506 results about "Mathematical morphology" patented technology

The invention discloses a method for locating a vehicle license plate in complex background. The method comprises the following steps: (1). Differentiating a gray scale mean of a fixed-size region and extracting a vehicle body image from a video; (2). preprocessing the vehicle body image such as gradation, gradient information enhancing, binarization and the like; (3). performing mathematical morphology operation on the processed vehicle body image, and performing primary screening by combining structural features of the vehicle license plate to obtain a plurality of candidate vehicle license plate regions; (4). converting the obtained candidate vehicle license plate regions into an HSV color spatial image; (5). adopting a BP neural network to perform color identification on the HSV color spatial candidate vehicle license plate image, and performing color edge detection by computing the mean color distance between two adjacent regions of a certain pixel; (6). judging edge color pair of an edge point, and eliminating the false vehicle license plate edge; and (7). performing final vehicle license plate location by the texture feature of the vehicle license plate. The method is suitable for complex background, and has higher location accuracy rate and higher robustness.

The invention relates to a removing method for a glasses frame in a face image. The method comprises the steps of: detecting and positioning eyes in the input face image with a pair of glasses, geometrically correcting the face image on the basis of the positioned result to form a first image; extracting an eye estimation area from the first image to obtain a second image, and conducting mathematical morphology processing on the second image so as to determine the position of a glasses frame; based on the first image, establishing a face skin color model; in the second image, interpolating each pixel point of the glasses frame the position of which is determined, correcting the interpolation result in a part of pixel points of the glasses frame according to the skin color model, and conducting de-noising processing to form a ninth image; and replacing the eye estimation area in the first image by the ninth image, synthesizing and outputting the face image the glasses frame of which is removed. According to the technical scheme, the glasses frame in the face image with a pair of glasses can be conveniently and effectively removed, and the face recognition rate can be further improved.

The invention provides a novel license plate identification method and a system and particularly relates to an integral license plate identification system which utilizes various image processing algorisms to complete license plate image pretreatment and the final license plate character identification. The identification method comprises the following steps: in license plate positioning, the filtering technique based on the math morphology is adopted; in license plate character segmentation and extraction, the methods such as perpendicular projection are adopted; lastly, the license plate characters are indentified through the template matching technique based on multiple-characteristic extraction. Compared with the prior art, the novel license plate identification method and the system can identify the license plate quickly, robustly and accurately.

The invention belongs to the image processing technique field and particularly relates to the license plate identifying technique and further to a license plate locating method under a complex background. Firstly, a license plate original image of RBG format is converted into HSI format to realize the separation between color information and brightness information; then the obtained saturation component chart and brightness component chart are binarized; then pixels of the original image are classified according to the color information of the license plate and a license plate locating template binary image can be obtained according to the classification results and a mathematical morphology computing is adopted to remove noise from the license plate locating template binary image; then a region growing method is adopted to extract each communication zone of the license plate locating template binary image and an inspection to the license plate size is also carried out, and the communication zones passing the size inspection become candidate license plate zones; the Hough transformation is adopted to correct inclined license plates and then license plate vertical texture characteristics are adopted to further inspect each candidate license plate zone so as to remove the fake candidate zones. The adoption of the invention can effectively enhance the universality and the locating accuracy of the system.

The invention discloses a navigation information acquisition method and an intelligent space system with multiple mobile robots. The intelligent space system comprises a distributed vision system and a Zigbee technology based wireless sensor network system. The method includes: carrying out image stitching based on a maximum gradient similar curve and an affine transformation model, and then carrying out image segmentation based on Otsu threshold segmentation and the mathematical morphology so that an environmental map is obtained. The mobile robots in navigation are positioned mainly by means of visual positioning, and Zigbee and DR (dead reckoning) combined positioning are used for making up visual blind zones in visual positioning supplementarily. The visual positioning refers to that positioning is realized by processing images including robot positions and direction signs mainly based on an HIS color model and the mathematical morphology. Combined positioning refers to that information fusion positioning of Zigbee and DR is achieved by the aid of a federal Kalman filter.

