Wild animal channel parameter determination method based on image recognition

Abstract

The invention relates to the technical field of traffic engineering and wild animal protection, in particular to a wild animal channel parameter determination method based on image recognition. In the process of designing a wild animal channel, firstly, wild animal monitoring images are collected, monitoring data are input into a first image recognition model, wild animal segmented images and wild animal species are output, then information value recognition is conducted on the wild animal segmented images, and whether a second image recognition model is entered or not is judged accordingly; therefore, image recognition efficiency is improved while wild animal recognition accuracy is improved, and accurate and efficient determination of wild animal channel design is facilitated; meanwhile, the invention provides a brand-new method for determining the number of the convolutional layers of the scale feature block, that is, the number of the convolutional layers is determined according to the information value, so that the accuracy of identifying wild animal species is greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical fields of traffic engineering and wildlife protection, and particularly to a method for determining wild animal passage parameters based on image recognition. Background Art

[0002] In the prior art, when designing wild animal passages according to monitoring data, there is a problem of low recognition accuracy for incomplete monitoring data, resulting in a lack of scientific nature in the wild animal passage design method and reducing the usage efficiency of the passages. Summary of the Invention

[0003] To solve the above technical problems, the present invention provides a method for determining wild animal passage parameters based on image recognition to solve the problems existing in the prior art.

[0004] The present invention provides a method for determining wild animal passage parameters based on image recognition, including the following steps: S1: Obtain the monitoring data of the designed passage system and the net height and net width data of highway bridges and culverts; S2: Perform image recognition on the monitoring data to identify wild animal species and population distribution; Specifically, S2 is as follows: S2.1: Input the monitoring data into the first image recognition model to output a wild animal segmentation image and wild animal species; S2.2: Perform information value recognition on the wild animal segmentation image. If the information value is higher than a preset threshold, use the wild animal species obtained in S2.1 as the final wild animal species; otherwise, proceed to the next step. Here, the information value is a value obtained based on the wild animal segmentation image and a standard wild animal image; S2.3: Input the monitoring data into the second image recognition model to output the wild animal species as the final wild animal species; S3: Determine the distribution data of wild animals according to the final wild animal species; S4: Determine the parameters of the wild animal passage according to the distribution data of the wild animals.

[0005] Preferably, in S2.2, the information value is obtained by: determining a standard wild animal image according to the wild animal species and wild animal segmentation image recognized in S2.1, comparing the similarity between the wild animal segmentation image and the standard wild animal image, and using the similarity value as the magnitude of the information value.

[0006] Preferably, the standard wild animal image determined according to the wild animal species identified in S2.1 and the wild animal segmentation image is specifically: select a wild animal image that is the same as the identified wild animal species and has the same shooting angle as the wild animal segmentation image as the standard wild animal image; Comparing the similarity between the wild animal segmentation image and the standard wild animal image, and taking the similarity value as the magnitude of the information value is specifically: align the key parts in the wild animal segmentation image with the standard wild animal image, then enlarge or reduce the wild animal segmentation image to have the highest degree of repetition with the standard wild animal image, and then calculate the similarity between the wild animal segmentation image and the standard wild animal image.

[0007] Preferably, the preset threshold is 80%.

[0008] Preferably, the second image recognition model is the SSD model.

[0009] Preferably, the SSD model includes a pre-trained network and a multi-scale feature block; The pre-trained network is a VGG model, and the multi-scale feature block is composed of multiple convolutional layers; in this step, the method for determining the number of convolutional layers of the multi-scale feature block is:

[0010] In the formula, N is the number of convolutional layers of the multi-scale feature block, round() is the rounding function, S is the similarity between the wild animal segmentation image and the standard wild animal image, and a, b are coefficients, where a = 7 and b > 0.

[0011] Preferably, S4 is specifically: according to the distribution data of protected wild animals and the net height and net width data of highway bridges and culverts, select the bridges and culverts that meet the requirements to determine the location of the animal passage, and determine the wild animal passage parameters.

[0012] Preferably, the passage parameters include: passage size, passage bottom material.

