CN110677640B - In-loop testing method and device for unmanned vehicle-mounted intelligent camera - Google Patents

Abstract

The invention discloses an in-loop testing method and device for an unmanned vehicle-mounted intelligent camera, the device comprises a box body and a box cover, an LED display screen is arranged on the box body, a fan accommodating device and an LED irradiation lamp accommodating device are respectively arranged on two sides of the box body, a water storage tank is arranged below the rear end of the box body, a guide rail, a graduated scale and a water collecting tank are longitudinally arranged on the bottom surface of the box body, a sliding block is arranged on the guide rail, a threaded rod is fixedly connected to the center of the sliding block, the upper end of the threaded rod is inserted into a camera fixing disc through a height adjusting disc, and the performance of the intelligent camera under different weather conditions is tested by simulating the weather environment in a. The testing method and the testing device provided by the invention can realize indoor testing of the performance of the intelligent camera, and the testing method is simple, the testing device is convenient to install, the cost is lower, the work is safe and reliable, and the popularization is stronger.

Description

In-loop testing method and device for unmanned vehicle-mounted intelligent camera

Technical Field

The invention belongs to the technical field of unmanned driving, and particularly relates to an in-loop testing method and device for an unmanned vehicle-mounted intelligent camera.

Background

With the development of intelligent networking technology for automobiles, the development of unmanned automobiles has been regarded as an important strategic reserve in all countries in the world. In particular, the rapid development of machine learning and deep learning technologies in recent years greatly promotes the progress of automobile unmanned. 3D map reconstruction is one of the most important technologies in the field of unmanned driving, and the mainstream unmanned vehicles generally adopt a multi-line radar as a 3D sensor to realize the technology. However, radar equipment is expensive, only depth information can be identified, texture and color cannot be acquired, and peripheral perception is insufficient.

The method for acquiring the depth information by adopting the camera for stereo matching is a classical idea method. After the internal and external parameters of the two cameras are determined, depth information can be directly obtained theoretically by means of a similar triangle theorem, but in practice, particularly in outdoor scene use, the cameras are greatly interfered by external light, a large amount of invalid noise and information exist, numerical value accuracy is always poor, and the digital depth measuring device can only be used as an auxiliary sensor of a radar. With the maturity of deep learning technology and the introduction of twin networks, a great number of researchers try to solve the stereo matching problem of the camera through deep learning. With the adoption of the twin network for the depth learning model of stereo matching, object depth information with high accuracy can be obtained through the binocular camera, so that a basis is provided for the camera to replace expensive radar for distance measurement.

At present, smart cameras capable of detecting pedestrians, vehicles, and obstacles have appeared based on the deep learning technology, and such smart cameras can mark not only the positions of the corresponding pedestrians, vehicles, and obstacles, but also the distances therebetween. The accuracy of the labeling of the distances of the detected pedestrians, vehicles and obstacles by the intelligent cameras can meet the requirement of the unmanned automobile on the accuracy of the depth information, and the accuracy of the unmanned automobile is tested. However, camera manufacturers test various parameters of cameras, and only ensure that the performance of the cameras reaches the indexes and the performance of the cameras is stably developed, and the test is not suitable for the field of automobiles. In the field of automobiles, an automobile factory performs real-time test on a camera, and although the test accuracy is high, the test method not only consumes a large amount of manpower, material resources and financial resources, but also takes too long time and has certain dangerousness. And because the real vehicle test is limited by external light, weather conditions, test sites and the like, the real vehicle test cannot meet various different test scenes and test working conditions easily, and the operation is complex.

In summary, the following problems exist in the test and evaluation of the current unmanned vehicle-mounted intelligent camera:

1. the test of camera manufacturers is a test of various parameters of cameras, and the performance index requirements of the cameras on unmanned automobiles are difficult to meet.

2. The automobile manufacturer tests the camera in real time, which consumes a lot of manpower, material resources and financial resources, consumes a long time and has certain danger.

3. When the camera is used in an outdoor scene, the camera is limited by external light, weather conditions, test fields and the like, real vehicle tests cannot easily meet various different test scenes and working conditions, and the operation is complex.

4. With the development of the depth learning technology and the stereo matching technology, the depth of field measuring capability of the binocular camera is remarkably improved, so that the evaluation on the depth of field measuring accuracy of the binocular camera is more and more important.

Disclosure of Invention

In order to solve the technical problems, the invention provides the unmanned vehicle-mounted intelligent camera in-loop testing device, which realizes the indoor testing of the performance of the intelligent camera.

