establishing a single background distribution model, updating the background distribution model if a new depth value for the pixel can be represented by the background distribution model, skipping update of the background distribution model if the pixel is before the background, and replacing the background distribution model if the pixel is behind the background. In case that the background distribution model does not exist initially, a new background distribution model is created. In one embodiment of the present invention, the non-meaningful pixels are handled. In another embodiment, fluctuation of the depth value due to noise is handled by using a candidate background distribution model. In yet another embodiment, the noise for pixels around object edges is handled by using a mixture of two background distribution models.

The invention belongs to the radar target detection technology field and relates to a radar target constant false alarm rate detection method. The radar target constant false alarm rate detection method mainly comprises the following steps: 1) the Weibull background distribution is converted to the Location-Scale distribution through the logarithmic transformation; 2) the orderly truncated (or head cut) treatment is carried out to a background sample to suppress the impact of an interference target, a largest unbiased estimator is adopted for estimating the location and the scale parameters; 3) the normalization of samples in a unit to be tested is carried out by using the estimated parameters; 4) the judgment of whether the background distribution is obeyed or not is carried out by utilizing the Anderson-Darling test, if the background distribution is not obeyed, the target is judged to exist, otherwise, no target is judged to exist. The radar target constant false alarm rate detection method utilizes the difference of the distribution characteristics of the target echo and the background clutter, compared with the traditional detection method based on a self-adaptive threshold, the characteristics of the background distribution and the interference target have little impact on the radar target constant false alarm rate detection method, thereby having very strong adaptability to the non-Gaussian environment and the multi-objective interference environment.

The invention provides a distribution network line inspection system based on a tablet computer. Each device of a distribution network line is provided with an identification code which is used for storing license plate information of the corresponding device and only identifying the corresponding device, the identification codes are read through a hand-held terminal which is connected with a background distribution network line inspection server through a wireless network, the hand-held terminal takes identified information as a key field for querying a distribution network line database on the distribution network line inspection server to acquire detailed information about the line at a server side, and the hand-held terminal transmits on-site pictures and written records back to the distribution network line inspection server through the wireless network. The distribution network line inspection system has the advantages that for performing inspection operation, the hand-held terminals (the tablet computer, a smartphone and the like) are used for scanning the bar codes to query the database and acquire the relevant information, so that the real-time quick query of the records and the modification and the maintenance of data are realized, and the efficiency of the inspection operation as well as the data entry and maintenance in the late period are improved.

The invention relates to an ATM self-service business hall behavior analysis method based on depth information. Behavior analysis based on common two-dimensional cameras is adopted in the prior art, and the situation that targets cannot be accurately positioned exists. According to the ATM self-service business hall behavior analysis method, firstly, binocular cameras are adopted for serving as acquisition equipment for the depth information, background modeling is carried out on depth maps, a Gaussian mixture model of each pixel is learned and updated, and background distribution is determined; secondly, the probability value of each pixel in each new depth map is worked out, and clustering segmentation is carried out on a front depth map according to the DENCLUE algorithm; finally, pixels in each human body region in each camera are projected onto the ground, association of projected targets is carried out by using multi-camera offset calibration, and therefore target detection under the multiple cameras is achieved under a global coordinate system. The depth information is used and background modeling of RGB information is combined, so that the stability of target detection is greatly improved, and a good basis is provided for following behavior analysis.

Provided are an apparatus and method for generating an image with a defocused background. According to various aspects, a preview image is used as the basis for extracting a background distribution and a defocused background is generated based on the extracted background distribution. Accordingly, it is not necessary to photograph two or more images to generate a defocused background effect.

The invention provides a wavelet transform-based thresholding method of an image, and belongs to the field of image processing. In the method, a gray scale of a complete natural scenery image is subjected to wavelet decomposition by utilizing the excellent denoising characteristic of wavelet; foreground characters in the image are removed as noise by virtue of low-pass filter so as to acquire approximate background distribution and foreground distribution; a global threshold is calculated according to the foreground distribution, the global threshold and the background distribution are superposed to form a local threshold which is finally used for image thresholding. The thresholding method provided by the invention can be used for quickly and effectively separating the character part as the foreground to eliminate interference of a complex background, and provides advantages to subsequent character cutting and identifying work. According to the thresholding method provided by the invention, the problem of OCR (optical character recognition) of the natural scenery image can be effectively solved.

