Multi-target parameter testing method and system, computer equipment and storage medium

Abstract

The invention provides a multi-target parameter testing method, which belongs to the field of military testing, and comprises the following steps: six detection light curtains which are arranged in sequence intersect in pairs in a space to form a testing area, and a multi-parameter calculation model is constructed according to arrangement parameters of the testing area. And acquiring multiple groups of imaging information when multiple targets pass through the test area, and sequentially performing feature extraction, fusion and deconvolution operation on the imaging information through the deformable convolutional network to obtain different images when different targets pass through the multiple detection light curtains. And determining pixel coordinates when different targets pass through different detection light curtains according to the processed images, searching time information according to the pixel coordinates, inputting multiple pieces of time information of the same target into a target multi-parameter calculation model, and outputting speed and position information of the same target. According to the invention, the final speed and position information of the target can be accurately obtained under the condition that multiple targets pass through the detection light curtain at the same time and are shielded.

Description

Technical Field

[0001] The invention belongs to the field of military testing technology, and in particular relates to a multi-target parameter testing method, system, computer equipment and storage medium. Background Art

[0002] With the rapid development of intelligent weapons, the spatial position, speed and other parameters for evaluating their performance have become particularly important. The test methods for obtaining these parameters are challenging and have put forward higher requirements. In the field of range testing, after the weapon launches the target, the measurement methods for the target position, speed and other information are mainly concentrated on the high-speed camera shooting method and the multi-light curtain intersection measurement method. Although the high-speed camera can clearly capture the flying target, it cannot process and calculate the target's parameter information in real time. At the same time, the high-speed camera is expensive and cannot be widely used in various research units. Although the four-light curtain intersection method in the multi-light curtain intersection measurement method can construct the measurement of two-dimensional coordinate parameters, it has strict requirements on the flight direction of the flying target. It requires that the target must pass through the sky curtain array vertically, otherwise a large error will be generated and the parameters cannot be corrected in real time. More importantly, the error generated by the four-light curtain measurement system is difficult to eliminate, which makes the measurement result error relatively large.

[0003] In order to solve this problem, researchers proposed the coordinate measurement principle of the six-light screen measurement system, which can measure two-dimensional coordinates and deflection angles, and can also test multiple targets fired at single or intervals. However, there are still some difficulties in testing the scattered positions of multiple target groups. Because during the weapon launch process, due to the differences between the multiple targets themselves, the speed and attitude angles of the multiple targets are not exactly the same, making it very likely that the previous target will catch up with the next target, which means that multiple targets will block each other in a certain area of ​​the space. For test systems with optical imaging or photoelectric detection as the core, it is not possible to intuitively obtain parameter information of multiple targets that block each other. Summary of the invention

[0004] In order to solve the problem that multiple target groups occlude each other in a certain area and parameter information of the multiple targets that occlude each other cannot be obtained intuitively, the present invention provides a multi-target parameter testing method, specifically a multi-target parameter testing method based on the linear array CCD intersection space-time resolution mechanism.

[0005] In order to achieve the above object, the present invention provides the following technical solutions:

[0006] A multi-objective parameter testing method, the specific steps are as follows:

[0007] Constructing a target multi-parameter calculation model of a test area; the test area is formed by six detection light curtains arranged in sequence intersecting in space in pairs, and the multi-parameter calculation model is constructed according to the arrangement parameters of the six detection light curtains;

[0008] Obtain multiple sets of imaging information of multiple targets passing through the test area;

[0009] The deformable convolution module of the deformable convolutional network is used to extract the shape and scale features of different images in multiple groups of imaging information. The multi-scale convolution module separates multiple imaging features of multiple targets when they are mutually occluded according to the shape and scale features of different images. The feature fusion module adds and fuses the multiple imaging features when they are mutually occluded, and then performs a deconvolution operation on the fused features to obtain different images of different targets when they pass through multiple detection light curtains.

[0010] Based on the different images of the same target when it passes through different detection light curtains, multiple pixel coordinates of the same target when it passes through different detection light curtains are determined; based on the multiple pixel coordinates of the same target when it passes through different detection light curtains, multiple moment information of the same target passing through the multiple detection light curtains is found; the multiple moment information of the same target is input into the target multi-parameter calculation model, and the speed and position information of the same target is output.

[0011] Preferably, the six detection light curtains are detection light curtains M arranged in sequence. 1 ~Detection light curtain M 6 , six detection light curtains arranged in sequence intersect in space to form a test area, specifically:

[0012] The detection light curtain M 1 And detection light curtain M 5 Placed vertically on the trajectory, and M 1 and M 5 Parallel to each other, M 1 and M 5 The target distance between them is S;

[0013] The detection light curtain M 2 And detection light curtain M 6 It is arranged in an inclined state with the ground, and M 2 and M 6 Respectively with M 1 and M 5 Intersection, the intersection angle is α;

[0014] The detection light curtain M 3 And detection light curtain M 4 All are arranged vertically and crosswise on the ground, M 3 and M 4 Respectively with M 1 and M 5 The intersection angle is β, forming a test area.

