37results about How to "Reduce false matches" patented technology

The invention discloses a crop spraying positioning method based on a binocular vision gridding partition matching algorithm and belongs to the technical field of medicine application of crops. The method comprises the following steps of: utilizing double cameras to calibrate target crops and obtaining an image of the target crops; respectively obtaining a binary image of a left camera target crop and a binary image of a right camera target crop; respectively carrying out gridding division on the binary image of the target crop obtained by the left camera and the binary image of the target crop obtained by the right camera; carrying out matched searching on each point of a left side gridding region in a right side gridding region to obtain mutually-matched points to form a matched point pair; calculating a left and right image vision difference by each matched point pair and solving a three-dimensional coordinate of the point of the target crop corresponding to the matched point pair; deleting an error point and carrying out fitting treatment on the point of the target crop to obtain a fitting curve or curve surface; and planning a sprayer path according to the fitting curve or curve surface. The crop spraying positioning method disclosed by the invention realizes extraction and positioning of a three-dimensional information outline of the target crop.

Devices, methods, and programs correct an estimated path found by dead reckoning navigation. A point correlation value s is calculated based on a distance to a reliability circle whose radius is a speculative maximum error e that indicates a distance of an error range with respect to a GPS position. A correlation value Sp of an estimated path and a correlation value Sq of a candidate path are calculated from cumulative values of the point correlation values s. A correction that moves the estimated path is not performed if the correlation value Sq of the candidate path becomes worse than the correlation value Sp of the estimated path, and the correction of the estimated path is performed if the correlation value Sq of the candidate path does not become worse than the correlation value Sp of the estimated path. Thus, false matching can be further reduced, and the travel path can be suitably corrected.

Embodiments of the invention provide an error correction method, an error correction device and an error correction system for an inertial navigation system. The inertial navigation system revises positional parameters of a moving object on the basis of a navigation map and GPS so as to realize inertial navigation recursion. The method includes: determining the working state of the inertial navigation system according to acquired current positional parameters, map reference positional parameters, and GPS reference positional parameters of the moving object; and if it is determined that the inertial navigation system is in an abnormal working state, using the GPS reference positional parameters to update the inertial navigation initial parameters of the inertial navigation system, and resetting the inertial navigation system in an initial state. Wrong navigation system match caused by map deviation resulted from untimely navigation map road network information updating or inaccurate surveying and mapping and the like can be reduced, and the accuracy of inertial navigation is substantially improved.

The embodiment of the invention provides an image processing method and electronic equipment, which are applied to the technical field of communication and are used for solving the problem of poor image effect when images containing portraits are processed in the prior art. The method comprises the following steps: performing portrait segmentation on a target image frame meeting a preset conditionin N first image frames acquired by a camera for the same shooting object to obtain a portrait mask of the shooting object; performing image segmentation on the N first image frames according to theportrait mask to obtain N first portrait image frames and N first background image frames, and combining one first portrait image frame and one first background image frame to form a first image frame; aligning the N first portrait image frames and the N first background image frames; and carrying out image fusion on each aligned first portrait image frame and the corresponding aligned first background image frame, and outputting a multi-frame registered target image. The embodiment of the invention is applied to electronic equipment to shoot portrait scenes.

The invention relates to a microscopic sequence image splicing method and device, and the method comprises the steps: employing a template matching algorithm to match overlapped regions of adjacent images, and calculating the position difference of the adjacent images; And establishing a mathematical model according to the adjacent relation between the images, and calculating global position coordinates of the images by combining a BP algorithm to generate a global image. According to the method, mismatching caused by blank or repeated areas and the like of the images in the local splicing process can be effectively processed, mismatching of local splicing is reduced, and high robustness is achieved; And the calculation speed is high, the automation degree is high, and the splicing effectis good.

The invention discloses a heterologous image registration method based on FAST feature extraction and PIIFD feature description. Through FAST feature extraction, PIIFD feature description, nearest neighbor proportion threshold coarse matching, RANSAC mismatching elimination and least square affine transformation calculation, a spatial transformation relation between heterologous images can be obtained. The technical solution of the invention has the advantages of perfect process, high robustness and strong practicability. The method is not easily influenced by the quality of the heterologous image, can overcome the interference of the image background and other factors on feature extraction, and can well solve the problems of low precision, large error, low speed and the like in the registration of the heterologous image at present.