Based on standard of faulty work and light scattering characteristic of faulty work, new type computer aided digitized testing device suitable to large caliber, sub aperture scattering imaging is built. Features of the testing device are as following: Kohler cold light source arranged in multiple optical fiber and multiple azimuth angles; reflecting imaging of scattered light from faulty work on surface to be tested; scattered light collected by micro zooming system and imaged on CCD; Movable operating table in X, Y directions controlled by computer through multiple sub apertures scans surface to be tested in large caliber; building up mathematical model pattern recognition based on mathematical morphology, and software for calibrating size of faulty work. The invention builds objective digital evaluation system, raises working efficiency. The device is suitable for recognizing and evaluating faulty work in size larger than several micros.

The invention discloses a non-contact contact line geometrical parameter detecting method. The method includes a first step of collecting high-definition images at equal time intervals by means of collection of control signals and finishing image preprocessing by means of the median filtering technique, the image graying technique and the like, a second step of locating a laser spot center point and extracting a coordinate of the center point by means of the iterative thresholding algorithm and a method of removing isolated noise in mathematical morphology, a third step of extracting a matched target region and detecting a transverse gray singular value of the target region, a fourth step of giving the wire height and a stagger value of a contact line by means of conversion from an image coordinate system to a camera coordinate system and from the camera coordinate system to a detection vehicle coordinate system, and compensating vibration of a vehicle body, and a fifth step of giving precise detection values of the wire height and the stagger value and displaying information of a plurality of parameters in a developed graphic monitoring interface. The method effectively improves detection efficiency of geometrical parameters of a contact net, simplifies the algorithm, improves precision of fault detection, and specifically improves safe reliability of the contact net of a high-speed train.

The present invention discloses a road detection algorithm of fusion of area and edge information. The algorithm comprises: collecting road image, and employing self-adaption median filtering and noise reduction to obtain enhanced images; selecting the R component of the enhanced images in the RGB color space, employing a maximum between-cluster variance (Otsu) method to realize the road area segment of the images to obtain a binary segmentation image, and optimizing the binary segmentation image through adoption of the serial mathematical morphology; detecting the binary segmentation image through adoption of the optimal edge detection Canny detector to detect the binary segmentation image and obtain the road edge detection information, and employing the Otsu method to calculate the dual threshold in the optimal edge detection Canny detector; and employing the binary segmentation image and the road edge detection information to obtain a road border line in the image. The road detection algorithm of fusion of area and edge information employs the image processing technology to detect and extract the road information for the autonomous navigation system of a cable tunnel robot; and moreover, the road detection algorithm of fusion of area and edge information is a self-adaption road border line segmentation fusion method and can obtain a smooth and accurate road edge.

The invention discloses a method and system for nighttime vehicle detection on the basis of vehicle light identification. The method includes: inhibiting a camera through strong light, obtaining traffic video data, and determining analyzed area-of-interest; performing filtering processing on images on the basis of mathematical morphology, and obtaining communication area information on the basis; then building vehicle light screening and matching rules, and screening and matching vehicle lights; performing position prediction on a target, setting vehicle confidence degree, updating stored vehicle position information, and achieving tracking of vehicles; and finally performing calculation and statistics on parameters of traffic flow, vehicle speed and the like of the vehicles left the area-of-interest. The method and system for the nighttime vehicle detection on the basis of the vehicle light identification adopts a large number of vehicle light image data statistic priori knowledge to identify vehicle light pairs and selects the vehicle light pair with highest matching degree from a connection area. The vehicle identification method enables the vehicle lights to be separated from nighttime scenes effectively and has good scene adaptability.

The invention provides a machine vision method for identifying the dial plate information of a multi-pointer instrument. The machine vision method mainly comprises the following steps of: acquiring an image containing information of the multi-pointer instrument, preprocessing the acquired image, positioning and marking a dial plate of the multi-pointer instrument in the image, segmenting the dialplate of the instrument in the image of the dial plate of the multi-pointer instrument, carrying out mathematical morphological filtering, extracting pointer characteristics in the dial plate by means of a human-eye-vision based jumping searching pointer and a human-eye-vision staring characteristic, calculating readings of the pointers of the instrument and the like. The machine vision method provided by the invention can be used for automatically carrying out vision identification on the information of the dial plate of the multi-pointer instrument in one image, and has the advantages of high degree of automation, accuracy in measurement, high efficiency and the like; and compared with the detection method in which one image and one instrument are provided in single-pointer instrument detection at present, the invention improves the utilization rate of the pointer instrument detection equipment and reduces the instrument detection cost.

