Test method of vehicle depot staff practical operation analysis system
By building a practical analysis system for vehicle depot employees and using deep learning models to identify video data, the problem of insufficient quality assurance of practical training has been solved, effective management and quality control of employee practical training has been achieved, recognition accuracy has been improved, and the time cost of manual confirmation has been reduced.
Patent Information
- Application Number
- CN202510802986.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-16
- Publication Date
- 2025-09-16
AI Technical Summary
In the existing technology, the quality of practical training for railway depot employees cannot be effectively measured and controlled, resulting in the inability to ensure the quality of actual operations and posing safety hazards.
Build a practical analysis system for vehicle depot employees. By creating a test video dataset, use a deep learning model to recognize video data, automatically identify and confirm operation status and time, and generate training evaluation files.
It achieves effective management and quality control of employees' practical training, improves recognition accuracy, reduces the time cost of manual confirmation, and ensures the quality of actual work.
Smart Images
Figure CN120653564A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of system testing, and in particular relates to a testing method for a vehicle depot employee operation analysis system. Background Art
[0002] Freight cars are one of the primary vehicles for railway transportation, carrying a vast amount of freight traffic. Ensuring their normal and efficient operation is crucial for safe, stable, and economically efficient railway operations. Currently, the effectiveness of practical training for freight car operators within railway bureaus and rolling stock depots is relatively limited. There is no effective way to measure and control the workload and quality of employee training. This makes it impossible to monitor and control the training process in real time and determine whether it meets requirements. This inability to effectively manage the training process and quality leads to a lack of guaranteed quality during actual train inspections, posing a safety hazard to vehicles. Summary of the Invention
[0003] The purpose of the present invention is to solve the problem that the quality of actual work cannot be guaranteed due to the inability to ensure the quality of employee training, and to propose a testing method for the vehicle depot employee practice analysis system.
[0004] The technical solution adopted by the present invention to solve the above technical problems is: A testing method for a vehicle depot employee performance analysis system, the method specifically comprising: Step 1: Create a test video dataset including normal process video data and abnormal process video data, and store each test video data in the video dataset in the corresponding path; Step 2: The employee implementation analysis system uses the server to load the test model. The server extracts the test video data from the folders in each path and uses the test model to identify each piece of test video data until all the test video data are identified. The server then confirms the identification results of the test video data. Step 3: If the recognition results of all test video data are passed, the final test model is obtained; Otherwise, modify the parameters of the test model, and then use the test model with the modified parameters to return to step 2.
[0005] Furthermore, the video data specifically includes the red flag plugging and unplugging video, the operation video of replacing the 13-type coupler, the operation video of replacing the hose, the operation video of replacing the brake shoe, and the operation video of replacing the 120 valve.
[0006] Furthermore, the specific process of step one is: Step 1: Use on-site cameras to collect video data of the correct operating procedures for each operation project, and also collect video data of flag-raising and flag-removing during the practical training of each operation project; Step 1 and 2: Determine all abnormal conditions of each operation item, and then generate video data of various abnormal operation conditions based on the video data of the correct operation process of each operation item; Step 13: Store the video data of the correct operation process and the abnormal operation process in the folder under the corresponding path.
[0007] Furthermore, based on the video data of the correct operation process of each operation project, video data of various abnormal operation conditions are produced. The specific method is as follows: Step 1: Edit the video data of the correct operation process of each operation project and edit it into MP4 format video; Step 2. Convert the edited MP4 video into a PNG image at the set frame rate. Step 3: Perform PS processing on the PNG format image according to various abnormal operation conditions; Step 4. Restore the PS-processed image to an MP4 format file.
[0008] Furthermore, the method further includes step 4, which is specifically: Staff members use facial recognition to log in to the employee practice analysis system and select the project that requires training on the button page. When executing the selected practice project on site, the on-site camera is used to collect video data of the staff's practice process. After the video data is collected, it is stored in the corresponding recognition path, and the final test model is used to identify the video data extracted from the recognition path.
[0009] Furthermore, confirming the recognition result of the test video data specifically includes confirming the recognition result of the operation state and confirming the operation time.
[0010] Furthermore, the recognition result confirmation of the operation status specifically includes: recognition result confirmation of the operation status of each detection item point and recognition result confirmation of the final status of each detection item point.
[0011] Furthermore, the operation time confirmation specifically includes: the operation time confirmation of each detection item point.
[0012] Furthermore, the definition of passing the recognition results of all test video data is: The operation status recognition results and operation time recognition results of all test cases are correct.
[0013] The beneficial effects of the present invention are: 1. Through a complete test data set construction method, using a combination of normal scenario implementation and possible abnormal scenarios, it can maximize the coverage of scenarios involved when employees carry out practical projects and ensure the accuracy of the recognition program.
[0014] 2. Easier data management and maintenance. For additions or modifications to standard implementation steps for each project, only some test data needs to be replaced or added, avoiding duplication of work and improving the use of test cases. Regular updates to test data sets ensure the timeliness of test data.
[0015] 3. By using the method of testing recognition accuracy and recognition time, the burden on the system can be reduced to the greatest extent, and the time cost of manually confirming the completion of employee training tasks can be reduced.