The invention provides a hyperspectral image anomaly target detection method based on a pixel selection process. The method comprises the following steps: 1) normalization preprocessing is carried out on the hyperspectral image, and the size of a detection window is set; 2) a sliding window technique is adopted, a to-be-detected hyperspectral image block X is acquired, wherein X belongs to R<N*D>, R is real number space, N is the pixel point number in the image block, and D is the spectral waveband number for the hyperspectral data; 3) a reconstruction error epsiloni of each pixel point Xi is calculated; 4) a vertex-edge weight graph is constructed; 5) a pixel selection process model is constructed; and 6) steps from the second to the fifth are repeated until detection on the anomaly target on the whole hyperspectral image is completed, and a final anomaly probability graph is acquired. The hyperspectral image anomaly target detection method has the advantages of solving technical problems that background distribution needs to be hypothesized in the existing detection technology, the adaptability is poor, differences between the background and the target are not enough, and the detection ability is low.

The invention discloses an image salient region detection method, apparatus and device, and a computer readable storage medium. The method comprises the steps of pre-segmenting an obtained to-be-detected image, thereby segmenting the to-be-detected image into multiple sub-regions; according to preset rules, calculating preliminary region saliency values of the sub-regions; according to region roundness and region distribution characteristics of the sub-regions, calculating background distribution measurement of the sub-regions; and by utilizing the preliminary region saliency values and the background distribution measurement of the sub-regions, determining target region saliency values of the sub-regions.

The invention discloses a knowledge-aided nonparametric constant false alarm detection method to solve detection problems which may occur in the non-uniform environment with unknown distribution, and especially relates to the knowledge assisting, nonparametric estimation and constant false alarm detection technologies, belonging to the technical field of radar knowledge assisting. The method mainly comprises four steps: the first step is to acquire a reference unit which is as independently and identically distributed as possible with a to-be-detected unit by using priori supplementary knowledge; the second step is to obtain an amplitude probability density function of the to-be-detected unit by using a nonparametric PDF estimation method; the third step is to calculate a detection threshold based on the PDF obtained in the previous step and a set false alarm probability; the last step is to compare the echo of the to-be-detected unit and the threshold obtained in the third step, and determining whether there is a target. In the invention, a uniform reference unit is obtained according to the priori information, and then background distribution is accurately estimated and the detection threshold is calculated, thereby improving the CFAR detection performance when the distribution is unknown in the non-uniform environment.

The invention discloses a sectional non-linear enhancement method for a urinary sediment image. The method comprises the following steps of: a first step: calculating section points by using a genetic algorithm according to distribution ranges of cells in the image and background in a gray space, dividing the whole image into a background section, a target section and a transition section according to the section points, calling a gray-level zone of distributed background as the background section, counting the gray-level zone of each distributed cell, ensuring that the gray-level zones in which the backgrounds and the cells are distributed have overlapped and crossed areas to find out the most proper thresholds to divide the two zones, and selecting a gray-level zone with less than 100 pixels as the transitional section; and a second step: performing different gray-level transformation method on the background section, the target section and the transition section. The method has the beneficial effects that the enhancement method is high in processing speed and stable and reliable; and the noise of the background is well inhibited, and the whole image is obvious in vision effect.

The invention relates to a method and a device for processing user data. The method mainly comprises the following steps of: according to behavior data of a user in a certain product, calculating biased interest type distribution information of the user on the product; calculating a difference value between the biased interest type distribution information and user interest background distribution information of the product and obtaining unbiased interest type distribution information of the user on the product; and integrating the unbiased interest type distribution information of the user on various products to obtain integrated interest type distribution information of the user. Due to the utilization of the method and the device for processing the user data, which are disclosed by the embodiment of the invention, the integrated unbiased interest type distribution information of the user can be effectively obtained.