An unmanned vehicle-mounted intelligent camera in-loop testing device comprises a box body and a box cover, and is characterized in that an LED display screen is arranged at the center of the front side of the box body, a fan accommodating device and an LED irradiation lamp accommodating device are respectively arranged at the left side and the right side of the box body, a water storage tank is arranged below the rear end of the box body, and a guide rail, a graduated scale and a water collecting tank are longitudinally arranged on the bottom surface of the box body; the water storage tank is characterized in that a water inlet hole and a water outlet hole are respectively arranged at two ends of the water storage tank, the water inlet hole is arranged above the water storage tank, the water outlet hole is arranged below the water storage tank, and a water storage tank groove is formed in the water storage tank; the guide rails are double guide rails and are symmetrically distributed about the longitudinal axis of the box body, the cross sections of the guide rails are I-shaped, and sliding blocks are arranged on the guide rails; the two graduated scales are symmetrically distributed about the longitudinal axis of the box body and are positioned on the outer side of the guide rail, the 0-graduation position of the graduated scale is flush with the screen of the LED display screen and extends to the tail end of the box body, and the minimum graduation of the graduated scale is mm; the front end part of the water collecting tank is aligned with the front end part of the guide rail, the rear end part of the water collecting tank is communicated with the water storage tank, and the depth of the water collecting tank is gradually increased from front to back; the box cover is positioned above the box body, and the wedge-shaped fine sand container and the rainfall spray head container are arranged on the box cover.

Furthermore, the device comprises four supporting columns which are distributed on four feet at the bottom of the box body and are directly connected with the box body.

Furthermore, the LED display screen is connected with a power supply through a power line and is provided with a USB interface.

Furthermore, the fan containing devices are symmetrically arranged left and right relative to the longitudinal axis of the box body, and a fan is placed in each fan containing device; the LED irradiation lamp accommodating devices are arranged in a bilateral symmetry mode, and one LED irradiation lamp is placed in each LED irradiation lamp accommodating device.

Furthermore, the slide block is in clearance fit with the guide rail, two positioning screw holes are respectively arranged on the left side and the right side of the slide block, the slide block can be fixed at any position on the guide rail through the positioning screw holes by the positioning bolts, a distance reading pointer is respectively arranged at the middle position of the edges of the two sides of the slide block, and the distance reading pointer is used for indicating the scale size on the graduated scale.

Furthermore, a threaded rod is fixedly connected to the center of the sliding block, and the upper end of the threaded rod is inserted into the camera fixing disc through the height adjusting disc; the middle of the height adjusting disc is provided with a threaded hole for allowing a threaded rod to pass through; the center of the camera fixing disc is provided with a camera fixing disc through hole which is larger than the outer diameter of the threaded rod and allows the threaded rod to pass through directly, and the camera fixing disc is connected with a fixing block through a connecting bolt; the camera fixing block is characterized in that two limiting bolts are connected to the middle of the fixing block in a threaded mode, a camera fixing block is arranged on the inner side of the fixing block, and the limiting bolts are inserted into the unthreaded holes of the camera fixing block through the fixing block.

Furthermore, the wedge-shaped fine sand container is provided with openings at the upper part and the lower part, the upper opening is larger than the lower opening, two movable blocks are arranged in the wedge-shaped fine sand container, and a rectangular groove is arranged below each movable block.

Further, the rainfall spray head container is in a conical shape with a small upper part and a big lower part, and the upper part and the lower part of the rainfall spray head container are both opened.

Furthermore, the size of case lid is greater than the size of box, and the case lid edge has a round guide way, guarantees that the case lid can cover on the box steadily.

The invention provides an in-loop test method for an unmanned vehicle-mounted intelligent camera, which comprises the following steps of:

step S1: acquiring the visual screens to be detected in different scenes, storing the visual screens in a storage device, connecting the storage device carrying the video to be detected with a USB interface on an LED display screen, and opening and playing the video;

step S2: adjusting the position of the intelligent camera to be detected relative to the LED display screen to enable the center of the intelligent camera to be aligned with the center line of the LED display screen;

step S3: changing the distance of the intelligent camera relative to the center of the LED display screen, and observing the numerical value of the distance reading pointer pointing to the graduated scale to obtain a specific distance t;

step S4: the illumination condition in the box is changed by controlling the brightness and the illumination range of the LED illumination lamp, and different illumination environments in a real scene are simulated;

step S5: the sand raising environment in a real scene is simulated by controlling the spreading amount of the fine sand in the wedge-shaped fine sand container above the box body and simultaneously adjusting the wind power of the fans at two sides of the box body;

step S6: adding a proper amount of water into the water collecting tank, pumping out the water by a water pump, sending the water to a rainfall sprayer above the tank cover, spraying the water by the rainfall sprayer to form rainfall, adjusting the wind power of the fan, and simulating a rainy day environment in a real scene;

step S7: simulating different weathers in a real scene by adjusting illumination, rainfall, sand blowing strength or size, acquiring photos shot by an intelligent camera under different weather conditions, and recording classification types of objects at the marked position and correct and wrong quantity in the classification types;

step S8: and comparing the image depth information given by the intelligent camera with the recorded distance t, evaluating the detection effect and the depth calculation precision of the camera in different scenes, and realizing the measurement of the performance of the camera.

The invention provides an in-loop test method and device for an unmanned vehicle-mounted intelligent camera, which have the following beneficial effects:

1. the invention solves the problems that the automobile manufacturer consumes a large amount of manpower, material resources, financial resources, consumes a long time and the like when carrying out real-time test on the camera.

