36results about How to "Achieve coarse positioning" patented technology

The invention discloses a method based on monocular vision for detecting and roughly positioning the edge of a road, and relates to the field of machine vision and intelligent control. Aiming at a continuous road with different edge characteristics, two road edge detection methods are supplied and suitable for semistructured and nonstructured roads and can be applied to vision navigation and intelligent control over a robot. The invention provides a method for detecting the edge of the road based on colors and a method for detecting the edge of the road based on threshold value partitioning. On the basis of the obtained edge of the road, an image is subjected to inverted perspective projection transformation, so that a front view is transformed into a top view; and according to a linear corresponding relation between a pixel and an actual distance in the image, a perpendicular distance from the current position of the robot to the edge of the road and a course angle of the robot can be calculated. The method is easy to implement, high in anti-interference performance, high in instantaneity and suitable for the semistructured and nonstructured roads.

The invention provides a remote sensing image registration method of a multi-source sensor, relating to an image processing technology. The remote sensing image registration method comprises the following steps of: respectively carrying out scale-invariant feature transform (SIFT) on a reference image and a registration image, extracting feature points, calculating the nearest Euclidean distances and the nearer Euclidean distances of the feature points in the image to be registered and the reference image, and screening an optimal matching point pair according to a ratio; rejecting error registration points through a random consistency sampling algorithm, and screening an original registration point pair; calculating distribution quality parameters of feature point pairs and selecting effective control point parts with uniform distribution according to a feature point weight coefficient; searching an optimal registration point in control points of the image to be registered according to a mutual information assimilation judging criteria, thus obtaining an optimal registration point pair of the control points; and acquiring a geometric deformation parameter of the image to be registered by polynomial parameter transformation, thus realizing the accurate registration of the image to be registered and the reference image. The remote sensing image registration method provided by the invention has the advantages of high calculation speed and high registration precision, and can meet the registration requirements of a multi-sensor, multi-temporal and multi-view remote sensing image.

The invention discloses a face alignment method based on cascade position regression of random forests. The face alignment method is characterized in that 1, a normalized face image is acquired; 2, the average shape of the faces can be calculated; 3, a candidate feature point of a face alignment frame is generated; 4, a face shape index grey level is generated; 5, a face shape index feature X is generated; 6, a face alignment frame is built; 7, the face shape is initialized, and after the continuous iteration, the final estimated face shape can be output. The good robustness can be kept under the conditions of the changing illumination, the changing expressions, and the shielding, and the precision can be improved, and the failure rate can be reduced.

The invention discloses a full-automatic stiffening plate false sticking machine, and belongs to the technical field of flexible printed circuit processing devices. The full-automatic stiffening plate false sticking machine comprises a machine frame, an FPC positioning heating platform driven by an X-direction movement mechanism is installed on the machine frame in a sliding mode, and a feeding mechanism, a servo blanking mechanism and a sticking mounting mechanism are arranged on the machine frame. The servo blanking mechanism comprises a material overturning mechanism. The sticking mounting mechanism comprises a supporting beam, a Z-direction and angle theta rotating movement mechanism, an upper vision camera and a lower vision camera, wherein a Y-direction movement mechanism is arranged on the supporting beam, the Z-direction and angle theta rotating movement mechanism is installed on the Y-direction movement mechanism, a stiffening plate suction nozzle is installed on the Z-direction and angle theta rotating movement mechanism, the upper vision camera is installed on the Y-direction movement mechanism, and the lower vision camera is installed on the machine frame. According to the full-automatic stiffening plate false sticking machine, the stiffening plate can be accurately stuck and mounted on an FPC, production efficiency and sticking mounting precision are greatly improved, the production cost is reduced, and device stability is high and product quality is reliable during actual production.

The invention provides a medical image abnormal region segmentation method and device. The method comprises the steps of obtaining a medical image; dividing the medical image into a plurality of to-be-classified image blocks, and utilizing a first machine learning model to classify each to-be-classified image block, the classification result being a normal image block or an abnormal image block; respectively taking to-be-segmented image blocks from the medical image based on the abnormal image blocks, and respectively performing semantic segmentation on each to-be-segmented image block by using a second machine learning model, the segmentation result being an abnormal pixel point; and marking the abnormal pixel points in the medical image.