The invention discloses a building area extraction method based on LiDAR data, which belongs to the field of extraction of building through the LiDAR data and includes early-age processing of the LiDAR data, resampling of original LiDAR data, reverse iteration mathematical morphological filtering and separation of building and dense trees. The building area extraction method directly processes three-dimensional (3D) point cloud data instead of converting the point cloud into depth images, avoids information loss and added calculation quantity in the conversion process, simultaneously utilizesdifferent windows to conduct mathematical morphological filtering operation through reverse gradual iteration, basically removes effects of terrain fluctuation on the building in the mathematical morphological filtering and obtains high extraction accuracy. The building area extraction method can quickly and accurately extract LiDAR points belonging to the building area from a large range of LiDAR points and is capable of providing reliable data support for urban three-dimensional modeling.

Signal “peak” and “valley” removal properties of mathematical morphology operators are exploited in a method and apparatus for detecting, removing, or improving fidelity of pacemaker signal components of the Electrocardiogram (ECG) at sampling rates well below the pacemaker signal Nyquist rate. The method works for the wide variability in pacemaker signal amplitude, width, firing frequency, and other characteristics encountered in practice, and it works for malfunctioning pacemakers that may fire at any point relative to the QRS complex of the ECG signal. Filtering operations require minimal digital storage and are computationally inexpensive (no multiplications), mainly involving “maximum” and “minimum” type signal detections over a finite time history of the input signal. This implies that the method can be inexpensively implemented on small instruments in either hardware or software. The method may also be used to estimate the height and polarity of the pacemaker voltage spike, even though the base sampling period may be longer than the width of the spike.

The invention discloses a vision-based quick LED module brightness uniformity detection method. The method comprises the following steps: S1, a light emitting image of the LED module is acquired, gray processing is carried out on the image, the image is filtered after gray processing, the image after being filtered is subjected to morphological processing, and the image after morphological processing is subjected to histogram equalization processing; S2, each LED lamp point is positioned through block positioning algorithm, the LED lamp point is subjected to gray processing, the LED lamp point after gray processing is subjected to mathematical morphological open operation, and the brightness value of the LED lamp point and an average value are acquired; and S3, after each LED lamp point in the LED module is traversed, the brightness value of each LED lamp point is displayed, and the final processing result is displayed. The method plays an important role in yield detection during the LED lamp production process.

The invention discloses a rapid lens flaw detection method and an apparatus thereof. The method comprises the following steps: equally dividing the surface of a lens to be detected into n<2> parts, respectively obtaining the images of all the parts, and obtaining n<2> images within a preset time which is used for controlling the image obtaining rate; 2, carrying out median filtering, black hat transformation and binary processing on the n<2> images respectively, and partitioning off defective images; 3, carrying out opening operation on the defective images to filter out non-flaw interference parts; and 4, drawing contours of the defective images subjected to non-defective interference filter, there are flaws on the surface of the lens to be detected if there is a counter, and no flaws are on the surface of the lens to be detected if there is no counter. The rapid lens flaw detection method provided in the invention allows scratches on the lens to be clearly detected through a surface defect detection method based on the mathematical morphology under the high speed motion of a platform, and defect types to be effectively discriminated, and has the characteristics of fast speed, objective detection result and the like.

The invention discloses an electrocardiosignal denoising method based on morphology and EMD (empirical mode decomposition) wavelet threshold value. The electrocardiosignal denoising method comprises the following steps: firstly, the mathematical morphology filter method is adopted to perform primary filtering on an electrocardiosignal to filter baseline drift in the electrocardiosignal, then, the EMD method is adopted to decompose the obtained electrocardiosignal and obtain a noise dominant component and a useful signal dominant component through classification, the threshold value denoising method is adopted to perform secondary filtering on the electrocardiosignal, and finally, the EMD method is adopted to reconstruct the electrocardiosignal to obtain the noise-filtered electrocardiosignal. The invention has the following remarkable effects: the method is simple and easy to implement; the morphological filter method, the EMD method and the threshold de-noising method are combined in an organic manner; compared with the traditional denoising method, the electrocardiosignal denoising method can comprehensively and effectively remove electrocardiosignal noise, keeps the useful information in the electrocardiosignal, and provides more valuable reference for change of cardiac functions and diagnosis of cardiac diseases.