[0015] Preferably, the multi-parameter calculation model is constructed according to the arrangement parameters of the six detection light curtains, specifically:

[0016] Define the i-th target among multiple targets passing through the first detection light curtain M 1 The position (x i ,y i ), speed v across the test area i and attitude angle (θ i ,γ i ), the first detection light curtain M 1 The coordinate system of the plane is oxyz; assuming o i1 o i2 is the actual flight path of the ith target, o′ i1 o′ i2 For vertical flight path, detection light curtain M 3 and M 4 The intersection line passes through the detection light curtain M 1 and M 5 The midpoint of o i1 o i2 The angles θ are formed with the xoy plane and the xoz plane respectively. i and γ i ;

[0017] According to the law of uniform linear motion of multiple targets in the test area, combined with the intersection angles α and β between the detection light curtains arranged at the test site, and the target distance S, a target multi-parameter calculation model is constructed. The specific model is:

[0018]

[0019] Where, t i1 -t i6 They are the time information of the target passing through each detection light curtain; x i represents the position of the i-th target in the x direction; y i represents the position of the i-th target in the y direction; v i Represents the flight speed of the i-th target.

[0020] Preferably, a light curtain array composed of 6 intersecting linear CCDs is used to construct a test area.

[0021] Preferably, after obtaining multiple sets of imaging information of multiple targets passing through the test area, the image is further enhanced using a Retinex algorithm based on illumination component correction, specifically including:

[0022] Decomposing images in the plurality of sets of imaging information into illumination components and reflection components;

[0023] The platform histogram method is used to process the irradiation component to obtain a balanced image;

[0024] The image noise of the reflection component is suppressed by a non-local mean filter to obtain the filtered image reflection component, and the local variance is used to determine the size of the high-frequency component in the filtered image reflection component to obtain the image reflection component after local detail enhancement;

[0025] The information enhanced image is input into the deformable convolutional network for processing.

[0026] Preferably, the deformable convolution module using the deformable convolution network extracts the shape and scale features of different images in the multiple groups of imaging information, the multi-scale convolution module separates the multiple imaging features of multiple targets when they are mutually occluded according to the shape and scale features of different images, the feature fusion module adds and fuses the multiple imaging features when they are mutually occluded, and then performs a deconvolution operation on the fused features to obtain different images of different targets when they pass through multiple detection light curtains, specifically including:

[0027] A convolution layer is added to the original convolution of the deformable convolution module, and the shape and scale features of different images in multiple sets of imaging information are extracted through the deformable convolution module after the added convolution layer;

[0028] Convolution kernels of different sizes are introduced into the multi-scale convolution module. By using convolution kernels of different sizes for the shape and scale features of different images, multiple imaging features when multiple targets are mutually occluded are extracted. The multiple imaging features when mutually occluded are added and fused through the feature fusion module.

[0029] The fused features are then deconvolved to obtain different images of different targets passing through multiple detection light curtains.

[0030] Preferably, the method of determining multiple pixel coordinates of the same target when it passes through different detection light curtains according to different images of the same target when it passes through the multiple detection light curtains, finding multiple time information of the same target passing through the multiple detection light curtains according to the multiple pixel coordinates of the same target when it passes through the different detection light curtains, inputting the multiple time information of the same target into the target multi-parameter calculation model, and outputting the speed and position information of the same target specifically includes:

[0031] Determine multiple pixel coordinates of the same target when it passes through different detection light curtains according to different images of the same target when it passes through multiple detection light curtains;

[0032] According to the multiple pixel coordinates of the same target when it passes through different detection light curtains, the imaging time information of the target on the non-mutually occluded detection light curtain and the parallel detection light curtain is first obtained, and the time difference set is calculated according to the arrangement and combination method; then the time difference set of the target on the mutually occluded detection light curtain and the parallel detection light curtain is calculated, and the time difference of the real target in the time difference set is found according to the law of linear motion of the target in finite space and the principle of time difference approximation between parallel light curtains, and the time information of the same target passing through the detection light curtain among the mutually occluded targets is determined;

[0033] According to the planar symmetry characteristics of parallel detection light curtains, the principle that the target imaging position is related to the attitude angle is adopted to determine the same target passing through all detection light curtains, and give the corresponding moment information of the target; then combined with the target multi-parameter calculation model, the multi-target position and velocity information is obtained.

