Target identification method and device based on bionic insect vision and electronic equipment

Abstract

The invention belongs to the field of visual target recognition, and particularly discloses a bionic insect vision target recognition method and device and electronic equipment, and the method comprises the steps: collecting a video of a target object, and extracting seven continuous images in the video; the seven continuous images are preprocessed; inputting the seven preprocessed continuous images into a lobular giant motion detector LGMD array to obtain six groups of motion feature vectors; based on the six groups of motion feature vectors, obtaining an LGMD feature map of the target object; and inputting the LGMD feature map of the target object into a convolutional neural network to obtain a recognition result of the target object, so that the recognition capability of targets with non-significant static features such as high motion and weak imaging can be improved.

Description

Technical Field

[0001] The present application belongs to the field of visual target recognition, and more specifically, to a target recognition method, device and electronic equipment based on bionic insect vision. Background Art

[0002] Identifying moving targets in scenes through visual technology is the focus of state monitoring and environmental perception. However, with the continuous expansion of visual surveillance application scenarios and the increasing complexity of target motion, the imaging conditions of targets are diverse, highly variable, and weak, making the captured image targets prone to imaging blur, large morphological changes, and much noise interference. The image recognition method of the existing technology mainly analyzes the static features of the target image, such as texture, contour, and color. When analyzing target images whose static features are not significant due to low imaging quality and fast motion speed, it is easy to have low performance in extracting and recognizing target features, resulting in poor recognition ability. Summary of the invention

[0003] In view of the defects of the prior art, the purpose of this application is to provide a target recognition method, device and electronic equipment that imitates the vision of bionic insects, aiming to solve the problem that the visual target recognition method of the prior art has poor recognition ability for target images with insignificant static features.

[0004] To achieve the above objectives, in a first aspect, the present application provides a target recognition method based on bionic insect vision, comprising:

[0005] Collecting a video of a target object and extracting 7 consecutive images from the video;

[0006] Preprocessing the 7 consecutive images;

[0007] The preprocessed 7 consecutive images were input into the lobula giant movement detector (LGMD) array to obtain 6 groups of motion feature vectors;

[0008] Based on the six groups of motion feature vectors, obtaining a LGMD feature map of the target object;

[0009] The LGMD feature map of the target object is input into a convolutional neural network to obtain a recognition result of the target object.

[0010] This application is based on the mechanism by which insects use dynamic vision to identify targets in the natural environment. It uses the LGMD array of the bionic locust compound eye to detect the multi-directional motion characteristics of the target, and uses a convolutional neural network with a simple neural structure of bionic insects to analyze the target motion characteristics. This can effectively extract and identify the target motion characteristics from the image sequence, improve the recognition ability of targets with unobvious static features such as high motion and weak imaging, and overcome the limitations of static feature recognition methods in increasingly complex and changeable scenes.

[0011] According to the target recognition method of bionic insect vision provided by the present invention, the preprocessing of the seven consecutive images includes:

[0012] Taking the origin of the image as the center point, cropping the image into a hexagonal image consistent with the size of the LGMD;

[0013] The hexagonal image is preprocessed.

[0014] This application crops the image into a size that can be input into the LGMD, and then performs preprocessing to ensure that the target image has uniform imaging quality, reduce interference from noise, light, etc., and highlight the motion change information in the image.

[0015] According to the target recognition method of bionic insect vision provided by the present invention, the seven pre-processed continuous images are input into the lobule giant motion detector LGMD array to obtain six groups of motion feature vectors, including:

[0016] Taking each two adjacent images of the seven consecutive images as input, and inputting them into the six LGMDs of the LGMD array respectively;

[0017] The motion feature vectors respectively output by the six LGMDs are obtained.

