TWI826777B - Row-crop type unmanned vehicle automatic navigation system and method thereof - Google Patents

Abstract

A row-crop type unmanned vehicle automatic navigation system and method thereof are provided. The method, applied to a mobile vehicle, includes steps of: using at least one image capturing device to capture a plurality of road images, generating a plurality of labels on the road images to identify the terrain illustrated in the road images, establishing an inference model, and then, with at least one semantic segmentation artificial intelligence algorithm and the road images having the labels, training the inference model to output a terrain recognizing result. Based on the terrain recognizing result, a set of trajectory information suitable for automatically navigating the mobile vehicle may be calculated.

Description

Qitian type unmanned vehicle automatic navigation system and method thereof

The present invention relates to an automatic navigation system, in particular to a border type unmanned vehicle automatic navigation system and a method thereof.

As shown in Figure 2, the border-style farmland 10 has a raised border 12 and a concave border 14. Crops are planted on the border 12, while farmers walk, use agricultural machinery, and irrigate on the border 14. Currently, tasks such as spreading seeds, weeding, fertilizing, etc. all need to be done manually. Subsequently, various agricultural machinery have been developed to replace purely manual work, with humans driving the agricultural machinery to save labor burden.

Although there are agricultural machinery that can replace human labor on foot, human driving is still required. Even if there are agricultural machinery in vast fields (such as soybean farms), farmers still need to perform long periods of physical labor under the scorching sun. If only one person can drive agricultural machinery, each stage of farming can only be undertaken by a single person, which cannot improve productivity and farm management efficiency. Moreover, long-term manpower expenditure may lead to fatigue driving, concentration disorders, and safety problems. Therefore, automating agricultural machinery can significantly improve agricultural productivity and management efficiency.

In order to improve agricultural productivity and management efficiency, Xinzhi Technology has proposed patents related to agricultural machinery automation. For example, the National Patent No. 202030497 "Navigation System and Operation Method of Navigation System" uses a radar signal to be emitted, and the distance between the local area and the object is calculated based on the time difference between the radar signal being reflected by the object, and obstacles are detected. Object detection to obtain the distribution information of the location of surrounding objects. However, in actual agricultural applications, both radar devices and navigation devices are expensive. Therefore, if the radar device is installed on agricultural machinery, the cost will increase significantly. Another example is the "Compound Guided Work Vehicle" patented by the country's patent No. I389635, which uses predetermined tracks to move the vehicle to meet different task requirements. However, this method can only allow the work vehicle to walk on a predetermined physical track. If you want to use the work vehicle in a new place, you must reinstall the track, which is quite inconvenient. Also, the track is an additional cost.

Therefore, in view of the shortcomings of the above-mentioned conventional technologies and future needs, the present invention proposes a border-type unmanned vehicle automatic navigation system and its method, which uses an unmanned vehicle to navigate forward autonomously. The specific architecture and implementation will be described in detail below. :

The main purpose of the present invention is to provide a field-type unmanned vehicle automatic navigation system and its method, which utilizes trained artificial intelligence models to identify the terrain of farmland, distinguish between fields and ditches, and calculate appropriate movement trajectories to provide to the mobile vehicle. , for mobile vehicles to automatically navigate forward and reduce farmers’ physical labor.

Another object of the present invention is to provide a field-type unmanned vehicle automatic navigation system and its method, which uses a small number of road images under various environmental conditions (dry and wet soil, light intensity, etc.) and various growth stages of plants. The trained inference model combined with the navigation information algorithm can complete the automatic navigation of the mobile vehicle, and the navigation effect will not be affected by factors such as sunlight intensity, plant size, soil conditions, etc.

In order to achieve the above object, the present invention provides a border type unmanned vehicle automatic navigation system, which is installed on a mobile vehicle and includes: at least one image capture device to capture multiple road surface images; a label analysis module, Connect the image capture device to generate multiple annotations on the road surface image to identify the terrain in the road surface image; a semantic segmentation module connects the annotation analysis module and the at least one image capture device to establish an inference model and use at least one The semantic segmentation artificial intelligence algorithm is trained with labeled road surface images to generate an inference model to output a terrain recognition result; a travel information module is connected to the semantic segmentation module to calculate a set of trajectories based on the terrain recognition results. information; and a travel control module that connects the travel information module and the mobile vehicle, receives at least part of the trajectory information, and controls the movement of the mobile vehicle according to the trajectory information.