[0034] The present invention also provides a multi-objective parameter testing system, comprising:

[0035] A model building module, used to build a target multi-parameter calculation model of a test area; the test area is formed by six detection light curtains arranged in sequence and intersecting in space in pairs, and the multi-parameter calculation model is built according to the arrangement parameters of the six detection light curtains;

[0036] An image acquisition module, used to acquire imaging information of multiple targets passing through a test area;

[0037] An image segmentation module is used to extract the shape and scale features of different images in multiple groups of imaging information using a deformable convolution module of a deformable convolution network. A multi-scale convolution module separates multiple imaging features of multiple targets when they are mutually occluded according to the shape and scale features of different images. A feature fusion module adds and fuses multiple imaging features when they are mutually occluded, and then performs a deconvolution operation on the fused features to obtain different images of different targets when they pass through multiple detection light curtains.

[0038] The parameter prediction module is used to determine multiple pixel coordinates of the same target when it passes through different detection light curtains based on different images of the same target when it passes through multiple detection light curtains, find multiple moment information of the same target passing through multiple detection light curtains based on the multiple pixel coordinates of the same target when it passes through different detection light curtains, input the multiple moment information of the same target into the target multi-parameter calculation model, and output the speed and position information of the same target.

[0039] The present invention also provides a computer device, comprising a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement any one of the steps in the multi-objective parameter testing method.

[0040] The present invention also provides a computer-readable storage medium, on which a computer program is stored, characterized in that when the computer program is loaded by a processor, it can execute any one of the steps in the multi-objective parameter testing method.

[0041] The multi-objective parameter testing method provided by the present invention has the following beneficial effects:

[0042] The present invention utilizes six detection light curtains arranged in sequence to intersect in space to form a test area, and constructs a multi-parameter calculation model based on the arrangement parameters of the test area. On this basis, multiple groups of imaging information of multiple targets passing through the test area are obtained, and the shape and scale features of different images in the multiple groups of imaging information are sequentially extracted through a deformable convolutional network, and multiple imaging features of multiple targets when they are mutually occluded are separated according to the shape and scale features of different images, so that the target image with mutual occlusion in a certain area can be segmented and processed, and finally the multiple imaging features when they are mutually occluded are added and fused, and then the fused features are deconvolved to obtain different images when different targets pass through multiple detection light curtains, that is, the image of each target passing through each detection light curtain is identified through multiple groups of imaging information, and the complete image information of the target passing through the light curtain is obtained. Then, the pixel coordinates of different targets passing through different detection light curtains are determined according to the processed images, and then the time information is found according to the pixel coordinates, and the multiple time information of the same target is input into the target multi-parameter calculation model, and the speed and position information of the same target is output, so that the final speed and position information of the target can be accurately obtained when multiple targets pass through the detection light curtain at the same time and there is occlusion. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiment of the present invention and its design scheme, the following briefly introduces the drawings required for this embodiment. The drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0044] Figure 1 A flowchart of a multi-objective parameter testing method provided by an embodiment of the present invention;

[0045] Figure 2 A schematic diagram of a six-line CCD intersection test provided by an embodiment of the present invention;

[0046] Figure 3 A side view schematic diagram of a six-line array CCD intersection test provided by an embodiment of the present invention;

[0047] Figure 4 A schematic top view of a six-line array CCD intersection test provided in an embodiment of the present invention;

[0048] Figure 5This is a schematic diagram of multi-target linear array CCD unit detection light curtain imaging under the condition of no mutual obstruction provided by an embodiment of the present invention;

[0049] Figure 6 Schematic diagram of imaging of a multi-target linear array CCD unit detecting light curtain under mutual occlusion provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0050] In order to enable those skilled in the art to better understand the technical solution of the present invention and implement it, the present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the scope of protection of the present invention.

[0051] In the description of the present invention, it is to be understood that the terms “center”, “longitudinal”, “lateral”, “length”, “width”, “thickness”, “up”, “down”, “front”, “back”, “left”, “right”, “vertical”, “horizontal”, “top”, “bottom”, “inside”, “outside”, “axial”, “radial”, “circumferential”, etc., indicating orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, are only for the convenience of describing the technical solutions of the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0052] In addition, the terms "first", "second", etc. are used for descriptive purposes only and are not to be understood as indicating or implying relative importance. In the description of the present invention, it should be noted that, unless otherwise clearly specified or limited, the terms "connected" and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to the specific circumstances. In the description of the present invention, unless otherwise specified, "plurality" means two or more, which will not be described in detail here.

[0053] Example 1

[0054] With the development of various new types of weapons, the position, speed and other parameter information of the multi-target group formed by continuous shooting is used as an important indicator for evaluating performance. The characteristics of the multi-target group are high speed, uniformity and diversity. When multiple targets arrive at the test area at the same time, the present invention adopts a light curtain array formed by the intersection of multiple linear array CCDs with time and space resolution characteristics, and introduces a deep learning network model to effectively identify the imaging information of the real target and deduce the parameter information of the target.

