A Camera Offset Detection Method Based on Feature Point Matching

Abstract

This invention belongs to the field of computer artificial intelligence vision applications, specifically disclosing a camera offset detection method based on feature point matching, including an image preprocessing module, a feature point extraction module, a feature point matching module, and a threshold judgment module; the image preprocessing module, feature point extraction module, feature point matching module, and threshold judgment module are connected sequentially. The beneficial effect of this feature point matching-based camera offset detection method is that: by extracting feature points from the current monitoring screen of the camera and feature points from multiple pre-stored standard monitoring screens at various time points, the method uses distance calculation to match the feature point descriptors of the current monitoring screen with the feature point descriptors of the pre-stored standard monitoring screens at multiple time points, and then combines this with a set threshold for the number of matching points to determine whether the camera has shifted. This method can be used in the security field to ensure the normal operation of equipment.

Description

Technical Field

[0001] This invention belongs to the field of computer artificial intelligence vision applications, specifically relating to a camera offset detection method based on feature point matching. Background Technology

[0002] In existing image processing techniques, methods for detecting positional offset between a target image and a standard image typically compare the entire content of both images or segment the image into blocks for comparison to determine if the image position has shifted. This method matches or transforms almost every pixel in the entire image, which is highly complex and lacks sufficient feature representativeness. Moreover, surveillance camera footage is real-time, operating 24 / 7, and is affected by factors such as lighting, monitoring environment, and weather. The content of the surveillance footage affected by these factors is completely different, and changes in the surveillance footage can severely impact the accuracy of image matching and increase the difficulty of parameter adjustment during image matching.

[0003] Therefore, when detecting the offset of surveillance cameras, a method that is low in complexity but can improve detection accuracy is needed.

[0004] To address the aforementioned problems, this invention provides a camera offset detection method based on feature point matching. Summary of the Invention

[0005] Purpose of the invention: The purpose of this invention is to provide a camera offset detection method based on feature point matching, which solves the problems existing in the methods in the background art, and realizes camera offset detection with a simpler and more reasonable processing flow and lightweight features.

[0006] Technical solution: The present invention provides a camera offset detection method based on feature point matching, including an image preprocessing module, a feature point extraction module, a feature point matching module, and a threshold judgment module; the image preprocessing module, the feature point extraction module, the feature point matching module, and the threshold judgment module are connected in sequence.

[0007] In this technical solution, the image preprocessing module performs data preprocessing on the acquired camera monitoring image, converting the image into a grayscale image; the feature point extraction module extracts SIFT features from the preprocessed monitoring image to obtain image feature points and feature point descriptors; the feature matching module, based on the extracted camera monitoring image feature points and feature point descriptors, uses a distance calculation method to match the current monitoring image feature point descriptors with the feature point descriptors of pre-stored standard monitoring images at multiple time points; and the threshold judgment module compares the threshold of the set matching pair with the matching pair value of the current monitoring image feature point descriptors and the feature point descriptors of the pre-stored images to determine whether the camera has shifted.

[0008] In this technical solution, the current monitoring screen image is acquired and grayscale processed to convert the RGB three-channel image into a single-channel grayscale image.

[0009] The camera offset detection method based on feature point matching in this technical solution includes the following steps: Step (1) The image preprocessing module acquires the current monitoring screen image and preprocesses the image. Step (2) The feature point extraction module extracts SIFT features from the preprocessed monitoring screen image to obtain image feature points and feature point descriptors. Step (3) The feature matching module, based on the extracted camera monitoring screen image feature points and feature point descriptors, uses a distance calculation method to match the current monitoring screen feature point descriptors with the feature point descriptors of multiple pre-stored standard monitoring screens at various time points. Step (4) The threshold judgment module compares the threshold of the set matching pair with the matching pair value of the current monitoring screen feature point descriptors and the feature point descriptors of the pre-stored screens to determine whether the camera has shifted.

[0010] In this technical solution, the feature representation of the pre-stored standard monitoring screen is a set of SIFT feature point information extracted from camera monitoring screens at multiple time points collected in advance.

[0011] In this technical solution, the SIFT feature is a local feature of the image. It is invariant to rotation, scaling, and brightness changes, and maintains a certain degree of stability against viewpoint changes, affine transformations, and noise. It also maintains good matching ability against factors such as object motion, occlusion, and noise. This feature is based on the idea of ​​image feature scale selection, establishing a multi-scale space of the image. The same feature point is detected at different scales, and the scale at which the feature point is located is determined simultaneously to achieve scale resistance. Then, some low-contrast points and edge response points are removed, and rotation-invariant feature descriptors are extracted to achieve resistance to affine transformations.