[0018] According to the target recognition method of bionic insect vision provided by the present invention, the step of obtaining the motion feature vectors respectively output by the six LGMDs comprises:

[0019] For each LGMD, the absolute value of the difference between each pixel in the two input images is calculated;

[0020] Calculating the output values ​​of the excitatory cells and the inhibitory cells based on the absolute value of the difference between each pixel in the two images;

[0021] Calculating cell membrane potential based on the output values ​​of the excitatory cells and the inhibitory cells;

[0022] Processing the cell membrane potential using a low-pass filter, and filtering out cell membrane potentials whose potentials are lower than a preset threshold;

[0023] Based on the filtered cell membrane potential, the motion feature vector is calculated.

[0024] According to the target recognition method of bionic insect vision provided by the present invention, the LGMD feature map of the target object is obtained based on the six groups of motion feature vectors, including:

[0025] The six groups of motion feature vectors are combined to generate an LGMD feature map of the target object.

[0026] According to the target recognition method of bionic insect vision provided by the present invention, after preprocessing the seven consecutive images, the method further includes:

[0027] Based on the preprocessed image, the sampling rate of the target object video is adjusted.

[0028] According to the target recognition method of bionic insect vision provided by the present invention, the convolutional neural network includes 3 convolutional layers, 2 pooling layers and 1 fully connected layer.

[0029] In a second aspect, the present application provides a target recognition device based on bionic insect vision, comprising:

[0030] An acquisition module, used to acquire a video of a target object and extract 7 consecutive images from the video;

[0031] A preprocessing module, used for preprocessing the seven consecutive images;

[0032] An extraction module is used to input the preprocessed 7 consecutive images into the lobular giant motion detector LGMD array to obtain 6 groups of motion feature vectors;

[0033] An acquisition module, used for acquiring a LGMD feature map of the target object based on the six groups of motion feature vectors;

[0034] The recognition module is used to input the LGMD feature map of the target object into a convolutional neural network to obtain a recognition result of the target object.

[0035] In a third aspect, the present application provides an electronic device comprising: at least one memory for storing programs; and at least one processor for executing the programs stored in the memory. When the programs stored in the memory are executed, the processor is used to execute the target recognition method of bionic insect vision described in the first aspect or any possible implementation method of the first aspect.

[0036] In a fourth aspect, the present application provides a computer-readable storage medium, which stores a computer program. When the computer program runs on a processor, the processor executes the target recognition method of bionic insect vision described in the first aspect or any possible implementation of the first aspect.

[0037] In a fifth aspect, the present application provides a computer program product. When the computer program product runs on a processor, the processor executes the target recognition method of bionic insect vision described in the first aspect or any possible implementation of the first aspect.

[0038] It can be understood that the beneficial effects of the second to fifth aspects mentioned above can be found in the relevant description of the first aspect mentioned above, and will not be repeated here.

[0039] In general, the above technical solutions conceived by this application have the following beneficial effects compared with the prior art:

[0040] By referring to the optic nerve structure of the compound eyes of insects, an LGMD array model composed of 6 LGMDs is designed. The features of the collected target image sequence in 6 directions are extracted to generate the LGMD feature map of the target motion. The LGMD feature map is then input into the bionic convolutional neural network for learning and analysis. The target type is identified through simple convolution and pooling operations, thereby improving the recognition ability of targets with unremarkable static features such as high motion and weak imaging. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0042] Figure 1 is a schematic diagram of the flow of a target recognition method of bionic insect vision provided in an embodiment of the present application;

[0043] Figure 2 is a schematic diagram of cropping a target image provided by an embodiment of the present application;

[0044] Figure 3 is a schematic diagram of an LGMD array provided in an embodiment of the present application;

[0045] Figure 4 : is a schematic diagram of a process for obtaining motion feature vectors output by six LGMDs provided in an embodiment of the present application;

[0046] Figure 5 6 feature extraction area division diagrams of LGMD provided in the embodiment of the present application;

[0047] Figure 6 is a schematic diagram of the structure of a target recognition device of bionic insect vision provided in an embodiment of the present application;

[0048] Figure 7 It is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0050] The term "and / or" in this article is a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. The symbol " / " in this article indicates that the associated objects are in an or relationship, for example, A / B means A or B.