[0013] The embodiments of the present invention have the following technical effects: In the process of designing the wild animal passage in the present invention, first, wild animal monitoring images are collected, and the monitoring data is input into the first image recognition model to output the wild animal segmentation image and the wild animal species. Then, the information value of the wild animal segmentation image is recognized, and based on this, it is judged whether to enter the second image recognition model. In this way, while improving the accuracy of wild animal recognition, the image recognition efficiency is improved, which is beneficial to the accurate and efficient determination of the wild animal passage design; Meanwhile, when the present invention uses the SSD model to identify wildlife images that are not fully captured, the number of convolutional layers of the multi-scale feature blocks in the SSD model is determined according to the information value of the monitoring image, thereby greatly improving the accuracy of the multi-scale feature blocks in identifying wildlife species, and further improving the passing capacity of wildlife passages. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0015] Figure 1 is a flowchart of a method for determining wildlife passage parameters based on image recognition provided by an embodiment of the present invention; Figure 2 is a flowchart of image recognition of monitoring data to identify wildlife species and population distribution provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0016] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention.

[0017] Embodiment 1, as Figure 1 shown in a flowchart of a method for determining wildlife passage parameters based on image recognition. As Figure 1 shown, a method for determining wildlife passage parameters based on image recognition includes the following steps: S1: Obtain the monitoring data of the designed passage system and the net height and net width data of highway bridges and culverts; In this step, the wildlife activity area is obtained through regional wildlife research materials, and then the monitoring data of the wildlife passage design system such as wildlife species and population distribution is obtained by installing infrared cameras or using drones in the wildlife activity area; Infrared cameras mainly detect the temperature difference between the animal's body and the surrounding environment through a passive infrared sensor (PIR). When an animal enters the monitoring area, the sensor will detect this temperature difference change and trigger the camera to take pictures or videos. This triggering method enables the infrared camera to automatically record the animal's activities without disturbing the animal; Among them, the model of the infrared camera is Boli BG636-48K. This infrared camera is a high-definition infrared camera designed specifically for wildlife monitoring. It has characteristics such as high pixels, long standby time, and multiple functions, and is suitable for high-definition image shooting and wildlife monitoring. Its technical parameters are as follows: Image sensor: 16 million pixel CMOS; Photo pixel: Supports up to 48 million pixels, and 36 million, 24 million, or 12 million pixels can be selected; Video resolution: Supports 1920×1080 (1080P), 1280×720, and 640×480; Lens parameters: FOV = 70°, F / NO = 2.2; Infrared LED lights: 940nm invisible infrared rays, 4 high-power infrared LED lights; Illumination and detection distance: 30-meter ultra-long night vision detection distance; Display screen: Built-in 2.0-inch color screen; Memory card: Supports SD cards up to 64GB; Battery: Supports 5 18650 lithium batteries or 4 - 8 AA batteries; Operating temperature: -20°C to +70°C; Dust and waterproof level: IP68; Dimensions: 147×90×145mm; Among them, the unmanned aerial vehicle (UAV) can be a fixed-wing UAV or a multi-rotor UAV. Fixed-wing UAVs are suitable for large-scale wildlife monitoring, and multi-rotor UAVs are suitable for small-scale wildlife monitoring.

[0018] S2: Perform image recognition on the monitoring data to identify wildlife species and population distributions; In this step, perform image recognition on the monitoring data obtained by the infrared camera to obtain wildlife species information, and perform image recognition on the monitoring data obtained by the UAV to obtain wildlife population information; In wildlife monitoring image recognition, since there are cases where wildlife monitoring images are not fully captured, this leads to errors in the identified wildlife species and population distributions. Therefore, based on this, this embodiment proposes a new image recognition method for wildlife to improve the recognition accuracy of wildlife images that are not fully captured; Specifically, Figure 2 shows the flowchart of performing image recognition on the monitoring data to identify wildlife species and population distributions. As shown in Figure 2 shown, the specific content of S2 is as follows: S2.1: Input the monitoring data into the first image recognition model to output the wildlife segmentation image and the wildlife species; Among them, the first image recognition model is a convolutional neural network model (Convolutional Neural Network, CNN). A convolutional neural network is a deep learning model specifically used to process image and video data. Its core idea is to extract local features of the image through convolutional operations and gradually combine these features through a multi-layer network structure to finally achieve tasks such as detection or segmentation.