[0012] The SIFT feature extraction steps of this technical solution are as follows: 1) Constructing a scale space, detecting extreme points, and obtaining scale invariance; 2) Filtering feature points and performing precise positioning; 3) Specifying orientation parameters for each feature point; 4) Generating descriptors for feature points.

[0013] In this technical solution, the feature point matching is performed by calculating the distance between feature point descriptors. The matched feature pairs may contain some incorrect matching pairs. Some filtering conditions are set manually to remove abnormal matching pairs. For example, matching pairs in which the ratio of the nearest distance between feature point descriptors to the second nearest distance between feature point descriptors is greater than a certain value can be filtered out. This filtering process can greatly ensure the accuracy of the matching pairs.

[0014] In this technical solution, the distance between feature point descriptors is calculated based on the Euclidean distance between the descriptors of two feature points. Taking two feature points A and B as an example, assuming that the descriptor of feature point A is (x1, y1) and the descriptor of feature point B is (x2, y2), the formula for calculating the distance dist between these two feature point descriptors is as follows:

[0015]

[0016] In this technical solution, the threshold judgment module compares the threshold of the set matching pair with the matching pair value of the feature point descriptor of the current monitoring screen and the feature point descriptor of the pre-stored screen to determine whether the camera has shifted.

[0017] Compared with the prior art, the beneficial effect of the camera offset detection method based on feature point matching of the present invention is that: by extracting feature points of the current monitoring screen of the camera and feature points of multiple pre-stored standard monitoring screens at multiple time points, the feature point descriptor of the current monitoring screen is matched with the feature point descriptor of the pre-stored standard monitoring screens at multiple time points using the distance calculation method, and then combined with the set threshold of the number of matching points to determine whether the camera has shifted, it can be used in the security field to ensure the normal operation of equipment. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a structural block diagram of a camera offset detection method based on feature point matching according to the present invention;

[0020] Figure 2 This is a schematic diagram of the image SIFT feature extraction process of a camera offset detection method based on feature point matching according to the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] In the description of this invention, it should be noted that the terms "top," "bottom," "one side," "the other side," "front," "rear," "middle part," "inside," and "top" are used interchangeably.

[0023] The orientation or positional relationship indicated by terms such as "bottom end" is based on the orientation or positional relationship shown in the accompanying drawings and is only for the convenience of describing the invention and simplifying the description, and is not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention; the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance; furthermore, unless otherwise expressly specified and limited, the terms "installed," "connected," and "linked" should be interpreted broadly, for example, they can be fixed connections, detachable connections, or integral connections; they can be mechanical connections or electrical connections; they can be direct connections or indirect connections through an intermediate medium; they can be internal connections between two elements. For those skilled in the art, the specific meaning of the above terms in the invention can be understood according to the specific circumstances.

[0024] like Figure 1 and Figure 2 The method for camera offset detection based on feature point matching shown includes an image preprocessing module, a feature point extraction module, a feature point matching module, and a threshold judgment module; the image preprocessing module, feature point extraction module, feature point matching module, and threshold judgment module are connected in sequence.

[0025] The image preprocessing module preprocesses the acquired camera surveillance images, converting them into grayscale images.

[0026] The feature point extraction module extracts SIFT features from the preprocessed surveillance image to obtain image feature points and feature point descriptors.

[0027] The feature matching module, based on the extracted feature points and feature point descriptors of the camera monitoring screen image, uses the distance calculation method to match the feature point descriptors of the current monitoring screen with the feature point descriptors of the standard monitoring screen at multiple pre-stored time points.

[0028] The threshold judgment module compares the threshold of the set matching pair with the matching pair value of the feature point descriptor of the current monitoring screen and the feature point descriptor of the pre-stored screen to determine whether the camera has shifted.

[0029] Specifically, the process involves acquiring the current monitoring screen image and performing grayscale processing, converting the RGB three-channel image into a single-channel grayscale image.

[0030] The camera offset detection method based on feature point matching of the present invention includes the following steps: Step (1) An image preprocessing module acquires the current monitoring screen image and preprocesses the image. Step (2) A feature point extraction module extracts SIFT features from the preprocessed monitoring screen image to obtain image feature points and feature point descriptors. Step (3) A feature matching module, based on the extracted camera monitoring screen image feature points and feature point descriptors, uses a distance calculation method to match the current monitoring screen feature point descriptors with the feature point descriptors of multiple pre-stored standard monitoring screens at various time points. Step (4) A threshold judgment module compares the threshold of the set matching pair with the matching pair value of the current monitoring screen feature point descriptors and the feature point descriptors of the pre-stored screens to determine whether the camera has shifted.

[0031] Among them, the feature point information of the pre-stored standard monitoring screen is a set of SIFT feature point information extracted from the camera monitoring screens at multiple time points collected in advance.