[0051] In the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way.

[0052] In the description of the embodiments of the present application, unless otherwise specified, "multiple" means two or more than two. For example, multiple processing units refer to two or more processing units, etc.; multiple elements refer to two or more elements, etc.

[0053] First, the following contents are introduced:

[0054] Identifying targets in a scene based on visual information is the focus of target detection and scene monitoring. The purpose is to collect and analyze the size, shape, texture, speed and other characteristic information of the target in the image, identify the type of the target, and then assist the monitoring computer to grasp the current state of the target in time and make target processing decisions. Visual target recognition is an important part of realizing automated and intelligent monitoring using computers and machine vision, and plays an important role in the fields of automation and artificial intelligence.

[0055] However, with the in-depth application of visual technology in the field of target recognition, many visual targets in recognition scenes show the characteristics of small imaging surface, blurred image, but high speed and complex trajectory. For example, the production process in a manufacturing plant, the movement of goods in a logistics warehouse, urban traffic vehicles and pedestrians, and the monitoring of drones in the air. When using visual methods to recognize targets in these scenes, it is difficult to accurately extract the inherent static features of the target such as size, shape, texture, etc. for feature recognition. In order to be able to recognize targets in scenes where the above static features are insufficient, it is necessary to extract the target's motion features such as the trajectory, speed and direction of the target, and use the motion feature method to recognize the target. However, traditional visual recognition methods are mainly based on static image features, and the ability to extract and recognize motion features in complex scenes is not strong. Although small target recognition and optical flow technology can detect motion features, the efficiency is not high in complex environments. Therefore, there is an urgent need for recognition devices and methods that are more suitable for visual targets with high motion and low imaging quality.

[0056] The visual and nervous systems of insects such as locusts and bees are simpler than those of higher animals, and their ability to process information is not strong. In order to efficiently perceive targets in complex natural environments, they have evolved a visual recognition mechanism that focuses on detecting and identifying target motion characteristics. By drawing on the visual and neural structure mechanisms of insects, it is hoped that efficient visual recognition devices and methods for moving targets in complex environments can be constructed from a new bionic perspective, thereby providing a way to accurately identify targets in scenes where static features are not obvious.

[0057] Therefore, the present application provides a target recognition method, device and electronic device that imitates insect vision, aiming to solve the problem that the visual target recognition method in the prior art has poor recognition ability for target images with insignificant static features.

[0058] Next, combine Figure 1-Figure 5 The target recognition method of bionic insect vision provided in the embodiments of the present application is introduced.

[0059] Figure 1 FIG. 1 is a flow chart of a target recognition method of bionic insect vision provided in an embodiment of the present application. Figure 1 As shown, the method comprises the following steps:

[0060] Step 100, capturing a video of the target object and extracting 7 consecutive images from the video;

[0061] Optionally, a video acquisition device such as a camera may be used to capture a video of the target object, and 7 consecutive images I may be extracted from the video. t (t=1,2,3,…,7), as the input image sequence for target recognition.

[0062] Step 110, preprocessing 7 consecutive images;

[0063] Optionally, after the image is obtained, the image may be preprocessed in order to reduce interference from noise, illumination, etc. and to highlight the motion change information in the image.

[0064] Optionally, the sampling rate of the video acquisition device should be adjusted according to the target's motion speed. If the sampling rate is too high, the two adjacent images are highly similar, the target motion changes are not obvious, and it is difficult to extract the details of the motion features; if the sampling rate is too low, the two adjacent images are very different, the target motion changes too fast, and it is also difficult to extract the details of the motion features. Therefore, it is necessary to adjust the sampling rate of the video acquisition device according to the difference of the preprocessed images to ensure that the preprocessed images can extract the details of the motion features.