The invention discloses a convolutional neural network-based character recognition method, and mainly aims at solving the problem that existing manual engineering drawing data input is time-consuming,labor-consuming and high in error rate. The method comprises the following main steps of: 1) converting grayscale images into binary images by adoption of Otsu method-based optimal global threshold value processing; 2) rotating to correct inclined images; 3) extracting a table box by adoption of a mathematic morphological algorithm; 4) obtaining an area where characters are located; 5) rotating inclined character strings for a corresponding angle to a horizontal direction; 6) searching circles in the images by adoption of a circular Hough transformation algorithm, and extracting welding beadnumbers in an engineering drawing; 7) carrying out image-text recognition on the engineering drawing by adoption of a convolutional neural network method; 8) correcting simple errors through an encoding rule and automatically correcting recognition errors; and 9) outputting and storing the drawing data. In image-text recognition of engineering drawings, the method has high correctness and timeliness, and is capable of realizing efficient recording and management of engineering drawing data.

The invention discloses an intelligent power grid high voltage insulation damage monitoring system and method based on ultraviolet imagery. Image acquisition is conducted on a high voltage insulation device through an acquisition module via an ultraviolet imager, and an RGB color image of corona discharge and arc discharge on the surface of the device is obtained; the RGB color image is converted into a grayscale image through a pre-processing module; a threshold segmentation method is adopted by a threshold segmentation sub-module, a threshold is set to convert the grayscale image into a binary image, and a discharging area in the image is divided; mathematical morphology filtering processing is conducted on the binary image through a morphological filter sub-module, and scatters are removed; an edge detecting sub-module conducts edge detection on the filtered image, edge pixel point position information is stored, colors of edge pixel points in the original image are changed, and a dividing effect image is obtained. The intelligent power grid high voltage insulation damage monitoring method based on ultraviolet imagery has the advantages of being efficient, simple, visual, imagery, safe and convenient, and not influencing normal operation of the device. With combination of the image division technology, the discharging area is divided, and accurate detection of high voltage insulation damage is achieved.

The invention provides an automatic division method for pulmonary parenchyma of a CT image. According to the automatic division method, the CT is divided through carrying out a random migration algorithm for two times to obtain the accurate pulmonary parenchyma; in the first time, the random migration algorithm is used for dividing to obtain a similar pulmonary parenchyma mask; and in the second time, the random migration algorithm is used for repairing defects of the periphery of a lung and dividing to obtain an accurate pulmonary parenchyma result. Seed points, which are set by adopting the random migration algorithm to divide, are rapidly and automatically obtained through methods including an Otsu threshold value, mathematical morphology and the like; and manual calibration is not needed so that the working amount and operation time of doctors are greatly reduced. According to the automatic division method provided by the invention, a process of 'selecting the seed points for two times and dividing for two times' is provided and is an automatic dividing process from a coarse size to a fine size; and finally, the dependence on the selection of the initial seed points by a dividing result is reduced, so that the accuracy, integrity, instantaneity and robustness of the dividing result are ensured. The automatic division method provided by the invention is funded by Natural Science Foundation of China (NO: 61375075).

The invention discloses a method for automatically identifying the action of writing on blackboard of a teacher in class video recording, which belongs to image recognition technology in a long-distance educational system, which comprises the following processing steps: a, capturing a frame of image by using a frame grabber, subtracting a previous frame of image from the current image to obtain a binary frame difference image and analyzing the binary frame difference image; b, subjecting the binary frame difference image to closed operation in mathematical morphology, marking communicated regions and removing obvious noise regions in the binary frame difference image by setting a threshold; c, acquiring a target rectangular frame and calculating the area of the rectangle, determining the movement of a human body when the area of the rectangle meets a certain condition and detecting and identifying the behavior of the writing of the teacher on the blackboard; d, when the more than 5 successive frames of images of behavior of writing on the blackboard are detected, stretching a camera; and e, if no 50 successive frames of images of the behavior of writing on the blackboard are detected, restoring the stretching magnification to an initial magnification.

The invention provides a three-dimensional multimode medical image automatic registration method based on mutual information and image segmentation. The method comprises the following steps that stepS1, preprocessing is performed by using a threshold method and a mathematical morphology method; step S2, segmentation is performed by using the k-means method; step S3, the optimal registration parameter based on the mutual information is obtained through iteration by using the optimization algorithm; step S4, an original reference image and a floating image are superposed; step S5, the grayscalehistogram of the reference image A through image segmentation preprocessing is calculated, and the pixels having the same grayscale value are arranged in one group; step S6, the registration parameter is initialized, and the initial values of six parameters are set as zero; and step S7, linear interpolation is performed on the floating image B by using the registration parameter to generate the changed floating image, and the zero value is assigned to the pixel points mapped to the floating image outside the reference image in the iteration process. The method is high in accuracy, high in robustness and high in performance.