2. The rainfall simulation device can simulate rainfall and sandstorm weather conditions encountered by the vehicle-mounted camera in the working process through the coordinated matching of the rainfall simulation sprayer, the sandstorm leakage device and the fan, and can test the working states of the vehicle-mounted intelligent camera in different weather environments.

3. The camera in-loop test method can test the recognition effect of obstacles, pedestrians, vehicles and the like of the monocular camera, and can measure the depth of field measurement accuracy of the binocular camera.

4. The method is clear, simple in structure, convenient to install, low in cost, easy to market, safe and reliable in work and high in popularization.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.

Fig. 1 is a schematic flow chart of an in-loop test method of an unmanned vehicle-mounted intelligent camera according to the invention.

Fig. 2 is an axial side projection view of the unmanned vehicle-mounted intelligent camera on-board ring testing device.

Fig. 3 is a top view of an in-loop testing apparatus for an unmanned vehicle-mounted intelligent camera according to the present invention.

Fig. 4 is a front view of a case cover of an unmanned vehicle-mounted intelligent camera in a ring test device according to the present invention.

Fig. 5 is a rear view of a case cover of an unmanned vehicle-mounted intelligent camera in a ring test device according to the present invention.

Fig. 6 is a structural diagram of the inside of a box body of an unmanned vehicle-mounted intelligent camera in a ring test device.

Fig. 7 is a front view of a box body of an unmanned vehicle-mounted intelligent camera in a ring testing device.

FIG. 8 is a guide rail slider axonometric view of an unmanned vehicle-mounted intelligent camera in a ring test device according to the present invention.

Fig. 9 is a structure diagram of a guide rail slider of the unmanned vehicle-mounted intelligent camera in the loop test device.

Fig. 10 is a top view of a box of an unmanned vehicle-mounted intelligent camera in a ring testing device according to the present invention.

Fig. 11 is a structural diagram of a camera (monocular) attitude adjusting device of an unmanned vehicle-mounted intelligent camera in a ring test device according to the present invention.

Fig. 12 is a structural diagram of a camera (binocular) attitude adjusting device of an unmanned vehicle-mounted intelligent camera in a loop test device according to the present invention.

1. The intelligent camera comprises a box body, 2. a support column, 3. a fan accommodating device, 4. an LED irradiation lamp accommodating device, 5. a water storage tank, 6. a box cover, 7. a camera fixing disc, 8. a sliding block, 9. a guide rail, 10. a graduated scale, 11. an LED display screen, 12. a fan, 13. an LED irradiation lamp, 14. a water collecting tank, 15. a double (single) -mesh intelligent camera, 51. a water inlet hole, 52. a water outlet hole, 53. a water storage tank groove, 61. a rainfall spray head accommodating device, 62. a movable block, 63. a wedge-shaped fine sand accommodating device, 64. a guide groove, 65. a box cover round through hole, 66. a box cover rectangular through hole, 67. a rectangular groove, 71. a camera fixing disc through hole, 72. a camera fixing block, 73. a connecting bolt, 74. a limiting bolt, 75. a fixing block, 81. a threaded rod, 82. a positioning bolt, 83. a distance reading pointer.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

The invention provides an in-loop testing device of an unmanned vehicle-mounted intelligent camera, which has the following structure:

referring to fig. 2 and 3, an axis side projection view and a top view of the unmanned vehicle-mounted intelligent camera in the loop test device are shown: the LED illuminating lamp box comprises a box body 1, four supporting columns 2, two fan accommodating devices 3, two LED illuminating lamp accommodating devices 4, a water storage tank 5 and a box cover 6. The two fan accommodating devices 3 and the two LED illuminating lamp accommodating devices 4 are positioned on two sides of the box body 1 and are respectively and symmetrically arranged; the water storage tank 5 is positioned below the tank body 1, water inlet holes 51 are symmetrically arranged above two ends of the water storage tank 5, and a water outlet hole 52 is positioned below the water storage tank 5 and opposite to the water inlet holes 51; the four support columns 2 are directly connected with the box body 1 and play a role in supporting the whole device; the case cover 6 is located above the case body 1, and a wedge-shaped fine sand container 63 and a rainfall spray head container 61, both of which are located on the central axis of the case cover 6, are sequentially arranged from front to back on the case cover 6.

Referring to fig. 4 and 5, a front view of a case cover and a back view of the case cover of the unmanned vehicle-mounted intelligent camera in the loop test device are shown: the cover 6 is provided with a wedge-shaped fine sand receiver 63 and a rain spray receiver 61. The wedge-shaped fine sand container 63 is provided with openings at the upper and lower parts, the upper opening is larger than the lower opening, and two movable blocks 62 are arranged in the wedge-shaped fine sand container; the movable blocks 62 are completely attached to the inner wall of the wedge-shaped fine sand container 63, the lower ends of the movable blocks 62 are just matched with the rectangular groove 67, fine sand is placed between the two movable blocks 62, and the speed and the flow of the hourglass can be changed by adjusting the distance between the two movable blocks 62; the rainfall nozzle receptacle 61 is conical with a small top and a large bottom, both open at the top and bottom. The reverse side of the box cover 6 is respectively provided with a box cover rectangular through hole 66 and a box cover circular through hole 65 which respectively correspond to the rectangular groove 67 of the wedge-shaped fine sand container 63 and the lower hole of the rainfall spray head container 61. The size of the box cover 6 is slightly larger than that of the box body 1, and a circle of guide grooves 64 are formed in the edge of the box cover 6 to ensure that the box cover 6 can stably cover the box body 1.

