CN109632265B - A vision-based detection system and method for docking state of unmanned boat water harvesting device - Google Patents

Abstract

vision-based unmanned ship water collection device docking state detection systems and methods, and discloses systems and methods for detecting docking states of water collection devices in real time when applied to an unmanned water quality detection boat, wherein the detection systems comprise a video acquisition module, a wireless image transmission module and a terminal processing and display module, images are acquired by the detection methods through the video acquisition module, the wireless image transmission module is sent to the terminal processing and display module, and after the terminal processing and display module receives the images, an interest area threshold detection algorithm is operated to process and display the images, and finally a result of whether docking is successful is obtained.

Description

A vision-based detection system and method for docking state of unmanned boat water harvesting device

technical field

The invention relates to the technical field of unmanned equipment, in particular to an unmanned water quality testing boat, in particular to a system and method for real-time detection of the docking state of a water extraction device applied to an unmanned water quality testing boat.

Background technique

At present, my country's marine water quality testing mainly relies on manual collection and testing. The testing cost is high, the automation cost is low, and the risk of going to sea is high. At the same time, some sea areas are not suitable for manned operations due to pollution and high risks. With the increasing maturity of unmanned boat technology and the development and application of automatic water quality testing equipment, unmanned water quality testing boats have emerged as the times require, which can realize automatic collection, distribution and testing of water samples. ) Whether the docking mechanism of the water sample collection module of the water collector and the water sample distribution module is accurately docked directly affects the water sample collection and detection, so the detection of the docking mechanism is very necessary.

Since the unmanned water quality testing boat is in an unmanned autonomous process throughout the water quality testing task, people can only monitor remotely from the mother ship or shore-based. Simple video monitoring requires a lot of time and energy of the monitor, and in a certain environment from The poor picture effect under the monitoring screen will bring inconvenience or even mistakes to manual monitoring, which not only wastes human resources, but may also lead to errors in monitoring results. Therefore, an automatic detection system is needed to match the docking Institutions conduct real-time monitoring.

SUMMARY OF THE INVENTION

In order to solve the deficiencies of the prior art, the present invention proposes a detection system based on vision. The docking state of the water collection device on the unmanned water quality detection boat is monitored in real time, so as to improve the detection accuracy, reduce labor consumption, and ensure the water quality monitoring task. Work properly.

The technical problem to be solved by the present invention is achieved through the following technical solutions:

A vision-based detection system for the docking state of an unmanned boat water harvesting device, comprising a video acquisition module, a wireless image transmission module, a terminal processing and display module, wherein the video acquisition module is installed on the unmanned boat, and includes an alignment water harvesting device The network camera of the docking mechanism and the wireless image transmission module include two groups of wireless transceiver modules and antennas, one of which is set on the unmanned boat and connected to the video acquisition module, and the other set of wireless transceiver modules and antennas are set on the The mother ship or shore base is connected to the terminal processing and display module, and the two groups of wireless transceiver modules and the antenna are connected by wireless communication. The terminal processing and display module includes a computer terminal with embedded visual detection algorithm and a display screen.

In the present invention, the camera of the video acquisition module is a camera with a fixed focal length, and the docking mechanism of the water harvesting device is aligned with an upper docking column, a lower docking column and a rubber sealing ring, wherein the upper docking column is connected with the water harvesting device. The water pumping pipeline, the lower butt column is connected with the water outlet of the CTD water collector, the upper butt column and the lower butt column are connected by mutual insertion, and the rubber sealing ring is arranged at the connection of the upper butt column and the lower butt column.

In the present invention, the terminal processing and display module includes a main interface and a sub-interface, the main interface is used for real-time monitoring and detection and intelligent early warning, and is provided with a display frame, parameter setting buttons, detection switches and alarm lights, and the sub-interface passes through the main interface. The parameter setting button on the interface is activated, which is used to set training and detection parameters, and a parameter setting box is set on it.

A vision-based method for detecting the docking state of an unmanned boat water harvesting device, comprising the following steps:

(1) The camera of the video collection module collects the video and picture information at the docking point of the docking mechanism of the water collection device;

(2) The video information collected by the video acquisition module is uploaded from the unmanned boat to the terminal processing and display module on the mother ship or shore base through the wireless image transmission module;

(3) The terminal processing and display module runs the docking detection monitoring software, runs the docking detection algorithm, displays the results on the terminal, and performs dual alarms with flashing lights and voice.