[0055] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention is described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0056] Reference Figure 1-Figure 6 The present invention provides a multi-target parameter testing method, specifically a multi-target parameter testing method based on the time-space resolution mechanism of linear array CCD intersection, as a means of testing multiple parameters of multiple targets in a predetermined cross section when a weapon continuously launches multiple targets. The concept of the present invention is:

[0057] The test area is formed by a light curtain array composed of six detection light curtains intersecting in space. According to the law of uniform linear motion of multiple targets in the test area under continuous launch of weapons and combined with the test site layout parameters, a multi-parameter calculation model of multiple targets reaching a predetermined cross-section in the system test area (referred to as the target multi-parameter calculation model) is constructed. The detection light curtain is formed by a detection optical system with a high-speed and high-sensitivity linear array CCD as the core.

[0058] The image information of multiple targets passing through the test area is collected by linear CCD, and the Retinex algorithm based on illumination component correction is used to improve the image contrast. The image enhancement effect is evaluated by information entropy and average gradient.

[0059] Based on the enhanced image, a deformable convolutional network with spatial geometric deformation ability is used to learn the shape and scale features of multiple targets under mutual occlusion in the test area. A multi-target imaging segmentation method with a network model with deep learning ability is formed, and the weighted sum loss function is used to evaluate the segmentation results. According to the target segmentation results of the image acquired by the linear array CCD, an adaptive fuzzy classifier learning algorithm is used to classify the imaging information of the image segmentation targets.

[0060] The coordinates and time information of multi-target linear CCD imaging are established in the form of a matrix. The time information of the same target passing through multiple detection light curtains is found out according to the characteristics of spatiotemporal resolution of multi-target imaging. The speed and position information of multi-targets are calculated through the target multi-parameter calculation model.

[0061] Specifically, Figure 1 As shown, the multi-objective parameter testing method provided by the present invention comprises the following steps:

[0062] Step 1: Use high-speed and high-sensitivity linear array CCD as the core to form a light curtain array with 6 linear array CCD intersections, and define the i-th target among multiple targets passing through the first detection light curtain M 1 The position (x i ,y i ), speed v across the test area i and attitude angle (θ i ,γ i), according to the law of uniform linear motion of multiple targets in the test area, combined with the intersection angles α and β between the detection light curtains arranged at the test site, and the target distance S, a target multi-parameter calculation model is constructed.

[0063] Specific reference Figure 2 and Figure 3 The specific steps of constructing the target multi-parameter calculation model based on the light curtain array where 6 linear CCDs intersect are as follows:

[0064] Step 11: Use high-speed and high-sensitivity linear array CCD as the core to form a light curtain array with 6 linear array CCD intersections, such as Figure 2 As shown, the six detection light curtains are detection light curtains M arranged in sequence. 1 ~Detection light curtain M 6 , that is, M 1 ~M 6 Indicates 6 detection light curtains. Detection light curtain M 1 and M 5 The detection light curtains are perpendicular to the ground and parallel to each other. In the test arrangement, the two detection light curtains are placed on the trajectory line, and the target distance between them is S; the detection light curtain M 2 and M 6 They are arranged in an inclined intersection with the ground, and they are respectively connected to the detection light curtain M 1 and M 5 Intersection, the intersection angle is α; detection light curtain M 3 and M 4 is a detection light curtain vertical to the ground, respectively connected to the detection light curtain M 1 and M 5 The six detection light curtains arranged in sequence intersect in pairs in the space to form a test area.

[0065] Step 12: Define the i-th target among multiple targets to pass through the first detection light curtain M 1 The position of (x i ,y i ), speed v across the test area i and attitude angle (θ i ,γ i ), the first detection light curtain M 1 The plane is the coordinate system oxyz of the system; Figure 2 The six detection light screens are projected onto the yoz plane and the xoz plane respectively, as shown in Figure 3 and Figure 4 As shown, assuming o i1 o i2 is the actual flight path of the ith target, o′ i1 o′ i2 For vertical flight path, detection light curtain M 3 and M 4The intersection line passes through the detection light curtain M 1 and M 5 The midpoint of o i1 o i2 With the xoy plane and xoz plane, they form a certain angle θ i and γ i .

[0066] Step 13: According to the law of uniform linear motion of multiple targets in the test area, combined with the intersection α, β between the detection light curtains arranged at the test site, and the target distance S, a target multi-parameter calculation model is constructed. The specific model is:

[0067]

[0068] In the formula, t i1 -t i6 They are the time information of the target passing through each detection light curtain; x i represents the position of the i-th target in the x direction; y i represents the position of the i-th target in the y direction; v i represents the flight speed of the i-th target.

[0069] t is the time information when the target passes through each light curtain; x i represents the position of the i-th target in the x direction; y i represents the position of the i-th target in the y direction; v i represents the flight speed of the i-th target.