Referring to fig. 6 and 7, the structure diagram and the front view of the inside of the case of the unmanned vehicle-mounted intelligent camera in the loop test device are as follows: the internal structure of the box body comprises a camera fixing disc 7, a sliding block 8, a guide rail 9, a graduated scale 10, an LED display screen 11, a fan 12, an LED illuminating lamp 13, a water collecting tank 14 and a double (single) eye intelligent camera 15. The camera fixing disc 7 is used for placing a double (single) eye intelligent camera 15, so that the position of the double (single) eye intelligent camera 15 can be adjusted; the guide rails 9 are positioned in the middle of the box body 1, the two guide rails 9 are symmetrically arranged relative to the central axis of the box body 1, the guide rails 9 are provided with the sliding blocks 8, and the sliding blocks 8 can freely move on the guide rails 9; the two graduated scales 10 are symmetrically arranged about the central axis of the box body 1 and are positioned on the outer sides of the two guide rails 9, the minimum scale of the graduated scales 10 is mm, four water collecting grooves 14 are respectively arranged on the outer sides of the two graduated scales 10, the two water collecting grooves 14 are arranged between the two guide rails 9, the two water collecting grooves 14 in the middle of the guide rails 9 are wider than the eight water collecting grooves 14 on the two sides of the graduated scales 10, and the water flow is larger because the rainfall in the middle is larger than that on the two sides; the fan 12 is placed in the fan accommodating device 3, the fan 12 can swing left and right, can swing freely in the fan accommodating device 3, and has different gear sizes so as to obtain different wind power levels; the LED lamps 13 are arranged in the LED lamp accommodating device 4, the LED lamps 13 face the LED display screen 11, and the two LED lamps 13 are matched with each other and can irradiate the whole LED display screen 11; the LED display screen 11 is provided with two USB interfaces, and a storage device such as a U disk or a mobile hard disk can be inserted into the USB interfaces.

Referring to fig. 8 and 9, the guide rail and slider device of the unmanned vehicle-mounted intelligent camera on-loop testing device of the invention is as follows: the device comprises a guide rail 9, a sliding block 8, a positioning bolt 82, a distance reading pointer 83, a threaded rod 81, a camera fixing disc 7, a height adjusting disc 84 and a camera fixing disc through hole 71. The guide rail 9 is I-shaped, the guide rail 9 is in clearance fit with the sliding block 8, and lubricating grease is coated between the guide rail 9 and the sliding block 8, so that the sliding block 8 can move freely on the guide rail 9; two positioning screw holes 85 are respectively arranged on the left side and the right side of the sliding block 8, the positioning bolts 82 can fix the sliding block 8 at any position on the guide rail 9 through the positioning screw holes 85, distance reading pointers 83 are respectively arranged at the middle positions of the two sides of the sliding block 8, and the distance reading pointers 83 are used for indicating the scale size on the graduated scale 10; the threaded rod 81 is fixedly connected to the middle of the sliding block 8, and the threaded rod 81 is connected with the height adjusting disc 84 through the sliding block 8; a threaded hole is formed in the middle of the height adjusting disc 84 to allow the threaded rod 81 to pass through, and the camera fixing disc 7 is arranged above the height adjusting disc 84; the middle of the camera fixing disc 7 is provided with a camera fixing disc through hole 71 slightly larger than the outer diameter of the threaded rod 81, the threaded rod 81 is allowed to pass through the height adjusting disc 84, the height of the camera fixing disc 7 can be reduced by rotating the height adjusting disc 84 clockwise, and the height of the camera fixing disc 7 can be increased by rotating the height adjusting disc 84 counterclockwise.

Referring to fig. 10, a top view of an unmanned vehicle-mounted intelligent camera in-loop testing device box according to the present invention: comprises a graduated scale 10, a water collecting tank 14, a water storage tank 53, a guide rail 9 and a water storage tank 5. The 0 scale position of the graduated scale 10 is aligned with the screen position of the LED display screen 11 and extends to the tail end of the box body 1; the initial positions of the water collecting tank 14 and the guide rail 9 are aligned, the guide rail 9 extends to the extreme end of the box body 1, the water collecting tank 14 is intersected with the water storage tank 53, the depth of the water collecting tank 14 is deeper and deeper towards the extreme end of the box body 1 along the guide rail 9, so that water in the water collecting tank 14 flows back to the water storage tank 53 along the inner wall of the water collecting tank 14, the water flowing into the water storage tank 53 is stored in the water storage tank 5, flows into a rainfall nozzle along a water pipe under the action of a water pump, then enters the box body 1 through the rainfall nozzle container 61 and the circular through hole 65 of the box cover to form rainfall, and the falling raindrops flow into the water storage tank 5 through the water collecting tank 14, so that the water can be recycled.