[0065] Step 120, inputting the preprocessed 7 consecutive images into the lobular giant motion detector LGMD array to obtain 6 groups of motion feature vectors;

[0066] This application designs an LGMD array model composed of 6 LGMDs by referring to the optic nerve structure of the compound eyes of insects, extracts features in 6 directions from 7 consecutive target image sequences collected, and obtains 6 groups of motion feature vectors.

[0067] Step 130, obtaining a LGMD feature map of the target object based on the six groups of motion feature vectors;

[0068] Step 140: input the LGMD feature map of the target object into a convolutional neural network to obtain a recognition result of the target object.

[0069] This application refers to the simple visual nervous system of insects, uses a simple convolutional neural network to analyze the LGMD feature map, and outputs the recognition result of the target object.

[0070] Optionally, if the target recognition process needs to be terminated, the video acquisition device may be turned off, and a judgment condition may be added to the algorithm. If it is detected that the video acquisition device is turned off, the target recognition process is exited.

[0071] The target recognition method of bionic insect vision provided in the present application designs an LGMD array model composed of 6 LGMDs by referring to the optic nerve structure of the compound eyes of insects, extracts features in 6 directions of the collected target image sequence, generates an LGMD feature map of the target motion, and then inputs the LGMD feature map into the bionic convolutional neural network for learning and analysis. The target type is identified through simple convolution and pooling operations, thereby improving the recognition ability of targets with insignificant static features such as high motion and weak imaging.

[0072] In some embodiments, step 110 specifically includes:

[0073] Step 1101, using the origin of the image as the center point, cropping the image into a hexagonal image consistent with the size of the LGMD;

[0074] Step 1102, pre-processing the hexagonal image.

[0075] Figure 2 is a schematic diagram of cropping a target image provided by an embodiment of the present application, such as Figure 2 As shown in the figure, when preprocessing the image, firstly, according to the hexagonal structure of the insect compound eye, the target image I is taken as the center point. t The image was cropped to a hexagonal shape consistent with the size of the LGMD.

[0076] Optionally, preprocessing the hexagonal image includes using median filtering to denoise the image and improve image clarity; using histogram equalization to enhance image contrast, making the color difference between different areas in the image more obvious and showing image motion changes more clearly; using a sharpening filter to enhance image contours and highlight contour changes caused by motion in the image; and finally normalizing the pixel range to a fixed interval to eliminate brightness and contrast differences between different images.

[0077] Preprocessing ensures that the target image has uniform imaging quality, reduces interference from noise, light, etc., and highlights the motion change information in the image.

[0078] In some embodiments, step 120 specifically includes:

[0079] Step 1201, taking each two adjacent images of the seven consecutive images as inputs, and inputting them into the six LGMDs of the LGMD array respectively;

[0080] Step 1202, obtaining motion feature vectors outputted by six LGMDs respectively.

[0081] Figure 3 is a schematic diagram of the LGMD array provided in the embodiment of the present application, such as Figure 3 As shown, a LGMD array consists of 6 LGMDs, defined as LGMD t (t=1,2,3,…,6), each LGMD takes two adjacent images as input signals and outputs a set of target motion feature vectors.

[0082] In some embodiments, step 1202 specifically includes:

[0083] Step 12021, for each LGMD, calculate the absolute value of the difference between each pixel in the two input images;

[0084] Step 12022, calculating the output values ​​of the excitatory cells and the inhibitory cells based on the absolute value of the difference between each pixel in the two images;

[0085] Step 12023, calculating the cell membrane potential based on the output values ​​of the excitatory cells and the inhibitory cells;

[0086] Step 12024, using low-pass filtering to process the cell membrane potential, and filtering out the cell membrane potential whose potential is lower than a preset threshold;

[0087] Step 12025, calculate the motion feature vector based on the filtered cell membrane potential.