According to the embodiment of the present invention, an image processing module is further included. The signal is connected to the image capture device, the annotation analysis module and the semantic segmentation module. It receives the road surface image transmitted by the image capture device and performs pre-processing before transmitting it to Annotation analysis module or semantic segmentation module.

According to embodiments of the present invention, the preprocessing performed by the image processing module includes: adjusting the size of the road surface image, adjusting the color representation value of the road surface image, sampling the road surface image, or a combination of the above three.

According to embodiments of the present invention, the terrain recognition results include a plurality of terrain blocks on the road surface images, and the terrain blocks include border blocks and border blocks.

According to the embodiment of the present invention, the trajectory information includes the wheel trajectory, center of mass trajectory, traveling linear speed and angular velocity of the mobile vehicle. The linear speed is preset, and the traveling information module obtains the parameters of both sides of the ditch block. The center points are connected in a line to generate the wheel trajectory of the mobile vehicle. The center points of the two wheel trajectories are taken and connected in a line segment as the centroid trajectory of the mobile vehicle, and calculated based on the offset of the centroid trajectory. The angular velocity of the moving vehicle.

According to an embodiment of the present invention, the traveling information module transmits the linear velocity and angular velocity in the set of trajectory information to the traveling control module.

According to an embodiment of the present invention, a visualization module is further included, which is connected with the semantic segmentation module to display terrain regions. Then it is connected to the travel information module to display the wheel trajectory and center of mass trajectory on the terrain block.

According to embodiments of the present invention, the visualization module displays trajectory information in different colors, color blocks, and lines.

The invention also provides a borderland-type unmanned vehicle automatic navigation method, which is applied to a borderland-type unmanned vehicle automatic navigation system installed on a mobile vehicle, including the following steps: using at least one image capture device to capture a plurality of images The road surface image is sent to a label analysis module to generate plural labels on the road surface image to identify the terrain in the road surface image; a semantic segmentation module is used to receive the road surface image containing the label, and establish an inference model using at least one The semantic segmentation artificial intelligence algorithm and the received road surface image train the inference model to output a terrain recognition result; and use a travel information module to calculate a set of trajectory information suitable for the movement of the mobile vehicle based on the inference model. And sent to a travel control module to control the movement of the mobile vehicle according to the set of trajectory information.

The present invention provides a border-type unmanned vehicle automatic navigation system and a method thereof, which are applied to the traveling mode of unmanned agricultural tools on border-style farmland. By collecting a small amount of border-field agricultural field images, the border and fields can be detected and distinguished. Qigou has also trained an AI inference model that can identify terrain, and provides clear detection and positioning of gullies to provide trajectories and navigation for mobile vehicles (such as unmanned agricultural tools) moving between gullies.

Please refer to Figure 3 , which is a schematic diagram of the mobile vehicle 16 moving on the farmland 10 . The farmland 10 is a long strip of fields 12, and sunken furrows 14 are provided between the fields 12 for walking or water intake for irrigation. The mobile vehicle 16 is a four-wheeled unmanned vehicle, and its left and right wheels span at least one farmland 12 and land on two ditches 14 on both sides respectively. For example, the left and right wheels of the mobile vehicle 16 can span Two borders12. There is at least one image capture device 22 on the mobile vehicle 16, such as a camera, a video camera or any lens. One of them can be installed at the front and rear positions of the mobile vehicle 16. When the mobile vehicle 16 moves forward, the image capture device 22 in front is used. The capturing device 22 captures the road surface image, and when the mobile vehicle 16 retreats, the rear image capturing device 22 is used to capture the road surface image. And the mobile vehicle 16 is provided with a processor 24. The image capture device 22 and the processor 24 form the border type unmanned vehicle automatic navigation system 20 of the present invention. The software part of the present invention includes image processing, AI training, and trajectory calculation. etc. are all performed in the processor 24.