[0019] The convolutional neural network model includes an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer; among them, the input layer is used to input the image monitoring data to be recognized, the convolutional layer is used to extract the features of the image monitoring data to be recognized, and its principle is to slide a convolutional kernel (filter) on the image to calculate the weighted sum of the local area and generate a feature map (Feature Map); the pooling layer is used to reduce the size of the feature map, reduce the amount of calculation, and at the same time increase the robustness of the model; in the convolutional neural network model of this embodiment, the pooling layer adopts the maximum pooling method (Max Pooling), that is, taking the maximum value of the local area, the fully connected layer is used to combine the features extracted by the convolutional layer and the pooling layer, and the output layer is used to output the wildlife segmentation image and the wildlife species.

[0020] S2.2: Perform information value recognition on the wildlife segmentation image. If the information value is higher than the preset threshold, then use the wildlife species obtained in S2.1 as the final wildlife species, otherwise go to the next step, where the information value is a value obtained based on the wildlife segmentation image and the standard wildlife image; Among them, the acquisition method of the information value is: determine the standard wildlife image according to the wildlife species and the wildlife segmentation image recognized in S2.1, compare the similarity between the wildlife segmentation image and the standard wildlife image, and use the similarity value as the size of the information value; Specifically, determining the standard wildlife image according to the wildlife species and the wildlife segmentation image recognized in S2.1 is specifically: select a wildlife image with the same wildlife species as the recognized one and the same shooting angle as the wildlife segmentation image as the standard wildlife image; Specifically, comparing the similarity between the wildlife segmentation image and the standard wildlife image and using the similarity value as the size of the information value is specifically: align the key parts in the wildlife segmentation image with the standard wildlife image, then enlarge or reduce the wildlife segmentation image to have the highest repeatability with the standard wildlife image, and then calculate the similarity between the wildlife segmentation image and the standard wildlife image; Among them, according to the statistics of image occlusion, the key parts are the head, hooves, and tail; In this step, the preset threshold is 80%.

[0021] S2.3: Input the monitoring data into the second image recognition model, and output the wild animal species as the final wild animal species; Among them, in the above steps, if the wild animal monitoring image is not fully captured, it may lead to inaccurate identification of the wild animal species by the model. Therefore, in this step, the second image recognition model is used to identify the wild animal species; Among them, the second image recognition model is the SSD model. SSD (Single Shot MultiBox Detector) is a high-accuracy object detection algorithm. Due to its relatively complex model design, it can achieve more accurate detection results; Among them, the SSD model includes a pre-trained network and a multi-scale feature block; Among them, the pre-trained network is the VGG model, and the multi-scale feature block is composed of multiple convolutional layers; in this step, the method for determining the number of convolutional layers of the multi-scale feature block is as follows:

[0022] In the formula, N is the number of convolutional layers of the multi-scale feature block, round() is the rounding function, S is the similarity between the wild animal segmentation image and the standard wild animal image, and a and b are coefficients, where a = 7 and b > 0; In this step, when using the SSD model to identify wild animal images that are not fully captured, the number of convolutional layers of the multi-scale feature block in the SSD model is determined according to the information value of the monitoring image, thereby greatly improving the accuracy of identifying wild animal species by the scale feature block.

[0023] At the same time, in this embodiment, the monitoring image is first input into the convolutional neural network model, and the wild animal segmentation image and the wild animal type are output. Then, the information value of the wild animal segmentation image is identified, and based on this, it is judged whether to enter the second image recognition model. That is, the wild animal image is first input into a relatively simple image recognition model, and then it is judged whether to input into the complex image recognition model according to the information value. Compared with directly calling the complex image recognition model at the beginning, while ensuring the accuracy of wild animal recognition, the image recognition efficiency is improved, which is beneficial to the accurate and efficient determination of the wild animal passage design.