Referring to fig. 11, a camera (monocular) attitude adjusting device of an unmanned vehicle-mounted intelligent camera in a ring test device according to the present invention: the device comprises a camera fixing disc 7, a height adjusting disc 84, a threaded rod 81, a camera fixing block 72, a connecting bolt 73, a limiting bolt 74, a fixing block 75, a distance reading pointer 83 and a sliding block 8. The camera fixing disc 7 is connected with a fixing block 75 through a connecting bolt 73, the fixing block 75 is connected with two limiting bolts 74 through threads, a camera fixing block 72 is arranged on the inner side of the fixing block 75, the limiting bolts 74 are inserted into a light hole of the camera fixing block 72 through the fixing block 75, and the camera fixing block 72 is driven to move by adjusting the limiting bolts 74 on two sides, so that the left position and the right position of a camera can be adjusted.

Referring to fig. 12, the camera (binocular) attitude adjusting apparatus of the unmanned vehicle-mounted intelligent camera in-loop test apparatus according to the present invention is used for a binocular intelligent camera performance test.

The working principle of the device is as follows: the positioning bolt 82 is loosened, the slide block 8 is moved back and forth to a proper position along the guide rail 9, then the positioning bolt 82 is tightened to reliably position the slide block 8 relative to the guide rail 9, and if the slide block 8 is found to be difficult to slide during use, a proper amount of grease can be applied to the guide rail 9 and the slide block 8. The camera fixing disk 7 is fixed to a corresponding position by rotating the height adjusting disk 84 back and forth, and the angle of the camera can be properly adjusted by rotating the camera fixing disk 7 so that the center of the camera is directly opposite to the center of the LED display screen 11. The positions and angles of the fans 12 are manually adjusted to enable the positions of the two fans to be bilaterally symmetrical, the positions and angles of the LED lamps 13 can be adjusted, and the two lamps are adjusted in a matching mode to enable the irradiation range of the two lamps to cover one side of the box body 1 where the LED display screen 11 is fixed. In order to simulate the sandstorm environment, the distance between two movable blocks 62 in the wedge-shaped fine sand container 63 is manually adjusted, the effective area of the rectangular groove 67 is changed, which is equivalent to the change of the sand leakage speed into the box body 1, and the appropriate fine sand is put between the two movable blocks 62 of the wedge-shaped fine sand container 63 corresponding to the pollution level of a certain sandstorm dust environment, and is leaked into the box body 1 through the rectangular groove 67 in the middle of the wedge-shaped fine sand container 63 and blown away by the fan 12 to form the sandstorm dust environment, and the working condition of the double (single) -eye intelligent camera 15 in the sandstorm dust environment is tested. In order to simulate different illumination conditions, the brightness and the angle of the LED illuminating lamp 13 in the box body 1 can be adjusted, when the accuracy of judging the depth of field of the double (single) -eye intelligent camera 15 is measured, the sliding block 8 can be moved along the guide rail 9 by moving the sliding block 8, different distances can be obtained, and the degree of the graduated scales 10 on the two sides can be read by the distance reading pointer 83 on the sliding block 8. In order to simulate the rainfall environment, the water outlet 52 on one side of the water storage tank 5 is connected with the water inlet of the water pump, the water outlet of the water pump is connected to a rainfall sprayer through a pipe, the rainfall sprayer is opposite to the rainfall sprayer container 61, the water inlet of the water storage tank 5 on one side is opened, a proper amount of water (generally 2/3 which cannot exceed the volume of the water storage tank) is added into the water storage tank 5 through the pipe, then the water pump is started, the water is pumped out from the water storage tank 5 through the water outlet 52 and flows into the box body 1 through the circular through hole 65 of the box cover through the rainfall sprayer to form rainfall, and the fan 12. The falling water drops flow into ten water collecting grooves 14 arranged at the bottom of the tank body 1, and because the water collecting grooves 14 are deeper and deeper, the falling water drops flow through the water collecting grooves 14 into the water storage tank 5 under the action of gravity in the extended water storage tank groove 53, so that the water is recycled.

Correspondingly, the invention provides an in-loop testing method for an unmanned vehicle-mounted intelligent camera, which comprises the following specific implementation steps of:

step S1: the method comprises the steps of obtaining visual screens to be detected in different scenes, storing the visual screens in a storage device, connecting the storage device carrying the video to be detected with a USB interface on an LED display screen 11, and opening and playing the video.

According to the functional requirements of the tested intelligent camera, finding a video capable of meeting the relevant functions of the test intelligent camera, wherein the video can comprise pedestrians, vehicles, lane lines, obstacles and the like; the storage device is a USB flash disk or a mobile hard disk capable of being connected with a USB.

Step S2: and adjusting the position of the intelligent camera to be measured relative to the LED display screen 11, so that the center of the intelligent camera is aligned with the center line of the LED display screen 11.