The invention discloses a multi-graph matching method based on low-rank tensor recovery, and the method comprises the following steps: S1, carrying out the preprocessing of each frame of image, and carrying out the feature extraction, that is, extracting the features of interest points; S2, processing interest points of all frames of images, and extracting high-order information features of the interest points according to the topological relation of the interest points; S3, based on the multi-graph cyclic consistency, establishing a multi-graph high-order feature information tensor accordingto the global corresponding relationship between the replacement matrix and the image features; and S4, solving low-rank representation of the multi-image high-order feature information tensor based on an alternating direction multiplier method (ADMM) algorithm by adopting rank constraint as a standard, so that an optimal permutation matrix, namely a matching result matrix, corresponding to a plurality of images can be effectively calculated. According to the multi-graph matching method based on low-rank tensor recovery, graph matching consistency is achieved, matching precision is improved, and the multi-graph matching method has important significance in image matching application research, target recognition and target tracking technologies.

The invention relates to a high throughput screening and authenticating method for a biological effect substance on the basis of recombination nuclear receptor protein. Selected recombination nuclearreceptor protein with a tag is selectively adsorbed to a magnetic bead carrier; according to an affinity binding principle of a nuclear receptor and a micromolecule, the recombination nuclear receptorprotein is used for specifically capturing the effect substance with binding activities from complex environment media; and an ultra-high performance liquid chromatography-quadrupole static electricfield orbitrap high-resolution mass spectrum technology is combined to carry out high throughput and non-targeted effect substance analysis. By use of the method, the specificity and the efficiency ofsubstance authentication are improved, and the method can be applied to the quick screening and the structure authentication of the effect substance combined with the nuclear receptor in the environment, and also can be used for medicine development.

The invention relates to a binocular structured light three-dimensional reconstruction method based on mask improvement, and the method employs a threshold method to make masks, the masks smaller than a threshold are regarded as shadow regions, and the masks not smaller than the threshold are regarded as effective regions. In this way, invalid area calculation can be skipped during Gray code decoding and phase shift unwrapping, the calculation amount is reduced, and therefore the three-dimensional reconstruction speed is increased. And meanwhile, a mask is manufactured by using a threshold method, and a main phase value smaller than zero is increased, so that wrong matching of left and right absolute phases in three-dimensional reconstruction is reduced, and the precision of the three-dimensional reconstruction is improved.

The invention discloses an image matching method based on SIFT (scale invariant feature transform) features and distortion-preserving mapping, which belongs to the field of computer image processing and comprises the following steps of: acquiring corresponding feature points between a source image and a target image by using an SIFT algorithm, and taking the corresponding feature points as a candidate matching set; establishing triangular mesh mapping between a source image and a target image according to corresponding feature points in the candidate matching set; adding a distortion-maintaining constraint condition in the triangular mesh mapping; and optimizing an objective function formed by corresponding feature points in the candidate matching set under a distortion-maintaining constraint condition to realize matching between the source image and the target image. According to the invention, the accuracy of image matching is improved.

The invention relates to a hierarchical positioning method used for an industrial robot and applied to an industrial environment and belongs to the object positioning field. The method includes the following steps that: S1, the image information of an object to be positioned is acquired according to a binocular vision system; S2, the image information of the object is preliminarily processed through using the MeanShift algorithm, and target image information is obtained through segmentation; S3, as for the target image information, interest point pair matching and screening are performed on atarget region through using an improved SURF algorithm; and S4, the three-dimensional coordinates of points are calculated through using a triangulation algorithm according to interest point pairs which are obtained through matching and screening, and the three-dimensional coordinates of the object can be precisely located. The invention provides the novel hierarchical target object positioning method according to the problems of long time consumed during the object grabbing operation of the industrial robot and low-precision positioning; the influence of unrelated points on an overall resultcan be avoided; overall matching accuracy is improved; and overall matching speed is increased.