[0088] Figure 4 FIG. 1 is a flow chart of obtaining motion feature vectors output by six LGMDs according to an embodiment of the present application. Figure 4 As shown, firstly, the preprocessed image I t and I t+1 Enter LGMD t ; Then execute LGMD t , output 1×6 motion feature vector FeatureVector(t); then store LGMD t Calculated F t (x,y) is denoted as Set t = t + 1, and then repeat the above steps until t = 7. t The execution process is as follows:

[0089] Step 1a: Calculate image I t and I t+1 The absolute value of the difference between each pixel in:

[0090] F t (x,y)=|I t+1 (x,y)-I t (x,y)|

[0091] Step 2a, calculate the output values ​​of excitatory cells and inhibitory cells:

[0092] E(x,y)=F t (x,y)

[0093]

[0094] Among them, E(x,y) is the output value of the excitatory cell, and I(x,y) is the output value of the inhibitory cell. For the previous LGMD t The F obtained at the step t (x,y), the default value is 0 when executed for the first time, ω(i,j) is the suppression aggregation coefficient matrix of the local suppression weight.

[0095] Optionally,

[0096] Step 3a, calculate cell membrane potential:

[0097] S(x,y)=E(x,y)-I(x,y)W

[0098] Where W is the suppression weight

[0099] Step 4a: Low-pass filter the calculated membrane potential:

[0100]

[0101] where ω e is a convolution kernel with spatial low-pass filtering.

[0102] Optionally,

[0103] Step 5a: Set the threshold T G , filter the weaker excitement point G, that is:

[0104]

[0105] in is the membrane potential after filtering, and C is the attenuation coefficient of [0,1].

[0106] Step 6a. Calculate LGMD t The motion feature vector of is:

[0107] Figure 5 6 feature extraction area division diagrams of LGMD provided in the embodiment of the present application, such as Figure 5 As shown, the input image I t Divided into 6 areasA i (i=1,2,3,…,6), for each region A i Membrane potential in The total membrane potential in the region is calculated by summing up the following formula as the A of the region i Movement characteristics:

[0108]

[0109] From 6 areas A i The calculated six characteristic quantities are combined to obtain LGMD t The output 1×6 motion feature vector is:

[0110] FeatureVector(t)=[Feature(t,i)] 1×6 ,i=1,2,3,…,6

[0111] In some embodiments, step 130 specifically includes:

[0112] Step 1301, combine the six groups of motion feature vectors to generate an LGMD feature map of the target object.

[0113] Specifically, the LGMD feature map is a 6×6 matrix FeatureMap, which is normalized and the value range is mapped to [0, 255]. The value of each element in the matrix is:

[0114]

[0115] In some embodiments, after step 110, the method further includes:

[0116] Based on the preprocessed images, adjust the sampling rate of the target object video. Optionally, the sampling rate of the video acquisition device should be adjusted according to the target's movement speed. If the sampling rate is too high, the two adjacent images are highly similar, the target movement changes are not obvious, and it is difficult to extract the details of the motion features; if the sampling rate is too low, the two adjacent images are very different, the target movement changes too fast, and it is also difficult to extract the details of the motion features. Therefore, it is necessary to adjust the sampling rate of the video acquisition device according to the difference of the preprocessed images to ensure that the preprocessed images can extract the details of the motion features.

[0117] Optionally, according to the difference of the preprocessed images, the sampling rate of the video acquisition device is adjusted using the following three conditions:

[0118] 1b. If the pixel values ​​of the same pixel in the two images differ by more than 5%, it is considered that the pixel has changed.

[0119] 2b. When the target moves quickly, the number of pixels that change between two adjacent images should not be less than 5% and not more than 10% of the total number of image pixels. If it is less than 5%, the sampling rate should be reduced; if it is more than 10%, the sampling rate should be increased.