The invention relates to the technical field of biometric identification and discloses an iris positioning method and an iris identification system. The method and the system are based on mathematical morphology, Radon transform, canny operator and hough transform. The algorithm comprises the positioning of an inner edge and an outer edge, wherein in the positioning of the inner edge, based on a gray projection method of line detection, a binarized image is processed by using the principle of the mathematical morphology so as to remove the interference of noise; and in the positioning of the outer edge, an iris image is processed by using the principle of the mathematical morphology at first, then an upper eyelid and a lower eyelid are detected by the Radon transform to remove the interference, and finally, an edge is extracted by a canny algorithm and the external edge of the iris is detected by the hough transform. Therefore, large area blind searching can be avoided so as to improve positioning speed.

The invention particularly relates to a method for detecting the camber of plates based on machine vision, belonging to the automatic control field of steel rolling processes. The method comprises thefollowing steps: obtaining plate images through plane array CCD cameras installed right above a roller way outside of roller beds in front of and behind a rolling mill, and correcting the distortionof the images through camera calibration; performing Gaussian low-pass frequency domain filtering on the corrected images, and performing image sharpening differential operation on the filtered imagesto enhance the image boundary. Mathematical morphology transformation and threshold binarization are carried out on the images with enhanced boundaries to segment and completely extract the plate image. The pixel coordinates of the edge contour points are extracted by using the Canny edge detection operator after threshold segmentation and the sub-pixel coordinates of the plate edges are obtained by using the improved Zernike orthogonal moment sub-pixel edge detection algorithm, and the camber value is obtained by fitting.

The invention discloses a photoelectric volume wave-based breathing rate monitoring method. The method is characterized by comprising the following steps: first, acquiring a photoelectric volume wave digital signal of the change of the blood volume on a capillary bed on the surface of skin at a position to be monitored, wherein the signal is an original photoelectric volume wave signal which has not been filtered, contains breathing information and has a sampling frequency over 500 Hz; then, preprocessing an acquired photoelectric volume wave by using a mathematical morphology method; next, detecting a peak point and a valley point of each beat of a volume pulse wave by adopting a characteristic point positioning method by the preprocessed signal; later on, calculating time interval variable of each beat and the difference variable between the peak point and the valley point of each beat; and finally, detecting obtained the breathing rhythms of the time interval variable and an amplitude variable, and calculating a breathing rate.

The invention provides a license plate recognition and positioning method based on a deep neural network, and mainly solves the problem of inaccurate license plate recognition and positioning in a complex scene in an existing algorithm. Firstly, a license plate data set meeting specific requirements of license plate detection is established; an anchor frame is generated by using a K-means clustering algorithm, a license plate detection deep convolutional neural network structure is established by combining machine learning and introducing an attention mechanism, a network model is trained by using an established license plate data set, and an Adam algorithm is used as an optimization algorithm in the training process; testing the model by adopting the detection accuracy when the cross-parallel ratio IOU is equal to 0.8 as a measurement index of algorithm performance and adopting a HyperLPR algorithm and a mathematical morphology method as a comparison algorithm. Compared with the priorart, the license plate recognition and positioning method based on the deep neural network has the advantages that a channel attention mechanism is added, so that the detection accuracy is higher, the speed is higher, and the robustness to the environment is very high.

The invention discloses a video image skin-tone detection method based on multi-color gamut selective morphology processing and is characterized in that: the mathematical morphology processing technology is introduced, video image data is input and passes through two parallel skin-tone detection modules based on RGB color gamut and YUV color gamut, pixels in the image are analyzed, selected and roughly judged whether the pixels are skin-tone pixels of the human body respectively in an RGB color space and a YUV color space under the control of a register operation management module MCU, the detection result is made a comprehensive judgment and analysis by a comprehensive judgment module for skin-tone detection, then a binary image obtained through the judgment is made a selective morphology filtering based on the content information in a selective morphology filtering module, finally the detected skin-tone is protected or processed under the control of the register operation management module MCU according to the need of the user to obtain the image which meets the visual habit of human eyes.