[0024] S3: Determine the distribution data of wild animals according to the final wild animal species; In this step, the identification situation of key protected wild animals identified within a period of time is counted to obtain the distribution data of wild animals.

[0025] S4: Determine the parameters of the wildlife passage according to the distribution data of the wild animals; Among them, the specific content of S4 is: According to the distribution data of protected wild animals and the existing highway bridge and culvert clear height and clear width data, select the bridge and culvert that meet the requirements and determine them as the optimal site for the animal passage, and determine the wildlife passage parameters to make them integrate with the surrounding ecological environment; Specifically, the passage parameters include: passage location, structural parameters and construction quantity. Among them, the structural parameters include passage size, passage bottom material, etc.; Among them, in terms of the passage size, for the passage of amphibians and reptiles, the diameter or height-width should not be less than 0.3 m and not exceed 0.6 m, or for large-sized animals, the passage size is increased; In this embodiment, when designing the wildlife passage, the size of the wildlife passage is determined according to the type of the wildlife passage. Among them, the types of wildlife passages can be divided into overpass-type passages and underpass-type passages for mammals. Table 1 shows the recommended values of the sizes of overpass-type wildlife passages. When determining the size of the wildlife passage, Table 1 can also be used to determine the size of the wildlife passage; Table 1 Recommended values of the sizes of overpass-type wildlife passages

[0026] The underpass-type passages for mammals are divided into large-sized mammal passages and medium and small-sized mammal passages. The large-sized mammal passages are further divided into large-sized herbivores (mainly Tibetan wild asses) and large-sized carnivores (wolves, foxes, bears, etc.). Referring to foreign experience, the recommended values of the sizes of underpass-type passages for mammals are proposed (Table 2); Table 2 Recommended values of the sizes of underpass-type wildlife passages

[0027] In terms of the passage bottom material, single cement concrete should be avoided as much as possible, and river sand, gravel or mixed sand and gravel should be used as the bottom material to increase the passing probability of wild animals.

[0028] In terms of the passage location, through the above step S3, the passage location is determined after understanding the activity range, migration path and habitat of the target species; In terms of the construction quantity, mainly through the constraint of the corridor density, the density of wildlife corridors will determine whether the corridors can effectively maintain the connectivity of habitats. The factors affecting the corridor density mainly depend on the species and characteristics of wildlife in the area and the habitat type of the area. If the road passes through an area with dense wildlife distribution, the corridor density needs to be increased. For some wildlife distribution areas with a wide distribution range and relatively small population numbers, the corridor density can be appropriately reduced.

[0029] Embodiment 2. The present invention also provides an electronic device, including one or more processors and a memory.

[0030] The processor may be a central processing unit (CPU) or other forms of processing units with data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device to perform desired functions.

[0031] The memory may include one or more computer program products, and the computer program products may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory, etc. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage media, and the processor may run the program instructions to implement a method for determining wildlife corridor parameters based on image recognition according to any embodiment of the present application described above and / or other desired functions. Various contents such as initial external parameters, thresholds, etc. may also be stored in the computer-readable storage media.

[0032] In one example, the electronic device may further include: an input device and an output device, and these components are interconnected through a bus system and / or other forms of connection mechanisms (not shown). The input device may include, for example, a keyboard, a mouse, etc. The output device may output various information to the outside, including early warning prompt information, braking force, etc. The output device may include, for example, a display, a speaker, a printer, and a communication network and its connected remote output devices, etc.

[0033] Of course, for simplicity, components such as buses, input / output interfaces, etc. are omitted. In addition, according to specific application scenarios, the electronic device may further include any other appropriate components.

[0034] In addition to the above methods and devices, an embodiment of the present application may also be a computer program product, which includes computer program instructions. When the computer program instructions are run by a processor, the processor is caused to implement the functions of a method for determining wild animal passage parameters based on image recognition provided by any embodiment of the present application.