[0120] 3b. When the target moves slowly, the number of pixels that change between the first preprocessed image I1 and the seventh preprocessed image I7 is not less than 10% of the total number of image pixels and not more than 30% of the total number of image pixels. If it is less than 10%, the sampling rate is reduced; if it is more than 30%, the sampling rate is increased.

[0121] In some embodiments, the convolutional neural network includes 3 convolutional layers, 2 pooling layers and 1 fully connected layer. A 6×6 target dynamic feature map FeatureMap is input as a sample into the convolutional neural network.

[0122] Referring to the characteristics of the simple structure and few neurons of the visual nervous system of insects, this convolutional neural network adopts a simple structure constructed by 3 convolutional layers, 2 pooling layers, and 1 fully connected layer. The input layer of the convolutional neural network receives the target motion features, and each input node corresponds to a normalized feature FeatureMap(t,i) of LGMD. The convolutional layer is the core of the convolutional neural network, which contains the features of multiple convolution analysis inputs. Through the convolution operation, the convolution kernel slides on the input features to gradually extract local feature information. The ReLU activation function is added after the convolutional layer to introduce nonlinearity and help the network learn more complex patterns. The role of the pooling layer is to reduce the size of the feature map and the computational complexity, and improve the robustness of the model. The subsequent fully connected layer combines the feature maps extracted by the convolutional layer and the pooling layer. The output layer uses the softmax activation function to generate the probability distribution of each category recognition result, and outputs the target type with the highest probability as the target recognition result.

[0123] Combine the following Figure 6 The target recognition device of bionic insect vision provided by the present invention is described. The target recognition device of bionic insect vision described below and the target recognition method of bionic insect vision described above can be referred to each other.

[0124] Figure 6 is a schematic diagram of the structure of a target recognition device of bionic insect vision provided in an embodiment of the present application, such as Figure 6 As shown, the device includes a collection module 610, a preprocessing module 620, an extraction module 630, an acquisition module 640 and an identification module 650, wherein:

[0125] The acquisition module 610 is used to acquire a video of the target object and extract 7 consecutive images in the video;

[0126] A preprocessing module 620 is used to preprocess the seven consecutive images;

[0127] An extraction module 630 is used to input the preprocessed 7 consecutive images into a lobular giant motion detector LGMD array to obtain 6 groups of motion feature vectors;

[0128] An acquisition module 640 is used to obtain a LGMD feature map of the target object based on the six groups of motion feature vectors;

[0129] The recognition module 650 is used to input the LGMD feature map of the target object into the convolutional neural network to obtain the recognition result of the target object.

[0130] It should be understood that the above-mentioned device is used to execute the method in the above-mentioned embodiment. The implementation principle and technical effect of the corresponding program module in the device are similar to those described in the above-mentioned method. The working process of the device can refer to the corresponding process in the above-mentioned method, which will not be repeated here.

[0131] Based on the method in the above embodiment, Figure 7 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 7 As shown, an embodiment of the present application provides an electronic device, which may include: a processor 710, a communication interface 720, a memory 730, and a communication bus 740, wherein the processor 710, the communication interface 720, and the memory 730 communicate with each other through the communication bus 740. The processor 710 may call the logic instructions in the memory 730 to execute the target recognition method of the bionic insect vision in the above embodiment.

[0132] In addition, the logic instructions in the above-mentioned memory 730 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the target recognition method of bionic insect vision described in each embodiment of the present application.

[0133] Based on the method in the above embodiment, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program runs on a processor, the processor executes the target recognition method of bionic insect vision in the above embodiment.

[0134] Based on the method in the above embodiment, an embodiment of the present application provides a computer program product. When the computer program product runs on a processor, the processor executes the target recognition method of bionic insect vision in the above embodiment.

[0135] It is understandable that the processor in the embodiment of the present application can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. The general-purpose processor can be a microprocessor or any conventional processor.