The invention discloses a mine operation guiding system based on APPs. The mine operation guiding system comprises APP terminals, a plurality of radio frequency identification technology (RFID) signal identification base stations, RFID tag cards, a cloud server and a background server. The APP terminals are arranged on bodies of mine personnel in a worn mode; the multiple RFID signal identification base stations are arranged along the trend of a mine channel; the RFID tag cards are arranged inside the APP terminals, and each RFID tag card and the corresponding APP terminal are uniquely bound together; the cloud server establishes communication connection with the APP terminals and comprises a cloud receiving module, a cloud monitoring module and a cloud distribution module; and the background server establishes communication connection with the cloud server and comprises a background receiving module, a background monitoring module and a background distribution module. According to the mine operation guiding system, the real-time positions of the downhole mine personnel can be monitored in real time, the mine personnel are subjected to operation instructions, and the mine personnel can further be guided and helped in the event of unexpected conditions.

The invention discloses an APP based electric vehicle anti-theft tracking system. The system includes a user mobile terminal; a plurality of RFID signal identification base stations arranged along streets; an RFID label card arranged on an electric vehicle; and a cloud end server establishing communication connection with the user mobile terminal and including a cloud end receiving module, a cloud end monitoring module and a cloud end distribution module; and a background server establishing communication connection with the cloud end server and including a background receiving module, a background monitoring module and a background distribution module. The system can monitor whether the electric vehicle is theft or not in real time and can also perform fixed-point monitoring on the theft electric vehicle and perform real time display on the APP in the user mobile terminal.

The invention provides an ultra-pixel-based automatic segmentation method in a dynamic outdoor environment. The ultra-pixel-based automatic segmentation method mainly comprises the steps of calculating low-texture super pixels, generating a threshold hue image by using an Otsu method, estimating background distribution, allocating pixel tags and creating a mask. According to the ultra-pixel-based automatic segmentation method, an image is converted into a hue-saturation value color space and is segmented into super pixels, the super pixels are calculated by using an energy-driven sampling method, the threshold hue image is generated by using the Otsu method, probability distribution of background pixels is estimated by utilizing a super pixel group, the hue image is compared with the probability distribution to generate the pixel tags, and finally a mask image including regions of objects of interest and background objects is created. The ultra-pixel-based automatic segmentation method utilizes the low-texture super pixels, and generates the threshold hue image by using the Otsu method, so that precise segmentation can be carried out in the case of unstable lighting condition, the algorithm is simple and the precision is improved to some extent.

The invention relates to the technical field of video processing, in particular to a video processing method, a system, device and a medium. The method aims at solving the problems that an existing matting method is large in limitation and small in extraction range. In order to achieve the purpose, the video processing method comprises the steps of acquiring a target background image, image frames in a video and a depth distribution map corresponding to the image frames; segmenting the image frame to obtain an image segmentation result; performing thresholding operation on the depth distribution map to obtain a distribution map thresholding result; determining a foreground and background distribution map based on an image segmentation result and a distribution map thresholding result; performing synthesis based on the image frame, the target background image and the foreground and background distribution image to obtain a final image; and outputting the final image. According to the method, the requirement of real-time image matting under the condition of no green screen is met, foreground and background segmentation is carried out by introducing the depth distribution map, the complex foreground and background judgment problem is avoided, and accurate extraction of non-portrait objects in the image frames is achieved.

The invention discloses a data distribution method of an NFC card. The method comprises the steps: setting a target package name of a current distribution Intent as a foreground application when the NFC card is sensed, enabling the background distribution to point to the current foreground application, obtaining a preset NFC application list registered in the background according to the card type of the NFC card, and if the current foreground application exists in the preset NFC application list, carrying out background distribution on NFC data of the current foreground application to the current foreground application. According to the data distribution method provided by the embodiment of the invention, directional distribution of the data of the NFC card is realized in background distribution, and a plurality of preset NFC application lists are not displayed in an Activity selector for a user to select, so that some problems of inconvenience brought to the user due to automatic triggering of background distribution are effectively solved. The invention further discloses a data distribution device, electronic equipment and a storage medium.

The present invention provides a real-time detection method for the uniformity of stray light, comprising the following steps: 1) preprocessing the background and flat fields of the target image; 2) drawing a histogram, and performing star image deduction based on the size and frequency of the ADU; 3 ) Draw the contour map of the background stray light, and calculate the related parameters such as the uniformity of the stray light. The invention provides CCD image surface background distribution and statistical data by detecting the uniformity of stray light in real time, which can assist the selection of comparative reference stars in poor photometry and improve photometry accuracy.