Please also refer to Figure 1 , which is a block diagram of the border-type unmanned vehicle automatic navigation system 20 of the present invention. The border type unmanned vehicle automatic navigation system 20 of the present invention is installed on the mobile vehicle 16 and includes an image capture device 22 and a processor 24. The image capture device 22 is used to capture road images, including the border 12 , ditch 14 and possible obstacles; the processor 24 further includes an image processing module 242, an annotation analysis module 244, a semantic segmentation module 246, a visualization module 247, and a traveling information module. 248 and a travel control module 249. The image processing module 242 is connected to the image capture device 22 to pre-process the road surface image, such as adjusting the size of the road surface image to meet subsequent needs, or capturing only a part of the area of interest in the image. In addition, preprocessing also includes adjusting the color representation values of the road image, such as adjusting the hue (Hue), saturation (Saturation), value (Value), brightness (Lightness), or the HSV or HSL color space containing the aforementioned representation values, etc. . In addition, preprocessing also includes sampling road surface images, so only a small number of images are sent to the rear annotation analysis module 244. For example, assuming that the image capture module can capture 30 road surface images per second, in the training mode Next, various conditions are taken into consideration, including differences in growth stages of various plants, light changes, etc., and several images are taken for each field. For example, several images are taken. The number of images of each type will not be too many. , increasing the breadth of sampling can lead to complete training. When the interpretation is actually performed instead of the training mode, more images are sampled, and various conditions are no longer considered. For example, 6 or 10 images are taken per second, or 700 of the total 1,000 images are taken, and the These sampled images are input into the inference model trained by the semantic segmentation module 246 to form a continuous traveling result.

The annotation analysis module 244 generates multiple annotations on the road surface image. The annotation is used to identify the terrain in the road surface image. In one embodiment, the annotation analysis module 244 generates multiple annotations based on the annotation results of the user, for example Provide a user interface (not shown in the figure), and provide "Qi Tian", "Qi Gou" or other options on the user interface for the user to choose, or provide an input device (not shown in the figure) such as a keyboard, A handwriting pad, etc., allows the user to use the input device to input annotations on the user interface, for example, input annotations such as "Qi Tian" and "Qi Ditch" in Figure 1. In one embodiment, the so-called annotation is performed by the user specifying segments or line segments on the road surface image and connecting them into an area for frame selection to generate a terrain block.

The semantic segmentation module 246 is connected to the annotation analysis module 244 to train an AI model. More specifically, the semantic segmentation module 246 first establishes an inference model, and then uses at least one semantic segmentation artificial intelligence algorithm and annotation-based artificial intelligence algorithm. Pavement images are used to train the inference model. This inference model is usually a deep learning neural network model, and then an inference model with artificial intelligence is trained. Therefore, the inference model can use artificial intelligence to perform semantic segmentation on the image, automatically identify the terrain, and Output a terrain recognition result. For example, the trained inference model can identify where are the borders and where are the borders in the road image. During the training process, the semantic segmentation module 246 can receive representative road images of crops in different growth stages (germination, flowering, maturity, etc.), different light conditions, and soil in different dry and wet conditions for training.

The traveling information module 248 is connected to the semantic segmentation module 246. The traveling information module 248 calculates a set of trajectories suitable for the traveling of the mobile vehicle 16 based on the terrain recognition results output by the semantic segmentation module 246 and combined with at least one navigation information algorithm. The information includes the wheel trajectory, center of mass trajectory, linear velocity and angular velocity of the mobile vehicle 16 . The terrain recognition results include terrain blocks on the road surface image, and the terrain blocks include border blocks and border blocks. In this embodiment, the operator of the application field presets the feasible linear speed of the mobile vehicle to travel. Then, the edges on both sides of the ditch block can be judged first, and the center points of the two edges can be found and connected in a line to generate the wheel trajectory of the mobile vehicle 16 on the ditch block. The present invention requires two steps. Two wheel tracks are generated on each ditch block. Secondly, take the center points of the two wheel trajectories and connect them into a line. This line segment is located in the middle of the two wheel trajectories, which is the centroid trajectory of the mobile vehicle 16. Generally speaking, this centroid trajectory will fall in the Qitian area. superior. Then, the angular velocity of the mobile carrier 16 is calculated based on the offset of the center of mass trajectory. In another not-shown embodiment, after the terrain blocks are generated, other conventional navigation information algorithms can be used to find the trajectory information of the mobile vehicle 16 , but the invention is not limited thereto.

After generating the linear velocity and angular velocity, the traveling information module 248 transmits all or part of the trajectory information to the traveling control module 249 , for example, only the linear velocity and the angular velocity are provided to the traveling control module 249 . The travel control module 249 is connected to the travel information module 248 and the mobile vehicle 16. After receiving part or all of the trajectory information, it automatically navigates the mobile vehicle 16 according to the trajectory information. For example, the mobile vehicle 16 has multiple wheels (not shown) and a chassis (not shown). The travel control module 249 can control the rotation speed and direction of different wheels of the mobile vehicle 16 according to the trajectory information, so as to Change the traveling direction, speed, etc. of the mobile vehicle 16.