Wherein, the docking detection algorithm of step (3) is a region of interest threshold detection algorithm (RTD, roi threshold detection algorithm), including offline training and online detection stages, wherein the specific steps of the offline training stage include:

① Take the pictures collected by the video acquisition module as positive samples for successful docking, and negative samples for failed docking, construct a training data set, and use the sample color pictures as input to the algorithm, and the original image size is 1280×720;

② Extract the rectangular area at the intermediate docking position as the region of interest (roi), the size of the region is 440×215 resolution, the background in the region of interest is simple, and the foreground and background are easily separated by color; this step not only reduces the processing of images It can also make the algorithm simpler to process and improve the efficiency of the algorithm;

③ Perform grayscale processing on the color image of the region of interest obtained in the previous step to obtain a grayscale image;

④ Threshold the grayscale image obtained in the previous step, separate the low-brightness washer area and the high-brightness white butt column area, and count the number of low-brightness pixels in the area;

⑤ Run steps ① to ④ on the training data set obtained in step ① to obtain the number of low-brightness pixels C _i of each sample, take the number of pixels of the sample as a variable, and attach a positive sample label to the successfully connected sample, and to the failed sample The negative sample label is attached, and the classification threshold t is obtained by training, and the offline training process is completed;

The specific steps in the online detection stage of the region of interest threshold detection algorithm include:

⑥ The color image collected by the video acquisition module is used as the input of the algorithm, and the size of the image is 1280×720;

⑦ Extract the rectangular area at the intermediate docking position as the area of interest, and the area size is 440×215;

⑧ Perform grayscale processing on the color image of the region of interest obtained in the previous step to obtain a grayscale image;

⑨ Threshold the grayscale image obtained in the previous step, and count the number of low-brightness pixels Ct in the area;

⑩ Take the classification threshold t obtained in step ⑤ offline training as a parameter, and compare it with the number of low-brightness pixels C _t obtained by online detection. If C _t < t, the docking is successful; otherwise, if C _t ≥ t, the docking fails. , it needs to be reconnected.

In the present invention, because the structure of the butt joint of the docking mechanism of the water harvesting device is relatively simple, only the upper and lower white docking columns and a black sealing ring can be adjusted by a simple pixel statistics method to adjust the appropriate camera installation position. Docking detection. When the docking is successful, there are fewer low-brightness pixels in the image. After the docking fails, the gap between the docking columns is larger, and there are more low-brightness pixels in the image. For this kind of scene, the threshold detection algorithm of the region of interest is adopted, the region near the docking point is the region of interest, and the number of low-brightness pixels after binarization of the image in the candidate region of interest is the threshold value, and the docking success or failure can be easily judged. . This detection algorithm includes two stages: offline training and online monitoring. In the offline training process, the pictures collected at ordinary times are regarded as positive samples for successful docking, and negative samples for failed docking. The threshold is a judgment condition, and it can be judged whether the connection is successful, which is simple and efficient.

Compared with the prior art, the present invention is easy to build, easy to operate, and low in cost, realizes the automatic detection of the docking state of the docking mechanism of the water collection device on the unmanned water quality testing boat, improves the automation degree of the unmanned water quality testing boat, and reduces the cost of the unmanned water quality testing boat. Manpower consumption is more intelligent and efficient. At the same time, the display terminal interface is simple and easy to operate. The automatic docking detection and intelligent voice alarm are very user-friendly, and the human-computer interaction performance is good. Good portability.

Description of drawings

FIG. 1 is a schematic diagram of the system connection of the present invention;

FIG. 2 is a schematic diagram of the workflow of the present invention;

3 is a schematic diagram of the offline training process of the visual detection algorithm of the present invention;

FIG. 4 is a schematic diagram of the online detection process of the visual detection algorithm of the present invention;

5 is a schematic diagram of the main interface of the terminal processing and display module of the present invention;

FIG. 6 is a schematic diagram of a sub-interface of the terminal processing and display module of the present invention.

In the figure: the upper docking column 1, the rubber sealing ring 2, the lower docking column 3, the docking mechanism of the water collection device 4, the camera 5, the wireless transceiver module and antenna 6, the wireless image transmission module 7, and the terminal processing and display module 8.

Detailed ways

The present invention will be further described below with reference to the accompanying drawings and specific preferred embodiments, but the protection scope of the present invention is not limited thereby.