[0070] Step 2: Collect multiple sets of imaging information of multiple targets passing through the test area based on the linear array CCD, where the i-th image is

[0071] Indicates the jth image in the i'th linear array CCD i' The pixel coordinates of the target on the left and right sides of the trajectory And the target imaging point In time value Photographed by the linear array CCD, i' represents the number of group targets imaged in the i'th linear array CCD, k i ′ represents the number of group targets at the i'th linear array CCD imaging moment; Indicates the time value of the target passing through the detection light curtain of the test area; v(t total ) represents the pixels in the width direction of the image formed after the linear array CCD completes the acquisition, where v(t total ) and the acquisition time t set by the linear array CCD totalAccording to the line scan rate and resolution of the linear array CCD, the time value of any pixel of the linear array CCD imaging can be calculated.

[0072] Step 3: Use the Retinex algorithm based on illumination component correction to improve the image contrast and obtain the enhanced image And through information entropy and the average gradient To evaluate the image enhancement effect, the specific steps are as follows:

[0073] Step 31: Image information collected by the linear array CCD Decomposition into radiation components and the reflected component The irradiation component was analyzed by using the platform histogram method. Processing to obtain an equalized image The image noise is suppressed by the non-local mean filter, and the filtered reflection component is obtained as follows: The local variance is used to determine the reflection component of the filtered image The size of the mid-high frequency components is used to obtain the image reflection component after local detail enhancement

[0074] Step 32: Enhanced Illumination Component With the reflection component Multiply to get the enhanced image And through information entropy and the average gradient To evaluate the effect of image enhancement. The information entropy can characterize the aggregation characteristics of the image grayscale distribution, r(χ) represents the distribution density of the enhanced image grayscale χ, γ′ is the grayscale level of the enhanced image; the average gradient reflects the rate of change of the contrast of the small details of the enhanced image, which is used to characterize the relative clarity of the image, and it is related to the grayscale value h′ of the pixel after enhancement. m , and the width and length of the image. The larger the average gradient, the clearer the image.

[0075] Step 4: Use a deformable convolutional network with spatial geometric deformation capabilities to learn the shape, scale and other features of the enhanced image of multiple targets under mutual occlusion in the test area, form a multi-target imaging segmentation method with a network model with deep learning capabilities, segment the image, and obtain different images of different targets when passing through multiple detection light curtains.

[0076] Furthermore, the steps of the target imaging segmentation method based on the deformable convolutional network in step 4 are as follows:

[0077] Step 41: Use the deformable convolution module to extract the shape and scale features of different images in the multi-target imaging information after image enhancement. Specifically, for the network input, the output mapping Y(a,b) of a certain pixel point (a,b) on the image after multi-target imaging information enhancement is increased by an offset Δa. n Then it is Y((a,b),Δa n ); and add a convolution layer to the original convolution of the deformable convolution, and determine the value of the offset sampling point through the bilinear interpolation algorithm, and the sampling point can be dynamically adjusted according to the shape of the target imaging to obtain the offset Δa n And form an offset domain; during network training, the convolution kernel and offset of the output target features are generated, and the feature information of multi-target imaging is learned at the same time. The shape and scale characteristics of different images in multiple groups of imaging information are extracted through the deformable convolution module after the additional convolution layer.

[0078] Step 42: Use a multi-scale convolution module to extract the features of multi-target mutual occlusion imaging. Use a multi-scale convolution module to extract multi-target imaging features, especially for the extraction of target features where the imaging shapes and the number of imaging images of multiple targets are different. By introducing convolution kernels of different sizes into the multi-scale convolution module, multiple imaging features when multiple targets are mutually occluded are extracted by using convolution kernels of different sizes for the shape and scale features of different images, especially each layer of learning can obtain more multi-target mutual occlusion imaging features.

[0079] Step 43: Use the feature fusion module to refine the target segmentation result. The feature fusion module adds and fuses multiple imaging features when they are mutually occluded; then deconvolution is performed on the fused features to obtain different images of different targets when they pass through multiple detection light curtains.

[0080] Step 44: Establish an evaluation function for the image target segmentation result. Use the weighted sum loss function to evaluate the target segmentation result. The general binary classification problem is represented by the cross entropy loss function. N i is the number of pixels of the image collected by the i-th linear array CCD, y′ i″ is the mark value of the i″th pixel, is the prediction result of the i″th pixel point; for the linear array CCD image acquisition with a single target imaging, the target imaging accounts for a small proportion, and the cross entropy loss function is used. The network is more biased towards the background prediction and is prone to errors; the parameter Δε is introduced to reduce overfitting and establish the dice loss function Based on the imbalance problem of target imaging samples, the cross entropy loss function and the dice loss function Perform weighted summation and add the weight coefficient τ to obtain the weighted summation loss function L new(L loss ,L D-Loss ,τ) is used to evaluate the pixel-level and image-level saliency in image segmentation results.

[0081] Step 5: Use a matrix to establish the coordinates and time information of multi-target linear CCD imaging. According to the characteristics of multi-target imaging time and space resolution, find out the multiple time information of the same target passing through multiple detection light curtains, input the time information into the target multi-parameter calculation model, and calculate the speed and position information of the multiple targets through the target multi-parameter calculation model.