In particular, the backend of the image processing module 242 is connected to the annotation analysis module 244 and the semantic segmentation module 246 at the same time. If it is in the training stage of the inference model, the preprocessed road image is sent to the annotation analysis module 244. Annotations are generated and provided to the semantic segmentation module 246 to train the inference model. However, if the inference model has been trained and is in the actual application stage, the annotation analysis module 244 no longer needs to provide annotations for training the inference model. At this time, the image processing module 242 will directly transmit the preprocessed road surface image to Semantic segmentation module 246 uses inference models for terrain recognition.

An embodiment is provided below to illustrate the navigation method of the Qitian-type unmanned vehicle automatic navigation system 10 of the present invention. Please also refer to Figure 2 ~ Figure 4E. First, the image capturing device 22 captures multiple road surface images, and the road surface images include border fields 12 and border ditches 14, as shown in Figure 4A. The road surface image is then pre-processed in the image processing module 242, such as adjusting the size of the road surface image, reducing or enlarging it, as shown in Figure 4B. Next, as shown in Figure 4C, the annotation analysis module 244 is used to annotate the preprocessed road surface image, and mark "edge fields" and "edge ditches". In Figure 4D, the semantic segmentation module 246 uses algorithms related to semantic segmentation to frame a plurality of terrain blocks on the road surface image, including a border block 32 and a border block 34, which respectively represent the plantable border area 12 and mobile fields. The carrier 16 has a movable furrow 14 (as shown in Figure 2). The semantic segmentation module 246 trains an inference model that can analyze the terrain based on the terrain blocks containing the annotation. After the inference model training is completed, the newly captured road surface image is preprocessed by the image processing module 242 (as shown in Section 2). 4B) can be directly sent to the semantic segmentation module 246, and the semantic segmentation module 246 uses the inference model to generate a plurality of terrain blocks on the road image, and generates and outputs the terrain recognition results as shown in Figure 4D. Then, as shown in Figure 4E, the travel information module 246 calculates a set of trajectory information suitable for the movement of the mobile vehicle based on the terrain recognition results and the algorithm related to the navigation information based on the Figure 4D, including the mobile vehicle. 16. The track of the wheel moving on the furrow block 34, the trajectory of the center of mass on the furrow block 32, the linear velocity and the angular velocity of travel.

Please refer to Figure 5, which is a schematic diagram of the present invention's calculation of wheel trajectory, center of mass trajectory, traveling linear velocity and angular velocity on the road surface image. This road surface image is a pre-processed and resized road surface image. The wheel track is taken from the center line of the ditch block 34. Therefore, points A and B are the centers of the bottom of the screen of the ditch blocks 34 on the left and right sides respectively. point. The center of mass trajectory is the center line of the two wheel trajectories, so point C is the center point of points A and B. It can be seen from Figure 5 that C is not at the center position C’ at the bottom of the screen, and there is an offset d, which represents the offset of the traveling route of the mobile vehicle 16 and needs to be corrected. Therefore, the offset d is used to calculate the angular velocity of the mobile vehicle 16 so that the bottom starting point C of the center of mass trajectory returns to the center of the screen. The mobile vehicle 16 usually moves forward at a constant speed, so the linear speed is fixed.

Figure 5 cannot show the linear velocity and angular velocity, but only needs to provide the linear velocity and angular velocity to the travel control module 249, and the travel control module 249 can control the forward direction and braking accordingly, so essentially giving the travel control module 249 The navigation information is linear velocity and angular velocity.

In one embodiment, the visualization module 247 in the present invention is connected to the semantic segmentation module 246 and the traveling information module 248 to display the terrain blocks generated by the semantic segmentation module 246, and on the terrain blocks The wheel trajectory and center of mass trajectory generated by the travel information module 248 are displayed. The visualization module 247 includes a display interface that presents terrain blocks, wheel trajectories and center-of-mass trajectories as shown in Figure 4D and Figure 4E to the user for viewing. In particular, the visualization module can display trajectory information such as terrain blocks, wheel trajectories, and center-of-mass trajectories in different colors, color blocks, and lines, allowing users to intuitively see the movement trajectory of the mobile vehicle 16 .