The vertical height and the left-right horizontal position of the camera can be adjusted through the camera posture adjusting device, and the intelligent camera can rotate 360 degrees in the horizontal direction. Through clockwise rotation height control disc 84, can reduce the intelligent camera height, the anticlockwise rotation can rise the height of intelligent camera, through the intelligent camera height is adjusted to the mode, makes it keep at same level with LED display screen 11, moves the camera simultaneously through rotating limit bolt 74, realizes the regulation of the left and right sides position of camera, makes the camera can rotate certain angle through rotating camera fixed disc 7 to realize adjusting well in camera center and LED display screen 11 center.

Step S3: the distance between the intelligent camera and the center of the LED display screen 11 is changed, the numerical value of the distance reading pointer 83 pointing to the graduated scale 10 is observed, and the specific distance t is obtained.

The distance between the intelligent camera and the LED display screen 11 can be obtained by moving the sliding block 8 longitudinally along the guide rail 9, changing the distance between the intelligent camera and the center of the LED display screen 11, fixing the sliding block 8 at a specific position through the positioning bolt 82, and reading the scale of the distance reading pointer 83 pointing to the graduated scale 10.

Step S4: by controlling the brightness and the irradiation range of the LED irradiation lamp 13, the illumination condition in the box is changed, and different illumination environments in a real scene are simulated.

The illumination intensity in the box is changed by controlling the brightness of the LED illuminating lamp 13, the direction of the lamp holder of the LED illuminating lamp 13 is adjusted to adjust the illumination range, the light intensity in the box can also be changed, different illumination intensities in a real scene are simulated, and the performance of the camera under different illumination conditions is tested.

Step S5: the sand raising environment in a real scene is simulated by controlling the spreading amount of the fine sand in the wedge-shaped fine sand container 63 above the box body 1 and simultaneously adjusting the wind power of the fans 12 at two sides of the box body 1.

The sand leaking speed and flow are controlled by adjusting the distance between the two movable blocks 62 in the wedge-shaped fine sand container 63, the sand blowing conditions with different strengths are simulated, the wind power of the fans 12 on the two sides of the box body 1 is adjusted at the same time, a more real sand blowing environment is created, and the performance of the camera under different sand blowing conditions is tested.

Step S6: a proper amount of water is added into the water collecting tank 14, the water is pumped out through a water pump and is sent to a rainfall sprayer above the tank cover 6, rainfall is formed through spraying of the rainfall sprayer, the wind power of the fan 12 is adjusted, and the rainy day environment in a real scene is simulated.

Adding a proper amount of water into the water collecting tank 14, wherein under the action of gravity, the water in the water collecting tank 14 flows into the water storage tank 53, one water outlet 52 of the water storage tank 5 is connected with the water inlet pipe of the water pump, the water outlet pipe of the water pump is connected with the water inlet pipe of the rainfall sprayer, the rainfall sprayer is inserted into the rainfall sprayer container 61, under the action of the water pump, the water is pumped out of the water storage tank 53 and sprayed out through the rainfall sprayer to form rainfall, and meanwhile, the wind power of the fan 12 is adjusted, so that the rainfall scene is more vivid, and the performance of the camera in a rainy day environment is tested.

Step S7: different weather in a real scene is simulated by adjusting illumination, rainfall, sand blowing strength or size, photos shot by the intelligent camera under different weather conditions are obtained, and classification types of objects at the marked positions and correct and wrong quantity in the classification types are recorded.

Step S8: and comparing the image depth information given by the intelligent camera with the recorded distance t, evaluating the detection effect and the depth calculation precision of the camera in different scenes, and realizing the measurement of the performance of the camera.

The method comprises the steps of evaluating the object detection performance of the camera in different scenes by calculating the false detection rate of the camera in different scenes, evaluating the precision of the camera in the process of calculating the depth of field by comparing the recorded different distances t with the depth information given by the camera, and evaluating the overall performance of the camera by combining with the use requirement.

The invention provides an in-loop test method and device for an unmanned vehicle-mounted intelligent camera, which are convenient for automobile manufacturers to test the performance of the unmanned vehicle-mounted intelligent camera, have the advantages of simple structure, convenience in installation, low cost, high working safety and high test accuracy, and can completely replace real vehicle test.

Example 1

The device provided by the invention tests two intelligent double cameras A, B, and for the same video source, the intelligent cameras A, B have the same exposure starting time and the same exposure interval, so as to ensure that the two cameras can obtain the same video image. The driving speed of an automobile on an urban road is generally required to be not more than 50km/h, the speed is converted to be about 13.9m/s, the driving distance of the automobile is generally required to be not more than 30cm when the automobile acquires front and rear pictures, so the time interval of the acquired front and rear pictures is not more than 0.021s, the exposure time interval is set to be 0.02s, namely the number of the pictures acquired within 1 second is 50, the length of a played video source is 1 minute, 3000 pictures are acquired by two cameras, wherein each picture marks information such as pedestrians, traffic signs, vehicles and the like, and marks corresponding distance information, in the process of image information processing, considering that the scenes in a video are limited, the information acquired in adjacent pictures has great similarity, the scene overlapping ratio is serious, and therefore 300 pictures are selected as effective samples in a targeted manner, and counting the information of the vehicles, pedestrians, traffic signs and the like and corresponding distance information which are identified in the 300 photos by each camera. The same setting is carried out, the test is respectively carried out in rainy days and windy and sandy days, the recognition capability of the camera is tested, and the obtained data and conclusion are as follows:

in the video used in the test, there were a total of 86 vehicles, 122 pedestrians, and 15 traffic signs.