[0082] Specific reference Figure 5 and Figure 6 , Step 4 is to find out the moment information of the same target passing through multiple detection light curtains according to the characteristics of multi-target imaging time and space resolution as follows:

[0083] Step 51: There are two possibilities when multiple targets pass through the detection light curtain. One is that the group of targets do not reach the imaging pixels of the linear array CCD at the same time, and the other is that the group of targets appear on the imaging pixels of the linear array CCD and reach the detection light curtain at the same time. 1 To analyze the main detection light curtain, these two phenomena are Figure 5 and 5 As shown. Definition A 1 ~A 5 Five targets appear simultaneously on the vertical ground detection light curtain M 1 , A 1 '~A 5 ′ is the corresponding imaging pixel position of the target, Figure 5 It can be found that the imaging pixel positions of the five targets in the linear array CCD are distinguishable. According to the consistency of the target flight direction and the geometric structure relationship of the detection light curtain array, the imaging pixel positions of the five targets in the linear array CCD of other detection light curtains are also distinguishable. By using the target imaging pixel position information in each detection light curtain and combining the line scanning rate Vc of the linear array CCD, the time value of each target passing through each detection light curtain in the test area can be calculated. These time information are substituted into the calculation model formulas (1) and (2) established according to the spatial relationship of the intersection of the detection light curtains formed by the six linear array CCDs to obtain the position, velocity and attitude angle information of multiple targets.

[0084] But for Figure 6 , A that arrives simultaneously appears on the imaging pixel of the linear array CCD 1 , A 2 and A 3 , the pixel positions of these three targets on the linear array CCD are the same, namely A 1 ′、A 2 ′ and A 3 ′ are the same pixel point. At this time, the detection light curtain M1 and M 5 , or M 2 and M 6 It is unable to distinguish the target and needs to be combined with the detection light curtain M 3 and M 4 Can be quickly distinguished.

[0085] Step 52: Based on the characteristics of spatial and temporal resolution of group target imaging, a matrix is ​​used to describe the pixel information of n targets in 6 detection light curtains, and k targets appear on a certain detection light curtain and reach the same time and are imaged on the linear array CCD.

[0086]

[0087] Among them, M 1 -M 6 is the pixel coordinate.

[0088] Step 53: Forming a method for matching information at the same target time, specifically: for the detection light curtain M 1 There are nk targets simultaneously on the pixels of the linear array CCD image. First, obtain their positions on the detection light curtain M. 2 With M 6 Imaging time information in and And find their time difference set Τ according to the permutation and combination method 61 ; Then calculate the mutual occlusion targets on the detection light curtain M 5 The time value t (n-k)5 With detection light curtain M 1 The time value t (n-k)1 difference According to the law of linear motion of the target in limited space, the time difference set Τ is found according to the principle of time difference approach between parallel light curtains. 61 A time difference in Determine that the same target among the mutually occluded targets passes through the detection light curtain M 1 、M 2 、M 5 and M 6 Second, according to the detection light curtain M 4 and M 3 Along the parallel detection light curtain M 1 The plane symmetry feature of the target imaging position is multiplied by the calculation result of ctgβ. The principle of determining the M through the detection light curtain 3 and M 4 The same target is detected and the target is calculated to pass through the detection light curtain M 3 and M 4 The time information, similarly, using or According to the principle of , the same target passing through all detection light curtains is determined, and the corresponding time information of the target is given; finally, based on the obtained time information of the same target, combined with formulas (1) and (2), the position, velocity and attitude angle information of n targets are calculated respectively.

[0089] The present invention uses six linear array CCDs to intersect in space to form a detection light curtain array with a spatial geometric relationship as a test area for multi-target parameters, extracts target imaging information and performs segmentation through image enhancement, segmentation and other methods, finds the moment information of the same target passing through multiple detection light curtains based on the characteristics of the time-space resolution of multi-target imaging of the light curtain array, and then accurately calculates the position and speed of each target, which is conducive to objectively evaluating the performance of intelligent weapons. The key point of the present invention is to use a linear array CCD to obtain image information of the target before, during and after passing through the light curtain, use image processing to obtain information of the real target, and combine the spatial geometric relationship of the light curtain of the test equipment to find the same target passing through the light curtain array, and then deduce the moment information of the target passing through the light curtain, providing reference value for the evaluation of weapons.

[0090] Based on the same inventive concept, the present invention also provides a multi-objective parameter testing system, comprising:

[0091] The model building module is used to build a target multi-parameter calculation model of the test area; the test area is formed by six detection light curtains arranged in sequence intersecting in space, and the multi-parameter calculation model is built according to the arrangement parameters of the six detection light curtains.

[0092] The image acquisition module is used to acquire imaging information of multiple targets passing through the test area.