The invention relates to an internet photo geospatial positioning method based on street views, comprising the steps of preprocessing a street photo gallery, extracting and describing features, and establishing a feature index; querying a nearest-neighbor feature of each feature of a photo to be queried, performing voting, trimming and smoothening voting results to obtain a most similar photo, using the most similar photo as a center of circle to delimit a buffer area according to a distance between two known street view points; calculating similarity of street photos in the buffer area to the photo to be queried so as to screen out photos of high similarity to form a similar photo set; subjecting both the similar photo set and the photo to be queried to feature extraction and matching, using an SfM (structure from motion) algorithm to register the photos, and generating a relative positional relationship between a sparse point cloud and a camera; calculating exterior orientation elements of the unknown photo to be queried according to the known street view point coordinates so as to finish positioning and orientating. Practice proves that the method provided herein enables electronic photos of any source on the internet to be positioned accurately.

The invention relates to a local test vector generating and optimizing method based on an annular oscillator network, belongs to the technical field of trustiness detection of an integrated circuit, and aims at enhancing the circuit activity in the local area, further improving the resistance to hardware Trojans, improving the influencing signal to noise ratio of the hardware Trojans to certain extent under influence of technical noises, and improving the hardware Trojan testing efficiency greatly. According to the method, an original layout is not changed, active sensing units of a local area is inserted into a spare space in the layout, a test circuit is generated, and a local test vector is generated and optimized for the test circuit. The method comprises the following steps that 1) the test circuit is generated; 2) a local test vector generating and optimizing system is initialized; 3) a test vector set is generated randomly, and test vectors are applied; 4) data is collected; 5) determination is completed by test; and 5) the test vectors are optimized. The method is mainly applied to trustiness detection occasions of the integrated circuits.

The invention discloses a method for achieving face key point detection based on cascade MobileNet-V2 , which comprises the following steps: S1, obtaining a picture data set; S2, establishing a MobileNetV2 cascading network; S3, preliminarily determining the face key point, S4, cutting a face area, S5, determining the face area again, and S6, obtaining an accurate face key point. The invention adopts the MobileNet-V2 neural network with high speed and high precision, and improves it, and uses the two-stage neural network cascading method to improve the accuracy of face detection.; compared with an existing face key point detection model, the method has the advantages of being higher in speed and precision; Through a cascade mode, positioning of fine promotion and refinement is achieved, and cascade network training comprises the steps of firstly training a first-stage cascade network, achieving rough positioning of face key points, then cutting a face area, and then achieving accuratepositioning of the face key points through a second-stage cascade network.

The invention discloses a reinforcement plate disc type full-automatic false mounting machine, and pertains to the technical field of flexible circuit board machining equipment. The reinforcement plate disc type full-automatic false mounting machine comprises a rack. The rack is provided with an X and Y direction movement mechanism. The X and Y direction movement mechanism is provided with a FPC locating heating platform. The rack is provided with a material supply mechanism and a mounting mechanism. The mounting mechanism comprises: a Z direction rotary mechanism which is driven by a first power device; at least one Z direction and theta-angle rotary movement mechanism which is installed on the Z direction rotary mechanism, wherein the direction and theta-angle rotary movement mechanism is provided with reinforcement plate suction nozzles; an upper vision camera which is installed on the Z direction rotary mechanism; and a lower vision camera which is installed on the rack, wherein the lower vision camera is located in the rotary trajectory range of the reinforcement plate suction nozzles. The reinforcement plate disc type full-automatic false mounting machine of the invention can be used to replace the traditional manual mounting method, so the production efficiency can be significantly improved, and the production cost can be reduced; and through the rotary mechanisms, monomer reinforcement plate can be transported continuously, so the production efficiency is higher.

The invention discloses a method and device for detecting a liquid level of a PET bottle based on machine vision. The device comprises a camera, two area light sources, five reflection mirrors, an industrial control computer and a PLC control unit, wherein multi-angle images of the PET bottle are obtained through matched use of the five mirrors. The device has the advantages of a simple structure, convenient debugging, low cost and a rapid speed. The detection method is characterized in that a simple and efficient projection gradient method is used to obtain liquid level information, so that the liquid level of the PET bottle can be detected. The detection method disclosed by the invention has the advantages that accuracy is high; computation is quick and the liquid level of the PET bottle can be detected rapidly and accurately.