The invention provides a method of iris location based on multi-resolution analysis, comprising: locating out rough circle center (x, y) of a pupil via operations, such as binary, mathematical morphology, gray scale, searching for smallest pieces and so on; then searching for a coordinate of a maximum value of the gray scale and gradient in certain rage on a plurality of lines near (x<0>, y) as the coordinate of pupil boundary point, and locating out exact circle center and radius of the pupil via curve fitting; according to precise components of each line of gray serial near the rough circle center (x, y) in the resolution of 2<-5> and 2<-6>, searching for the coordinate of local minimum value point as an external border boundary point of the iris and processing curve fitting to gain exact circle center and radius of the external border of the iris. Adopting the method based on multi-resolution analysis which is provided by the invention can effectively process a positioning operation of iris image, and avoid a problem that traditional arithmetic is sensitive for occlusion of eyelid and eyelash.

The invention discloses a detection method and a detection system for video image text and relates to the technical field of image text detection. The method comprises the following steps of coarse positioning of video image text areas and verification of video image text areas. The specific steps are as follows: (A) coarse positioning of video image text area: the gaussian pyramid multi-scale image of a video image is calculated, the image text area is segmented by threshold segmentation, text areas are combined by the inflation and corrosion technology in mathematical morphology, and candidate text areas are analyzed and positioned by connected domain analysis, text line segmentation and elimination rules; and (B) verification of video image text areas: an adaboost classifier is trained by adopting harr characteristics, and the adaboost classifier is utilized to detect whether text exists in the candidate text areas and the candidate text areas the text of which is not detected are eliminated. The method and the system can improve the accuracy and the recall ratio of video image text detection.

The invention discloses a method and device for locating a fault on a direct current grounding electrode circuit. The method for locating the fault comprises the steps that a plurality of voltage sampling values and a plurality of current sampling values , corresponding to a plurality of preset sampling times, of the direct current grounding electrode circuit are obtained from the measuring points of the direct current grounding electrode circuit; a forward voltage traveling wave function and a backward voltage traveling wave function are obtained in a calculating mode; mathematical morphology gradient transformation is carried out to obtain a forward traveling wave gradient function and a backward traveling wave gradient function; the reflected wave of a fault point is confirmed through the forward traveling wave gradient function and the backward traveling wave gradient function; the time corresponding to the reflected wave of the fault point is used as fault time, and a fault distance is confirmed according to the fault time. The upper edge or the lower edge or the signal saltation point of transient signals can be extracted well, and therefore the reflected wave of the fault point can be distinguished accurately. Therefore, the accurate fault time can be obtained, and the accuracy of the fault distance can be greatly improved.

The invention relates to an image fusion technique, in particular to an infrared intensity image and infrared polarization image enhancement and fusion method, which can solve the inherent problems of edge distortion and low contrast in conventional infrared fused images. The method is carried out according to the following steps that support value transform is respectively carried out on an infrared intensity image and an infrared polarization image, so that low-frequency images and support value sequence images are obtained; the top-hat transform of mathematical morphology is used for respectively extracting bright information and dark information; the bright information images and the dark information images are enhanced; the two last-layer low-frequency average images and the two enhanced images are fused, so that a low-frequency enhanced fused image is obtained; inverse support value transform is carried out on the low-frequency enhanced fused image and a support value sequence image fused by a maximum method, so that a final fused image is obtained, the contrast between the targets and the background in the fused image is enhanced, and edge distortion is little.

The invention provides a gait recognition method based on a partitioned matrix. Firstly, extracting single-frame images from a video, then carrying out grey scale transformation on the single-frame images, using the background subtraction method to extract person body targets, using mathematical morphology to fill the holes of binary images, and extracting profiles of the person by means of single connection analysis so that the person bodies are positioned in the middle and are uniformly in the size of 64 * 64 pixels; observing the periodic change of the gait according to elliptical short axis and eccentricity fitted in image regions after the standard centralization of each frame image in a gait video sequence; using a gait energy diagram to extract the integral characteristic of the gait in the one period, dividing GEI into sub-blocks by means of the partitioned matrix, eliminating the sub-blocks which are useless to classification in a self-adapting manner, and adopting the method, which combines the two-dimensional principal component analysis of a sub-block mode with the two-dimensional linear discriminant analysis, to further extract local characteristics; and integrating the characteristics of each effective sub-block into a whole during the classification recognition, and adopting a nearest neighbor classifier to perform identification judgment. The method is effective for the recognition of the gait of knapsack change.