TABLE 1 comparison of the number of objects recognized by two cameras under normal conditions and their accuracy

Camera type	Pedestrian	Vehicle with a steering wheel	Traffic sign	Summary of the invention
					Binocular camera A	115(94.3％)	80(93.0％)	13(86.7％)	208(93.3％)
Binocular camera B	118(96.7％)	84(97.7％)	14(93.3％)	216(96.9％)

TABLE 2 comparison of the number of objects recognized by two cameras in rainy weather conditions and their accuracy

Camera type	Pedestrian	Vehicle with a steering wheel	Traffic sign	Summary of the invention
					Binocular camera A	107(87.7％)	76(88.4％)	12(80％)	195(87.5％)
Binocular camera B	110(90.2％)	81(94.2％)	12(80％)	203(91.0％)

TABLE 3 comparison of the number of objects recognized by two cameras under the condition of sand blowing and the accuracy

Camera type	Pedestrian	Vehicle with a steering wheel	Traffic sign	Summary of the invention
					Binocular camera A	110(90.2％)	79(91.9％)	12(80％)	201(91.1％)
Binocular camera B	114(93.4％)	82(95.3％)	14(93.3％)	210(94.2％)

TABLE 4 comparison of the number and accuracy of objects identified by two cameras under high-intensity illumination

Camera type	Pedestrian	Vehicle with a steering wheel	Traffic sign	Summary of the invention
					Binocular camera A	98(80.3％)	72(83.7％)	11(73.3％)	181(81.2％)
Binocular camera B	102(93.4％)	75(95.3％)	11(73.3％)	188(84.3％)

According to the data, the precision of the tested binocular camera A is lower than that of the tested binocular camera B, namely the performance of the tested binocular camera B is better, the target recognition rate marked by the A, B camera is 98%, the target recognition precision errors are respectively 4.8% and 1.2% in good weather, the target recognition precision errors are both within an error allowable range, under the rainy day condition, the target recognition precision errors are respectively 10.7% and 7.1%, the target recognition precision errors exceed the error allowable range, and the two types of target recognition precision errors do not meet the requirements; under the condition of sand lifting, the target identification precision errors are 7.1% and 3.9%, the binocular camera B meets the error requirements, and the camera A does not meet the requirements. Under the condition of strong light irradiation, the target identification precision errors are 17.1% and 13.9%, the error allowable range is checked by the target identification precision errors and the target identification precision errors, and both cameras do not meet the requirements.

As can be seen from this experiment, the recognition performance of the binocular camera B is entirely higher than that of the binocular camera a. The target recognition capability of the binocular camera A and the target recognition capability of the binocular camera B under good conditions meet the specification requirements, but the performances of the binocular camera A and the binocular camera B are greatly reduced under strong light irradiation, the specifications are not met, and meanwhile, experiments show that the influence of the light irradiation on the binocular camera is the largest.

All the embodiments in the present specification are described in a related manner, and the same and similar parts among the embodiments may be referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the system embodiment, since it is substantially similar to the method embodiment, the description is simple, and for the relevant points, reference may be made to the partial description of the method embodiment.

The above description is only for the preferred embodiment of the present invention, and is not intended to limit the scope of the present invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention shall fall within the protection scope of the present invention.

Claims (9)

1. An unmanned vehicle-mounted intelligent camera in-loop testing device comprises a box body (1) and a box cover (6), and is characterized in that an LED display screen (11) is arranged at the center of the front side of the box body (1), a fan accommodating device (3) and an LED irradiation lamp accommodating device (4) are respectively arranged on the left side and the right side of the box body (1), a water storage tank (5) is arranged below the rear end of the box body (1), and a guide rail (9), a graduated scale (10) and a water collecting tank (14) are longitudinally arranged on the bottom surface of the box body (1); a water inlet hole (51) and a water outlet hole (52) are respectively arranged at two ends of the water storage tank (5), the water inlet hole (51) is arranged above the water storage tank (5), the water outlet hole (52) is arranged below the water storage tank, and a water storage tank groove (53) is arranged in the water storage tank (5); the guide rails (9) are double guide rails and are symmetrically distributed about the longitudinal axis of the box body (1), the cross sections of the guide rails (9) are I-shaped, and the slide blocks (8) are arranged on the guide rails (9); the two graduated scales (10) are symmetrically distributed about the longitudinal axis of the box body (1), the graduated scales (10) are located on the outer side of the guide rail (9), the 0-graduation position of the graduated scales (10) is flush with the screen of the LED display screen (11) and extends to the tail end of the box body (1), and the minimum graduation of the graduated scales (10) is mm; the front end part of the water collecting tank (14) is aligned with the front end part of the guide rail (9), the rear end part of the water collecting tank (14) is communicated with the water storage tank (53), and the depth of the water collecting tank (14) is gradually increased from front to back; the box cover (6) is positioned above the box body (1), and the wedge-shaped fine sand container (63) and the rainfall spray head container (61) are arranged on the box cover (6);