Referring to the vision-based detection system and method for the docking state of the unmanned boat water harvesting device shown in Figures 1-6, wherein, in Figure 1, the vision-based detection system for the docking state of the unmanned water harvesting device includes a video acquisition module , a wireless image transmission module 7, a terminal processing and display module 8, wherein the video acquisition module is installed on the unmanned boat, including a fixed focal length network camera 5, which is set at the docking mechanism 4 of the water harvesting device, and the docking mechanism of the water harvesting device 4 includes an upper butt column 1, a lower butt column 3 and a rubber sealing ring 2, wherein the upper butt column 1 is connected with the pumping pipeline of the water extraction device, the lower butt column 3 is connected with the water outlet of the CTD water extraction device, and the upper butt column 1 is connected with The lower docking column 3 is connected by mutual insertion, and the rubber sealing ring 2 is arranged at the connection between the upper docking column 1 and the lower docking column 3; the wireless image transmission module 7 includes two groups of wireless transceiver modules and antennas 6, one of which is The wireless transceiver module and antenna 6 are arranged on the unmanned boat and are connected to the video acquisition module. Another group of wireless transceiver modules and antennas 6 are arranged on the mother ship or shore base, and are connected to the terminal processing and display module 8. The two groups of wireless transceiver modules and The antennas 6 are connected through wireless communication, and the terminal processing and display module 8 includes a computer terminal with an embedded visual detection algorithm and a display screen, as shown in Figures 5 and 6, including a main interface and a sub-interface, and the main interface is used for real-time monitoring. Detection and intelligent early warning. There are display frame, parameter setting button, detection switch and alarm light on it. The sub-interface is activated through the parameter setting button on the main interface, which is used to set training and detection parameters, and the parameter setting box is set on it.

Based on the above structure, a vision-based method for detecting the docking state of an unmanned boat water harvesting device, as shown in Figure 2, includes the following steps:

(1) The camera of the video collection module collects the video and picture information at the docking point of the docking mechanism of the water collection device;

(2) The video information collected by the video acquisition module is uploaded from the unmanned boat to the terminal processing and display module on the mother ship or shore base through the wireless image transmission module;

(3) The terminal processing and display module runs the docking detection monitoring software, runs the docking detection algorithm, displays the results on the terminal, and performs dual alarms with flashing lights and voice.

Wherein, the docking detection algorithm in step (3) is a region of interest threshold detection algorithm, including offline training and online detection stages, wherein the offline training stage is shown in Figure 3, and its specific steps include:

① Take the pictures collected by the video acquisition module as positive samples for successful docking, and negative samples for failed docking, construct a training data set, and use the sample color pictures as input to the algorithm, and the original image size is 1280×720;

② Extract the rectangular area at the intermediate docking position as the area of interest, the size of the area is 440×215 resolution, the background in the area of interest is simple, and the foreground and background are easily separated by color;

③ Perform grayscale processing on the color image of the region of interest obtained in the previous step to obtain a grayscale image;

④ Threshold the grayscale image obtained in the previous step, separate the low-brightness washer area and the high-brightness white butt column area, and count the number of low-brightness pixels in the area;

⑤ Run steps ① to ④ on the training data set obtained in step ① to obtain the number of low-brightness pixels C _i of each sample, take the number of pixels of the sample as a variable, and attach a positive sample label to the successfully connected sample, and to the failed sample The negative sample label is attached, and the classification threshold t is obtained by training, and the offline training process is completed;

The online detection stage of the region of interest threshold detection algorithm is shown in Figure 4, and its specific steps include:

⑥ The color image collected by the video acquisition module is used as the input of the algorithm, and the size of the image is 1280×720;

⑦ Extract the rectangular area at the intermediate docking position as the area of interest, and the area size is 440×215;

⑧ Perform grayscale processing on the color image of the region of interest obtained in the previous step to obtain a grayscale image;

⑨ Threshold the grayscale image obtained in the previous step, and count the number of low-brightness pixels Ct in the area;

⑩ Take the classification threshold t obtained in step ⑤ offline training as a parameter, and compare it with the number of low-brightness pixels C _t obtained by online detection. If C _t < t, the docking is successful, otherwise, if C _t ≥ t, the docking fails. , it needs to be reconnected.

Therefore, in combination with the above structures and methods, it can be found that the system of the present invention is easy to build, easy to operate, and low in cost, realizes automatic detection of the docking state of the docking mechanism of the water collection device on the unmanned water quality testing boat, and improves the performance of the unmanned water quality testing boat. The degree of automation, more intelligent and efficient.