[0136] The method steps in the embodiments of the present application can be implemented by hardware or by a processor executing software instructions. The software instructions can be composed of corresponding software modules, and the software modules can be stored in random access memory (RAM), flash memory, read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), registers, hard disks, mobile hard disks, CD-ROMs, or any other form of storage medium known in the art. An exemplary storage medium is coupled to a processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in an ASIC.

[0137] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted through the computer-readable storage medium. The computer instructions may be transmitted from a website site, a computer, a server or a data center to another website site, a computer, a server or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or a data center that includes one or more available media integrated. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid state disk (SSD)), etc.

[0138] It should be understood that the various numerical numbers involved in the embodiments of the present application are only used for the convenience of description and are not used to limit the scope of the embodiments of the present application.

[0139] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A target recognition method based on bionic insect vision, characterized in that: include: Collecting a video of a target object and extracting 7 consecutive images from the video; Preprocessing the 7 consecutive images; The preprocessed 7 consecutive images were input into the lobular giant motion detector LGMD array to obtain 6 groups of motion feature vectors; Based on the six groups of motion feature vectors, obtaining a LGMD feature map of the target object; The LGMD feature map of the target object is input into a convolutional neural network to obtain a recognition result of the target object.

2. The target recognition method of bionic insect vision according to claim 1, characterized in that: The preprocessing of the seven consecutive images comprises: Taking the origin of the image as the center point, cropping the image into a hexagonal image consistent with the size of the LGMD; The hexagonal image is preprocessed.

3. The target recognition method of bionic insect vision according to claim 1, characterized in that: The seven pre-processed continuous images are input into the lobule giant motion detector LGMD array to obtain six groups of motion feature vectors, including: Taking each two adjacent images of the seven consecutive images as input, and inputting them into the six LGMDs of the LGMD array respectively; The motion feature vectors respectively output by the six LGMDs are obtained.

4. The target recognition method of bionic insect vision according to claim 3, characterized in that: The obtaining of motion feature vectors respectively output by the six LGMDs comprises: For each LGMD, the absolute value of the difference between each pixel in the two input images is calculated; Calculating the output values ​​of the excitatory cells and the inhibitory cells based on the absolute value of the difference between each pixel in the two images; Calculating cell membrane potential based on the output values ​​of the excitatory cells and the inhibitory cells; Processing the cell membrane potential using a low-pass filter, and filtering out cell membrane potentials whose potentials are lower than a preset threshold; Based on the filtered cell membrane potential, the motion feature vector is calculated.

5. The target recognition method of bionic insect vision according to claim 1, characterized in that: The step of obtaining the LGMD feature map of the target object based on the six groups of motion feature vectors includes: The six groups of motion feature vectors are combined to generate an LGMD feature map of the target object.

6. The target recognition method of bionic insect vision according to claim 1 or 2, characterized in that: After preprocessing the seven consecutive images, the method further includes: Based on the preprocessed image, the sampling rate of the target object video is adjusted.

7. The target recognition method of bionic insect vision according to claim 1, characterized in that: The convolutional neural network includes 3 convolutional layers, 2 pooling layers and 1 fully connected layer.

8. A target recognition device based on bionic insect vision, characterized in that: include: An acquisition module, used to acquire a video of a target object and extract 7 consecutive images from the video; A preprocessing module, used for preprocessing the seven consecutive images; An extraction module is used to input the preprocessed 7 consecutive images into the lobular giant motion detector LGMD array to obtain 6 groups of motion feature vectors; An acquisition module, used for acquiring a LGMD feature map of the target object based on the six groups of motion feature vectors; The recognition module is used to input the LGMD feature map of the target object into a convolutional neural network to obtain a recognition result of the target object.

9. An electronic device, characterized in that: include: at least one memory for storing a computer program; At least one processor is used to execute the program stored in the memory. When the program stored in the memory is executed, the processor is used to execute the target recognition method of bionic insect vision as described in any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program runs on a processor, the processor is enabled to execute the target recognition method of bionic insect vision as described in any one of claims 1 to 7.

Images