the center of the sliding block (8) is fixedly connected with a threaded rod (81), and the upper end of the threaded rod (81) is inserted into the camera fixing disc (7) through a height adjusting disc (84); the middle of the height adjusting disc (84) is provided with a threaded hole for allowing a threaded rod (81) to pass through; the camera fixing disc (7) is provided with a camera fixing disc through hole (71) which is larger than the outer diameter of the threaded rod (81) in the center and allows the threaded rod (81) to pass through directly, and the camera fixing disc (7) is connected with a fixing block (75) through a connecting bolt (73); the camera fixing block is characterized in that two limiting bolts (74) are connected to the middle of the fixing block (75) in a threaded mode, the camera fixing block (72) is arranged on the inner side of the fixing block (75), and the limiting bolts (74) are inserted into light holes of the camera fixing block (72) through the fixing block (75).

2. The unmanned vehicle-mounted intelligent camera in-loop testing device of claim 1, characterized by comprising four supporting columns (2) distributed at four feet at the bottom of the box body (1) and directly connected with the box body (1).

3. The unmanned vehicle-mounted intelligent camera in-loop test device according to claim 1, wherein the LED display screen (11) is connected with a power supply through a power line and is provided with a USB interface.

4. The unmanned vehicle-mounted intelligent camera-in-loop test device according to claim 1, wherein the fan accommodating devices (3) are arranged in bilateral symmetry about a longitudinal axis of the box body (1), and a fan (12) is arranged in each fan accommodating device; the LED irradiation lamp accommodating devices (4) are arranged in a bilateral symmetry mode, and one LED irradiation lamp (13) is placed in each LED irradiation lamp accommodating device.

5. The on-board intelligent unmanned camera on-board test device according to claim 1, wherein the slider (8) is in clearance fit with the guide rail (9), the left side and the right side of the slider (8) are respectively provided with two positioning screw holes (85), the positioning bolts (82) can fix the slider (8) at any position on the guide rail (9) through the positioning screw holes (85), the middle positions of the edges of the two sides of the slider (8) are respectively provided with a distance reading pointer (83), and the distance reading pointer (83) is used for indicating the scale size on the graduated scale (10).

6. The on-board intelligent unmanned camera on-board testing device according to claim 1, wherein the wedge-shaped fine sand container (63) is open at both the top and bottom, and the top opening is larger than the bottom opening, and two movable blocks (62) are arranged therein, and a rectangular groove (67) is formed below the movable blocks (62).

7. The unmanned on-vehicle smart camera on-board test device as claimed in claim 1, wherein the rainfall sprayer receptacle (61) is in a conical shape with a small top and a large bottom, and is open at both top and bottom.

8. The unmanned vehicle-mounted intelligent camera on-board test device according to claim 1, wherein the size of the box cover (6) is larger than that of the box body (1), and a circle of guide groove (64) is formed in the edge of the box cover (6) to ensure that the box cover (6) can be stably covered on the box body (1).

9. The method for testing the unmanned vehicle-mounted intelligent camera in-loop testing device according to any one of claims 1 to 8, characterized by comprising the following steps:

step S1: acquiring the video screens to be detected in different scenes, storing the video screens in a storage device, connecting the storage device carrying the video to be detected with a USB interface on an LED display screen (11), and opening and playing the video;

step S2: adjusting the position of the intelligent camera to be measured relative to the LED display screen (11) so that the center of the intelligent camera is aligned with the center line of the LED display screen (11);

step S3: changing the distance between the intelligent camera and the center of the LED display screen (11), and observing the numerical value of the distance reading pointer (83) pointing to the graduated scale (10) to obtain a specific distance t;

step S4: the illumination condition in the box is changed by controlling the brightness and the illumination range of the LED illumination lamp (13), and different illumination environments in a real scene are simulated;

step S5: the sand raising environment in a real scene is simulated by controlling the spreading amount of the fine sand in the wedge-shaped fine sand container (63) above the box body (1) and simultaneously adjusting the wind power of the fans (12) at two sides of the box body (1);

step S6: adding a proper amount of water into the water collecting tank (14), pumping out the water through a water pump, conveying the water to a rainfall sprayer above the tank cover (6), spraying the water through the rainfall sprayer to form rainfall, adjusting the wind power of the fan (12), and simulating a rainy day environment in a real scene;

step S7: simulating different weathers in a real scene by adjusting illumination, rainfall, sand blowing strength or size, acquiring photos shot by an intelligent camera under different weather conditions, and recording classification types of objects at the marked position and correct and wrong quantity in the classification types;

step S8: and comparing the image depth information given by the intelligent camera with the recorded distance t, evaluating the detection effect and the depth calculation precision of the camera in different scenes, and realizing the measurement of the performance of the camera.

Images