If there is an obstacle ahead, as shown in Figure 6A, the obstacle 30 appears on the ditch 14 on the right. At this time, the terrain block generated by the semantic segmentation module 246 will be as shown in Figure 6B. The furrow 14 corresponds to the furrow block 34a in the terrain block, which is a terrain block inferred by the semantic segmentation module 246 with a blank split block situation. When the traveling information module 248 generates trajectory information based on the terrain blocks included in the terrain recognition results output by the semantic segmentation module 246, it will first find that the wheel trajectory of the ditch block 34a is interrupted when generating the wheel trajectory. Therefore, It will prompt that the traveling information module 248 needs to brake. In this case, the linear velocity and angular velocity provided by the traveling information module 248 to the traveling control module 249 will be zero to prevent the mobile vehicle 16 from hitting the obstacle 30 .

Therefore, through the field-type unmanned vehicle automatic navigation system and method provided by the present invention, it only needs to use a small amount of images of the road ahead under various environmental conditions (dry and wet soil, light intensity, etc.) and various growth stages of plants. , an inference model that can identify the terrain can be trained, and combined with the navigation information algorithm, the mobile vehicle can be controlled to move without being affected by factors such as sunlight intensity, plant size, soil conditions, etc. The navigation effect will not be affected. The present invention not only reduces the physical labor of farmers, but also enables multiple mobile vehicles to perform operations at the same time, thereby improving agricultural productivity. The calculated trajectory information can be used to count the daily movement distance and understand the application progress. In this way, This will improve farm management efficiency in the future.

The above descriptions are only preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Therefore, all equivalent changes or modifications made in accordance with the characteristics and spirit described in the scope of the present invention shall be included in the patent scope of the present invention.

10:Farmland 12: Qitian 14: Qigou 16:Mobile vehicle 20: Qitian type unmanned vehicle automatic navigation system 22:Image capture device 24: Processor 242:Image processing module 244:Annotation analysis module 246:Semantic segmentation module 247:Visualization module 248:Travel information module 249:Travel control module 30: Obstacles 32: Qitian block 34, 34a: Qigou block A: The center point of the bottom of the screen of the ditch area on the left B: The center point of the bottom of the screen of the ditch area on the right C: The center point of point A and point B C’: Center position at the bottom of the screen

Figure 1 is a block diagram of the Qitian type unmanned vehicle automatic navigation system of the present invention. Figure 2 is a schematic diagram of farmland including borders and ditches. Figure 3 is a schematic diagram of a mobile vehicle moving on farmland. Figures 4A to 4E are schematic diagrams of one embodiment of the process of the border-type unmanned vehicle automatic navigation system of the present invention. Figure 5 is a schematic diagram of the present invention's calculation of wheel trajectory, center of mass trajectory, traveling linear velocity and angular velocity on road surface images. Figure 6A and Figure 6B are schematic diagrams of obstacles ahead and corresponding terrain blocks.

20: Qitian type unmanned vehicle automatic navigation system

22:Image capture device

24: Processor

242:Image processing module

244:Annotation analysis module

246:Semantic segmentation module

247:Visualization module

248:Travel information module

249:Travel control module

Claims (16)