[0032] SIFT features are local image features that are invariant to rotation, scaling, and brightness changes. They also maintain a certain degree of stability against viewpoint changes, affine transformations, and noise, and exhibit good matching compatibility with object motion, occlusion, and noise. This feature is based on the idea of ​​image feature scale selection, establishing a multi-scale space for the image. It detects the same feature point at different scales, determining the feature point's location and scale simultaneously to achieve scale resistance. Then, it removes some low-contrast points and edge response points, and extracts rotation-invariant feature descriptors to achieve resistance to affine transformations.

[0033] The SIFT feature extraction steps are as follows: 1) Construct a scale space, detect extreme points, and obtain scale invariance; 2) Filter feature points and perform precise localization; 3) Specify orientation parameters for each feature point; 4) Generate descriptors for feature points.

[0034] Feature point matching is performed by calculating the distance between feature point descriptors. The matched feature pairs may contain some incorrect matches. By manually setting filtering conditions, abnormal matches are removed, such as those where the ratio of the nearest distance to the second nearest distance between feature point descriptors is greater than a certain value. This filtering process greatly ensures the accuracy of the matched pairs.

[0035] The distance between feature point descriptors is calculated based on the Euclidean distance between the descriptors of two feature points. Taking two feature points A and B as an example, assuming that the descriptor of feature point A is (x1, y1) and the descriptor of feature point B is (x2, y2), the formula for calculating the distance dist between these two feature point descriptors is as follows:

[0036]

[0037] The threshold judgment involves comparing the threshold of the set matching pair with the matching pair value of the feature point descriptor of the current monitoring screen and the feature point descriptor of the pre-stored screen to determine whether the camera has shifted.

[0038] Example

[0039] like Figure 1 and Figure 2 The method shown is a camera offset detection method based on feature point matching, which specifically includes the following steps:

[0040] To capture surveillance footage from the camera;

[0041] The current monitoring image is converted to grayscale, transforming the RGB three-channel image into a single-channel grayscale image. The transformation method is to sequentially read the R, G, and B values ​​of each pixel in the image, calculate the grayscale value Grey using a formula (converted to an integer), and assign the grayscale value to the corresponding position in the new image. The conversion is completed after traversing all pixels once.

[0042] Gray=0.299*R+0.587*G+0.114*B

[0043] SIFT feature extraction is performed on the preprocessed image. 1) Scale space is constructed, and extreme points are detected. Images in all scale spaces are searched, and potential scale-invariant key extreme points are detected using the Gaussian differential function. 2) Feature points are filtered and precisely located. At each candidate location, a fitted fine model is used to determine the location scale, and relatively stable feature points are selected. 3) Orientation parameters are assigned to each feature point. Based on the local gradient direction of the image, the gradient distribution characteristics of the neighboring pixels of the feature point are used to determine its orientation parameters. Then, the stable orientation of the local structure of the feature point is obtained using the gradient histogram of the image. 4) Descriptors of the feature points are generated. The feature point is rotated to the specified direction and the gradient magnitude and direction of the pixels in the neighborhood of the feature point are used to divide the pixels in the neighborhood into several blocks. Each block is a histogram of multiple directions within its pixels, which is statistically analyzed and normalized to form a feature point descriptor. Finally, multiple feature point descriptors are extracted from an image. These feature point descriptors constitute a feature representation S of the image. The feature representation S contains multiple feature point descriptors, the number of which is the same as the number of feature points extracted from the image, denoted as m.

[0044] Pre-collect surveillance footage from cameras at multiple time points, and extract corresponding SIFT feature point descriptors from each of these footage segments. The SIFT feature point descriptors from the footage at different time points are then used to construct a feature representation set {R1, R2, R3, R4, ..., R...}.i}, where R i It is a feature representation of a surveillance video at a specific point in time, pre-stored by the camera. Feature representation R i It contains multiple feature point descriptors, the number of which is the same as the number of feature points extracted from the image, denoted as n. i The feature set is {R1, R2, R3, R4, ..., R}. i The set of feature points corresponding to} is denoted as {n1, n2, n3, n4...., n i};

[0045] The distance between feature point descriptors is the Euclidean distance calculated based on the descriptors of two feature points. Taking two feature points A and B as an example, assuming the descriptor of feature point A is (x1, y1) and the descriptor of feature point B is (x2, y2), the formula for calculating the distance dist between these two feature point descriptors is as follows:

[0046]

[0047] By calculating the feature representation S of the current monitoring screen and the feature representation R of the monitoring screen at different time points stored in the database, i Feature point descriptor matching is performed based on the distance between them. The resulting feature representation pairs may contain some incorrect matches. By calculating the ratio of the nearest distance between feature point descriptors to the second nearest distance between them, matching pairs with a ratio greater than 0.7 are filtered out. This yields a set of matching pairs {N1, N2, N3, N4, ..., N...} between the feature representation of the current monitoring screen and the feature representations of monitoring screens at different pre-stored time points. i}, where N i It is the number of matching pairs between the feature representation of the current monitoring screen and the feature representation of the monitoring screen at a certain time point in the pre-stored data.