Claims (4)

1. a vision-based method for detecting the docking state of an unmanned boat water harvesting device, is characterized in that: comprise the following steps: The camera of the video acquisition module collects the video and picture information of the docking point of the docking mechanism of the water harvesting device; The video information collected by the video acquisition module is uploaded from the unmanned boat to the terminal processing and display module on the mother ship or shore base through the wireless image transmission module; The terminal processing and display module runs the docking detection monitoring software, runs the docking detection algorithm, displays the results on the terminal, and performs dual alarms with flashing lights and voice; Wherein: the docking detection algorithm is an interest region threshold detection algorithm, including offline training and online detection stages, wherein the specific steps of the offline training stage include: ① Take the pictures collected by the video acquisition module as positive samples for successful docking, and negative samples for failed docking, construct a training data set, and use the sample color pictures as input to the algorithm, and the original image size is 1280×720; ② Extract the rectangular area at the intermediate docking position as the area of interest, the size of the area is 440×215 resolution, the background in the area of interest is simple, and the foreground and background are easily separated by color; ③ Perform grayscale processing on the color image of the region of interest obtained in the previous step to obtain a grayscale image; ④ Threshold the grayscale image obtained in the previous step, separate the low-brightness washer area and the high-brightness white butt column area, and count the number of low-brightness pixels in the area; ⑤ Run steps ① to ④ on the training data set obtained in step ① to obtain the number of low-brightness pixels C _i of each sample, take the number of pixels of the sample as a variable, and attach a positive sample label to the successfully connected sample, and to the failed sample The negative sample label is attached, and the classification threshold t is obtained by training, and the offline training process is completed; The specific steps in the online detection stage of the region of interest threshold detection algorithm include: ⑥ The color image collected by the video acquisition module is used as the input of the algorithm, and the size of the image is 1280×720; ⑦ Extract the rectangular area at the intermediate docking position as the area of interest, and the area size is 440×215; ⑧ Perform grayscale processing on the color image of the region of interest obtained in the previous step to obtain a grayscale image; ⑨ Threshold the grayscale image obtained in the previous step, and count the number of low-brightness pixels Ct in the area; ⑩ Take the classification threshold t obtained in step ⑤ offline training as a parameter, and compare it with the number of low-brightness pixels C _t obtained by online detection. If C _t < t, the docking is successful; otherwise, if C _t ≥ t, the docking fails. , it needs to be reconnected. 2. The vision-based method for detecting the docking state of an unmanned boat water harvesting device according to claim 1, wherein the video acquisition module, the wireless image transmission module, the terminal processing and the display module constitute a vision-based unmanned aerial vehicle. The detection equipment for the docking state of the water harvesting device of the boat, wherein the video acquisition module is installed on the unmanned boat, including a network camera aligned with the docking mechanism of the water harvesting device, and the wireless image transmission module includes two groups of wireless transceiver modules and antennas, one of which is The wireless transceiver module and antenna are arranged on the unmanned boat and connected to the video acquisition module. Another group of wireless transceiver modules and antennas are arranged on the mother ship or shore base, and are connected to the terminal processing and display module. Between the two groups of wireless transceiver modules and the antenna Connected by wireless communication, the terminal processing and display module includes a computer terminal with a built-in visual detection algorithm and a display screen. 3. The vision-based method for detecting the docking state of an unmanned boat water harvesting device according to claim 2, wherein the camera of the video acquisition module is a camera with a fixed focal length, and the aligned water harvesting device docking mechanism comprises: The upper butt column, the lower butt column and the rubber sealing ring, wherein the upper butt column is connected with the pumping pipe of the water extraction device, the lower butt column is connected with the water outlet of the CTD water extraction device, and the upper butt column and the lower butt column are connected by mutual insertion. The rubber sealing ring is arranged at the connection between the upper butt column and the lower butt column. 4. The vision-based method for detecting the docking state of an unmanned boat water harvesting device according to claim 2, wherein the terminal processing and display module comprises a main interface and a sub-interface, and the main interface is used for real-time monitoring and detection and intelligent early warning , the display frame, parameter setting button, detection switch and alarm light are set on it. The sub-interface is activated through the parameter setting button on the main interface, which is used to set training and detection parameters, and the parameter setting box is set on it.

Images