[0093] The image segmentation module is used to extract the shape and scale features of different images in multiple groups of imaging information using the deformable convolution module of the deformable convolution network. The multi-scale convolution module separates multiple imaging features of multiple targets when they are mutually occluded according to the shape and scale features of different images. The feature fusion module adds and fuses the multiple imaging features when they are mutually occluded, and then performs a deconvolution operation on the fused features to obtain different images of different targets when they pass through multiple detection light curtains.

[0094] The parameter prediction module is used to determine multiple pixel coordinates of the same target when it passes through different detection light curtains based on different images of the same target when it passes through multiple detection light curtains, find multiple moment information of the same target passing through multiple detection light curtains based on the multiple pixel coordinates of the same target when it passes through different detection light curtains, input the multiple moment information of the same target into the target multi-parameter calculation model, and output the speed and position information of the same target.

[0095] Each module in the above multi-objective parameter test system can be implemented in whole or in part by software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each module.

[0096] The present invention also provides a computer device, including a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps in the multi-objective parameter testing method embodiment. The specific implementation method can be found in the method embodiment, which will not be repeated here.

[0097] Furthermore, the present invention also provides a non-temporary computer-readable storage medium containing instructions, and a computer program is stored on the storage medium. For example, a memory containing instructions, the above instructions can be executed by a processor of a computer device to complete the above method. For example, a non-temporary computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, and an optical data storage device. When the computer program is executed by the processor, the steps in the embodiment of the multi-objective parameter testing method can be implemented. The specific implementation method can be found in the method embodiment, which will not be repeated here.

[0098] It will be appreciated by those skilled in the art that embodiments of the present invention may provide methods, systems or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0099] The present invention is described with reference to flowcharts and / or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as a combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0100] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0101] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0102] It should be pointed out that the specific implementation methods described above can enable those skilled in the art to understand the invention more comprehensively, but do not limit the invention in any way. Therefore, although the invention has been described in detail in this specification and embodiments, those skilled in the art should understand that the invention can still be modified or replaced by equivalents; and all technical solutions and improvements that do not deviate from the spirit and scope of the invention are included in the protection scope of the patent for the invention. Any figure mark in the claims should not be regarded as limiting the claims involved. Any simple change or equivalent replacement of the technical solution that can be obviously obtained by any technician familiar with the field within the technical scope disclosed in the present invention belongs to the protection scope of the present invention.

Claims

1. A multi-objective parameter testing method, characterized in that: The following steps are involved: Constructing a target multi-parameter calculation model of a test area; the test area is formed by six detection light curtains arranged in sequence intersecting in space in pairs, and the multi-parameter calculation model is constructed according to the arrangement parameters of the six detection light curtains; Obtain multiple sets of imaging information of multiple targets passing through the test area; The deformable convolution module of the deformable convolutional network is used to extract the shape and scale features of different images in multiple groups of imaging information. The multi-scale convolution module separates multiple imaging features of multiple targets when they are mutually occluded according to the shape and scale features of different images. The feature fusion module adds and fuses the multiple imaging features when they are mutually occluded, and then performs a deconvolution operation on the fused features to obtain different images of different targets when they pass through multiple detection light curtains. Based on the different images of the same target when it passes through different detection light curtains, multiple pixel coordinates of the same target when it passes through different detection light curtains are determined; based on the multiple pixel coordinates of the same target when it passes through different detection light curtains, multiple moment information of the same target passing through the multiple detection light curtains is found; the multiple moment information of the same target is input into the target multi-parameter calculation model, and the speed and position information of the same target is output.

2. The multi-objective parameter testing method according to claim 1, characterized in that: The six detection light curtains are detection light curtains M1 to detection light curtains M6 arranged in sequence. The six detection light curtains arranged in sequence intersect in pairs in space to form a test area, specifically: Place the detection light curtain M1 and the detection light curtain M5 vertically on the ground on the trajectory line, and M1 and M5 are parallel to each other, and the target distance between M1 and M5 is S; The detection light curtain M2 and the detection light curtain M6 are arranged in an inclined intersection state with the ground, and M2 and M6 intersect with M1 and M5 respectively, and the intersection angle is α; The detection light curtain M3 and the detection light curtain M4 are arranged vertically and crosswise on the ground. M3 and M4 intersect with M1 and M5 respectively, and the intersection angle is β, so as to form a test area.

3. The multi-objective parameter testing method according to claim 2, characterized in that: The multi-parameter calculation model is constructed according to the arrangement parameters of the six detection light curtains, specifically: Define the position (x) where the i-th target in the multi-target passes through the first detection light curtain M1 i ,y i ), speed v across the test area i and attitude angle (θ i ,γ i ), the coordinate system oxyz of the first detection light curtain M1 plane; assuming o i1 o i2 is the actual flight path of the ith target, o′ i1 o′ i2 For a vertical flight path, the intersection line of the detection light curtains M3 and M4 passes through the midpoint line of the detection light curtains M1 and M5. i1 o i2 The angles θ are formed with the xoy plane and the xoz plane respectively. i and γ i ; According to the law of uniform linear motion of multiple targets in the test area, combined with the intersection angles α and β between the detection light curtains arranged at the test site, and the target distance S, a target multi-parameter calculation model is constructed. The specific model is: Where, t i1 -t i6 They are the time information of the target passing through each detection light curtain; x i represents the position of the i-th target in the x direction; y i represents the position of the i-th target in the y direction; v i Represents the flight speed of the i-th target.