[0048] Using the set of matching pairs {N1, N2, N3, N4...., N i The number of feature points {n1, n2, n3, n4...., n} in the feature representation of the monitoring images at different pre-stored time points. i The set of ratios of} To determine if the camera has shifted, when any element in the ratio set... When the value is greater than or equal to the threshold of 0.6, it is determined that the camera has not shifted; otherwise, when all elements in the ratio set are less than 0.6, it is determined that the camera has shifted.

[0049] It should be noted that, in this document, terms such as "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0050] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A camera shift detection method based on feature point matching, characterized in that: The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. SIFT feature extraction is performed on the preprocessed image, 1) a scale space is constructed, extreme points are detected, and all image scales are searched, potential key extreme points that are scale invariant are detected through a Gaussian differential function; 2) feature point filtering and accurate positioning are performed, a fitting fine model is used to determine the position scale at each candidate position, and relatively stable feature points are selected; 3) a direction parameter is specified for each feature point, based on the gradient direction of the local image, the gradient distribution characteristics of the feature point neighborhood pixels are used to determine the direction parameter of the feature point, and the stable direction of the local structure of the feature point is obtained by using the gradient histogram of the image; 4) a feature point descriptor is generated, the rotation is corrected to the specified direction with the feature point as the center, the gradient amplitude and direction of the pixels in the neighborhood of the feature point are divided into several blocks, the histogram statistics of multiple directions in each block are normalized, and the feature point descriptor is formed, finally a plurality of feature point descriptors are extracted from an image, and these feature point descriptors constitute a feature representation S of the image, the feature representation S contains a plurality of feature point descriptors, the number is consistent with the number of feature points extracted from the image, and is denoted as m; The monitoring pictures of the camera at multiple time points are collected in advance, and corresponding SIFT feature point descriptors are extracted from the monitoring pictures respectively, and the SIFT feature point descriptors of the monitoring pictures at different time points are constructed into a feature representation set wherein is a feature representation of a monitoring picture of the camera at a certain time point stored in advance, and the feature representation contains a plurality of feature point descriptors, the number of which is consistent with the number of feature points extracted from the image, denoted as , the feature representation set , and the corresponding feature point number set is denoted as ; The distance between feature point descriptors is the Euclidean distance calculated based on the descriptors of two feature points. Taking two feature points A and B as an example, assuming that the descriptor of feature point A is , and the descriptor of feature point B is , the calculation formula of the distance dist between the two feature point descriptors is as follows: The feature point descriptor matching is performed by calculating the distance between the feature representation S of the current monitoring picture and the feature representation of the monitoring picture at different time points stored in advance. The matching feature representation may contain some false matching pairs. By calculating the ratio of the nearest distance between the feature point descriptors to the second nearest distance between the feature point descriptors, the matching pairs with a ratio greater than 0.7 are filtered out, and the matching pair number set of the feature representation of the current monitoring picture and the feature representation of the monitoring picture at different time points stored in advance is obtained , wherein is the matching pair number of the feature representation of the current monitoring picture and the feature representation of the monitoring picture at a certain time point stored in advance. Using the set of matching pairs The number of feature points in the feature representation of the monitoring images at different pre-stored time points The set of ratios To determine if the camera has shifted, when any element in the ratio set... When the value is greater than or equal to the threshold of 0.6, it is determined that the camera has not shifted; otherwise, when all elements in the ratio set are less than 0.6, it is determined that the camera has shifted. 2.The camera shift detection method based on feature point matching according to claim 1, characterized in that: The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. 3.The camera shift detection method based on feature point matching of claim 2, wherein: The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected.

4. The camera shift detection method based on feature point matching according to claim 2, characterized in that: The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected.

5. The camera shift detection method based on feature point matching according to claim 4, characterized in that: The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point extraction module, the feature matching module and the threshold judgment module are sequentially connected. The image preprocessing module, the feature point 6. The camera offset detection method based on feature point matching according to claim 2, characterized in that: The feature matching is performed by calculating the distance between the feature point descriptors, and the matched feature pairs can contain some false matching pairs. Some screening conditions are set artificially to eliminate the abnormal matching pairs, and the matching pairs with the ratio of the nearest distance between the feature point descriptors to the second nearest distance between the feature point descriptors greater than a certain value.

7. The camera shift detection method based on feature point matching according to claim 1, characterized in that: The threshold judging module compares the threshold of the matched pairs with the matching pair value of the feature point descriptors of the current monitoring picture and the pre-stored picture, so as to judge whether the camera is deviated or not. ​

Images