A border type unmanned vehicle automatic navigation system, which is installed on a mobile vehicle. The border type unmanned vehicle automatic navigation system includes: at least one image capture device to capture multiple road surface images; a label analysis module , connected to the at least one image capture device, generating plural annotations on the road surface images to identify the terrain in the road surface images; a semantic segmentation module connected to the annotation analysis module and the at least one image capture device , establish an inference model, and use at least one semantic segmentation artificial intelligence algorithm and the road surface images containing the annotations to train the inference model, and the inference model is used to output a terrain recognition result, wherein the terrain recognition result A plurality of terrain blocks included on the road surface images, the terrain blocks including at least one border block and at least two border blocks; a traveling information module connected to the semantic segmentation module, based on the terrain recognition result Calculate a set of trajectory information; and a travel control module, connect the travel information module and the mobile vehicle, receive at least part of the set of trajectory information, and control the movement of the mobile vehicle according to the set of trajectory information, wherein The traveling information module takes the center points of both sides of the at least two ditch blocks and connects them into a line to generate the two wheel trajectories of the mobile vehicle, and takes the center points of the two wheel trajectories and connects them into a line segment as The trajectory of the center of mass of the mobile vehicle. The Qitian-type unmanned vehicle automatic navigation system as described in claim 1 further includes an image processing module connected to the at least one image capture device, the annotation analysis module and the semantic segmentation module to receive the at least one image Capture the road surface images transmitted by the device and perform prediction After processing, the preprocessed road surface images are sent to the annotation analysis module or the semantic segmentation module. The Qitian-type unmanned vehicle automatic navigation system as described in claim 2, wherein the preprocessing performed by the image processing module includes adjusting the size of the road surface images, adjusting the color representation values of the road surface images, and adjusting the road surface images. Carry out sampling or a combination of the above three. The Qitian-type unmanned vehicle automatic navigation system as described in claim 1, wherein the set of trajectory information includes the wheel trajectory, center of mass trajectory, traveling linear speed and angular speed of the mobile vehicle, the linear speed is preset, and the The travel information module calculates the angular velocity of the mobile vehicle based on an offset of the center of mass trajectory. The Qitian-type unmanned vehicle automatic navigation system as described in claim 4, wherein the travel information module transmits the linear velocity and angular velocity in the set of trajectory information to the travel control module. The Qitian-type unmanned vehicle automatic navigation system as described in claim 4 further includes a visualization module connected to the travel information module for displaying the wheel trajectory and the center of mass on the terrain blocks. trajectory. The Qitian-type unmanned vehicle automatic navigation system as described in claim 1 further includes a visualization module connected to the semantic segmentation module for displaying the terrain blocks. For the Qitian-type unmanned vehicle automatic navigation system described in claim 6 or 7, the visualization module displays the trajectory information in different colors, color blocks, and lines. A QiTian-type unmanned vehicle automatic navigation method, which is applied to a QiTian-type unmanned vehicle automatic navigation system installed on a mobile vehicle. The QiTian type unmanned vehicle automatic navigation method includes the following steps: Use at least one image capture device to capture multiple road surface images; use a label analysis module to receive the road surface images, and generate multiple labels on the road surface images to identify the terrain in the road surface images; use a semantic segmentation The module receives the road surface images containing the annotations, establishes an inference model, and trains the inference model using at least one semantic segmentation artificial intelligence algorithm and the road surface images containing the annotations to output a terrain Recognition results, wherein the terrain recognition results include a plurality of terrain blocks on the road surface images, the terrain blocks include at least one border block and at least two border blocks; and using a travel information module based on the terrain The identification result calculates a set of trajectory information suitable for the movement of the mobile vehicle, and transmits it to a travel control module to control the movement of the mobile vehicle based on the set of trajectory information, wherein the travel information module obtains the at least two ditches. The center points of the two edges of the block are connected in a line to generate the two wheel trajectories of the mobile vehicle. The center points of the two wheel trajectories are connected in a line segment as the center of mass trajectory of the mobile vehicle. The method for automatic navigation of a border type unmanned vehicle as described in claim 9, after the step of using the image capturing device to capture the road surface images, further includes: using an image processing module to receive the data from the image capturing device. Transmit the road surface images and perform preprocessing; and transmit the preprocessed road surface images to the annotation analysis module or the semantic segmentation module. As claimed in claim 10, the border-type unmanned vehicle automatic navigation method, wherein the preprocessing performed by the image processing module includes adjusting the size of the road surface images, adjusting the color representation values of the road surface images, and adjusting the road surface images. Carry out sampling or a combination of the above three. The Qitian-type unmanned vehicle automatic navigation method as described in claim 9, wherein the set of trajectory information includes the wheel trajectory, center of mass trajectory, traveling linear speed and angular speed of the mobile vehicle, the linear speed is preset, and the The travel information module calculates the angular velocity of the mobile vehicle based on an offset of the center of mass trajectory. As for the Qitian-type unmanned vehicle automatic navigation method described in claim 12, the traveling information module transmits the linear velocity and angular velocity in the set of trajectory information to the traveling control module. The border type unmanned vehicle automatic navigation method described in claim 9 further includes using a visualization module to display the terrain blocks on the road surface images. The borderland type unmanned vehicle automatic navigation method described in claim 12 further includes using a visualization module to display the wheel trajectory and the center of mass trajectory on the terrain blocks. As for the Qitian-type unmanned vehicle automatic navigation method described in claim 15, the visualization module displays the trajectory information in different colors, color blocks, and lines.