[0035] The computer program product may be written in any combination of one or more programming languages for programming code to perform the operations of the embodiments of the present application. The programming languages include object-oriented programming languages such as Java, C++, etc., and also include conventional procedural programming languages such as the "C" language or similar programming languages. The programming code may be executed entirely on the user computing device, partially on the user device, executed as an independent software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0036] In addition, an embodiment of the present application may also be a computer-readable storage medium, on which computer program instructions are stored. When the computer program instructions are run by a processor, the processor is caused to implement a method for determining wild animal passage parameters based on image recognition provided by any embodiment of the present application.

[0037] The computer-readable storage medium may adopt any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may, for example, include but is not limited to an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples (non-exhaustive list) of the readable storage medium include: an electrical connection having one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0038] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.

Claims

1. A method for determining wildlife passage parameters based on image recognition, characterized in that: The following steps are involved: S1: Obtain monitoring data of the designed channel system and clear height and width data of highway bridges and culverts; S2: performing image recognition on the monitoring data to identify wild animal species and population distribution; The S2 is specifically: S2.1: Input the monitoring data into a first image recognition model, and output a wildlife segmentation image and a wildlife species; S2.2: Identify the information value of the wildlife segmented image. If the information value is higher than a preset threshold, the wildlife species obtained in S2.1 is used as the final wildlife species, otherwise proceed to the next step; wherein the information value is a value obtained based on the wildlife segmented image and the standard wildlife image; S2.3: inputting the monitoring data into a second image recognition model, and outputting the wildlife species as the final wildlife species; S3: Determine wildlife distribution data based on the final wildlife species; S4: Determine wildlife passage parameters based on the wildlife distribution data.

2. The method for determining wildlife passage parameters based on image recognition according to claim 1, characterized in that: In S2.2, the information value is obtained by determining a standard wildlife image based on the wildlife species identified in S2.1 and the wildlife segmentation image, and comparing the wildlife segmentation image with the standard wildlife image for similarity, and using the similarity value as the size of the information value.

3. The method for determining wildlife passage parameters based on image recognition according to claim 2, characterized in that: Determining the standard wildlife image based on the wildlife species identified in S2.1 and the wildlife segmentation image specifically includes: selecting a wildlife image that is the same as the identified wildlife species and shot at the same angle as the wildlife segmentation image as the standard wildlife image.

4. The method for determining wildlife passage parameters based on image recognition according to claim 2, characterized in that: The wildlife segmentation image is compared with the standard wildlife image for similarity, and the similarity value is used as the size of the information value. Specifically, the key parts in the wildlife segmentation image are aligned with the standard wildlife image, and then the wildlife segmentation image is enlarged or reduced to the highest degree of repetition with the standard wildlife image, and then the similarity value between the wildlife segmentation image and the standard wildlife image is calculated.

5. The method for determining wildlife passage parameters based on image recognition according to claim 1, characterized in that: In S2.2, the preset threshold is 80%.

6. The method for determining wildlife passage parameters based on image recognition according to claim 2, characterized in that: The second image recognition model is an SSD model.

7. The method for determining wildlife passage parameters based on image recognition according to claim 6, characterized in that: The SSD model includes a pre-trained network and a multi-scale feature block; The pre-trained network is a VGG model, and the multi-scale feature block is composed of multiple convolutional layers; the method for determining the number of convolutional layers of the multi-scale feature block is: ； Where N is the number of convolutional layers of the multi-scale feature block, round() is the rounding function, S is the similarity value between the wildlife segmentation image and the standard wildlife image, a and b are coefficients, where a=7, b>0.

8. The method for determining wildlife passage parameters based on image recognition according to claim 1, characterized in that: The S4 is specifically as follows: based on the distribution data of the wild animals and the clear height and clear width data of the highway bridge and culvert, a highway bridge and culvert that meets the requirements is selected as the site selection for the animal passage, and the parameters of the wildlife passage are determined.

9. The method for determining wildlife passage parameters based on image recognition according to claim 8, characterized in that: The channel parameters include: channel size and channel bottom quality.

Images