The invention discloses an image segmentation method, a training method of an image segmentation network, electronic equipment and a storage medium. The image segmentation method comprises the following steps: sending an input image to an image segmentation network; obtaining a first foreground target box in the input image; acquiring a first region of interest and a first region of interest feature map in the input image based on the first foreground target box; dividing the first region of interest into a plurality of grids, predicting a feature corresponding to each grid in the first region of interest based on a first region of interest feature map, and obtaining a semantic feature of each pixel point in the first region of interest; and obtaining an instance segmentation result based on the feature of each grid in the first region of interest, the information of the first foreground target box and the semantic feature of each pixel point in the first region of interest. In this way, the memory occupied by image segmentation can be reduced, and the consumed time can be shortened.

According to the rapid global positioning method for the bounded area of the laser radar, coarse positioning of the map area where the robot is located is achieved by constructing and matching the first-level feature vectors of the global map and the local map; further according to a map where the robot is located, a secondary feature vector of the global map and a secondary feature vector of thelocal map are constructed; the precise positioning of the robot position and azimuth angle is realized, influence of illumination on positioning precision is effectively avoided, the convex polygon geometric map is utilized as map representation, the laser radar positioning algorithm based on map feature vector matching is acquired, algorithm complexity is reduced, and rapid global positioning ofthe robot in a bounded area is realized.

The invention provides a lung disease recognition method and device based on CT data, a computer readable medium and electronic equipment, and relates to the technical field of image processing. The method comprises the following steps: acquiring a 3D tensor corresponding to lung CT data to be identified, and partitioning the 3D tensor to obtain k partitioned tensors; performing lung disease classification identification on the k block tensors to obtain k block classification results; and outputting an identification result corresponding to the to-be-identified lung CT data according to the k block classification results, and determining a focus block tensor corresponding to the to-be-identified lung CT data in the k block tensors when the identification result is an abnormal result. According to the method, the 3D tensor is partitioned, so that sampling of the 3D tensor can be avoided, and then missing of key focus information or introduction of repeated invalid noise is avoided; and meanwhile, the focus block tensor can be directly determined in the k block tensors through the k block classification result, so that the coarse positioning of the focus is realized based on the focus block tensor.

The invention discloses a multi-direction drilling technology based building punching device. The multi-direction drilling technology based building punching device comprises a machine table, a drilling mechanism, a simple type machine base and a clamping mechanism, wherein the machine table comprises a processing table and brackets which are fixedly connected to front and rear sides of the bottomof the processing table; a rotary slot is formed in the middle of the right half part inside the processing table; left sides of front and rear sides of the processing table are fixedly connected toslide rails; the drilling mechanism comprises a turnover board which is rotatably connected to the inner part of the rotary slot. The invention relates to the technical field of mechanical processingequipment. According to the multi-direction drilling technology based building punching device disclosed by the invention, the drilling mechanism is rotatably arranged on the processing table, so thatparts can be processed onto the processing table to process while some movable parts are normally drilled; while workpieces which need fixing ground are drilled, the whole drilling mechanism can be turned over for directly drilling objects on ground; and an insertion sleeve is matched with an insertion board insertion sleeve, so that the drilling mechanism is easily fixed, operations are convenient, and range of application is wide.

The invention relates to the technical field of sensors, in particular to a sunlight direction detecting sensor within a large-angle range. The sunlight direction detecting sensor comprises a base and a shading barrel arranged on the base, a light through hole is formed in the top of the shading barrel, a four-quadrant sensor is arranged at the bottom of the shading barrel, a lens is arranged between the light through hole and the four-quadrant sensor, and at least eight phototubes are uniformly arranged in the circumferential direction of the shading barrel. The sensor is large in tracking range and capable of achieving blind-zone-free continuous tracking within the +/-80 degrees with the axis of the four-quadrant sensor as the center and has the advantages of being simple in structure, high in detection accuracy and convenient to use.

The invention discloses a flywheel housing coaxiality measuring device which comprises a crankshaft rotating assembly installed on a flywheel housing starting motor installation base and a measuring assembly fixed to a flywheel housing. The crankshaft rotating assembly comprises a driving motor, a connecting base and an output gear, and the driving motor is driven to drive the output gear to rotate so that a crankshaft can rotate synchronously. The measuring assembly comprises a mounting substrate, a sliding block, a sensor frame, a sensor mounting base and a sensor. The measuring device has the advantages that labor intensity is reduced, uniform rotation of the flywheel housing is guaranteed, measuring precision of the measuring assembly is improved, the measuring precision of the measuring device is greatly improved through automatic reading and measurement of the sensor compared with manual observation of a dial indicator, and in addition, due to the fact that the sensor frame can be movably fixed to the sliding block, the measuring assembly can measure the coaxiality of different types of flywheel housings, and the applicability is high.