4. The multi-objective parameter testing method according to claim 3, characterized in that: A light curtain array consisting of 6 linear CCDs intersecting each other is used to construct the test area.

5. The multi-objective parameter testing method according to claim 4, characterized in that: After obtaining multiple sets of imaging information of multiple targets passing through the test area, the image is enhanced by a Retinex algorithm based on illumination component correction, including: Decomposing images in the plurality of sets of imaging information into illumination components and reflection components; The platform histogram method is used to process the irradiation component to obtain a balanced image; The image noise of the reflection component is suppressed by a non-local mean filter to obtain the filtered image reflection component, and the local variance is used to determine the size of the high-frequency component in the filtered image reflection component to obtain the image reflection component after local detail enhancement; The information enhanced image is input into the deformable convolutional network for processing.

6. The multi-objective parameter testing method according to claim 5, characterized in that: The deformable convolution module using the deformable convolution network extracts the shape and scale features of different images in multiple sets of imaging information. The multi-scale convolution module separates multiple imaging features of multiple targets when they are mutually occluded according to the shape and scale features of different images. The feature fusion The module adds and fuses multiple imaging features when they are mutually blocked, and then performs a deconvolution operation on the fused features to obtain different images of different targets when they pass through multiple detection light curtains, including: A convolution layer is added to the original convolution of the deformable convolution module, and the shape and scale features of different images in multiple sets of imaging information are extracted through the deformable convolution module after the added convolution layer; Convolution kernels of different sizes are introduced into the multi-scale convolution module. By using convolution kernels of different sizes for the shape and scale features of different images, multiple imaging features when multiple targets are mutually occluded are extracted. The multiple imaging features when mutually occluded are added and fused through the feature fusion module. The fused features are then deconvolved to obtain different images of different targets passing through multiple detection light curtains.

7. The multi-objective parameter testing method according to claim 6, characterized in that: The method of determining multiple pixel coordinates of the same target when it passes through different detection light curtains according to different images of the same target when it passes through the multiple detection light curtains, finding multiple time information of the same target passing through the multiple detection light curtains according to the multiple pixel coordinates of the same target when it passes through the different detection light curtains, inputting the multiple time information of the same target into the target multi-parameter calculation model, and outputting the speed and position information of the same target specifically includes: Determine multiple pixel coordinates of the same target when it passes through different detection light curtains according to different images of the same target when it passes through multiple detection light curtains; According to the multiple pixel coordinates of the same target when it passes through different detection light curtains, the imaging time information of the target on the non-mutually occluded detection light curtain and the parallel detection light curtain is first obtained, and the time difference set is calculated according to the arrangement and combination method; then the time difference set of the target on the mutually occluded detection light curtain and the parallel detection light curtain is calculated, and the time difference of the real target in the time difference set is found according to the law of linear motion of the target in finite space and the principle of time difference approximation between parallel light curtains, and the time information of the same target passing through the detection light curtain among the mutually occluded targets is determined; According to the planar symmetry characteristics of parallel detection light curtains, the principle that the target imaging position is related to the attitude angle is adopted to determine the same target passing through all detection light curtains, and give the corresponding moment information of the target; then combined with the target multi-parameter calculation model, the multi-target position and velocity information is obtained.

8. A multi-objective parameter testing system, characterized in that: include: A model building module, used to build a target multi-parameter calculation model of a test area; the test area is formed by six detection light curtains arranged in sequence and intersecting in space in pairs, and the multi-parameter calculation model is built according to the arrangement parameters of the six detection light curtains; An image acquisition module, used to acquire imaging information of multiple targets passing through a test area; An image segmentation module is used to extract the shape and scale features of different images in multiple groups of imaging information using a deformable convolution module of a deformable convolution network. A multi-scale convolution module separates multiple imaging features of multiple targets when they are mutually occluded according to the shape and scale features of different images. A feature fusion module adds and fuses multiple imaging features when they are mutually occluded, and then performs a deconvolution operation on the fused features to obtain different images of different targets when they pass through multiple detection light curtains. The parameter prediction module is used to determine multiple pixel coordinates of the same target when it passes through different detection light curtains based on different images of the same target when it passes through multiple detection light curtains, find multiple moment information of the same target passing through multiple detection light curtains based on the multiple pixel coordinates of the same target when it passes through different detection light curtains, input the multiple moment information of the same target into the target multi-parameter calculation model, and output the speed and position information of the same target.

9. A computer device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is loaded into a processor, the computer program can execute the steps of the method according to any one of claims 1 to 7.

Images