[0016] 4. The method of the present invention has strong scalability and can be applied to image recognition testing for employee training on railway trains, high-speed railways, subway rail transit, etc., to ensure the quality of practical training, thereby effectively ensuring the quality of actual work during operation. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] Figure 1 The present invention is a flow chart of a testing method of a vehicle depot employee operation analysis system. DETAILED DESCRIPTION
[0018] It should be noted that, unless there is any conflict, the various embodiments disclosed in this application can be combined with each other.
[0019] Specific implementation method 1: Combination Figure 1 This embodiment describes a method for testing a vehicle depot employee performance analysis system, the method specifically comprising the following steps: Step 1: Create a test video dataset including normal process video data and abnormal process video data, and store each test video data in the video dataset in the corresponding path; Step 2: The employee practice analysis system uses the server to load the test model (a conventional deep learning model, such as a convolutional neural network model, can be used). The server extracts the test video data from the folders in each path and uses the test model to identify each piece of test video data until all the test video data are identified. The recognition results of the test video data are then confirmed. Step 3: If the recognition results of all test video data are passed, the final test model is obtained; Otherwise, modify the parameters of the test model, and then use the test model with the modified parameters to return to step 2.
[0020] Through image collection and the use of technical means such as "intelligent image recognition", we can automatically identify employees' personal information and training status, and control the process and quality of replacing certain parts of truck vehicles. We can also automatically statistically analyze employees' training hours, practical frequency and quality, and automatically generate employee practical training evaluation files as an important basis for employees' practical skills, and conduct quantitative analysis of employees' practical work.
[0021] Specific implementation method 2: The difference between this implementation method and specific implementation method 1 is that the video data specifically includes the red flag plugging and unplugging video, the operation video of replacing the 13 type coupler, the operation video of replacing the hose, the operation video of replacing the brake shoe and the operation video of replacing the 120 valve.
[0022] Other steps and parameters are the same as those in the first embodiment.
[0023] Specific implementation method three: This implementation method is different from specific implementation methods one or two in that the specific process of step one is: Step 1: Use on-site cameras to collect video data of the correct operating procedures for each operation project, and also collect video data of flag-raising and flag-removing during the practical training of each operation project; Step 1 and 2: Determine all possible abnormal situations that may occur in each operation item during the actual operation process based on the correct operation process of each operation item, and then generate video data of various abnormal operation situations based on the video data of the correct operation process of each operation item; Step 13: Store the video data of the correct operation process and the abnormal operation process in the folder under the corresponding path.
[0024] Other steps and parameters are the same as those in the first or second embodiment.
[0025] This invention uses a hemispherical network camera to capture video data demonstrating the correct operational procedures. Two cameras captured the insertion and removal of the red flag to determine the start and end times. The 13-type coupler demonstration project was filmed by two cameras, the hose demonstration project by one camera, the brake shoe demonstration project by seven cameras, and the 120 valve demonstration project by four cameras. For example, machine position D01 is responsible for red flag monitoring, machine position D02 is responsible for red flag monitoring, machine position D03 is responsible for upward monitoring of the coupler, machine position D04 is responsible for downward monitoring of the coupler and overall monitoring of the hose, machine position D05 is responsible for monitoring the rear inner brake shoe, machine position D06 is responsible for monitoring the fixture during the replacement of the 120 valve and brake shoe, machine position D07 is responsible for monitoring the shut-off gate of the 120 valve and brake shoe, machine position D08 is responsible for gate regulator monitoring, machine position D09 is responsible for monitoring the rear outer brake shoe, machine position D10 is responsible for monitoring the emergency valve nut, machine position D11 is responsible for monitoring the front inner brake shoe, machine position D12 is responsible for monitoring the front outer brake shoe, machine position D13 is responsible for monitoring the 120 valve nut, and machine position D14 is responsible for facial recognition.
[0026] Based on the standard operating procedures (SOPs) for each implementation project transmitted from the on-site camera, test engineers reviewed the overall project design and outline design specifications, identified all possible exceptions for each implementation project, and designed the use case scope based on the SOPs and exception information. Test case design was then performed for each inspection point. Once the test case design was completed, it was first reviewed and revised internally by the testing team, and then externally reviewed by the project team before the final test case version was finalized. The video data for the complete operation process was copied into three copies, and the original video data file was deleted. One copy of the video data for the complete operation process was stored with the file name "channel number.mp4." The other two copies of the video data for the complete operation process were then cropped, retaining only the last 1 second of video data. The remaining two copies were stored with the file names "1.mp4" and "0.mp4," respectively.
[0027] Specific embodiment 4: This embodiment differs from specific embodiments 1 to 3 in that: video data of various abnormal operation conditions are generated based on the video data of the correct operation process of each operation project. The specific method is as follows: Step 1: Edit the video data of the correct operation process of each operation project and edit it into MP4 format video; Step 2. Convert the edited MP4 video into a PNG image at the set frame rate. Step 3: Perform PS processing on the PNG format images based on various abnormal operation situations (including blocked parts that may appear during the training process, forgetting to perform a certain operation step, performing the operation steps in the wrong order, and ending the operation process before the entire process is completed); Step 4. Restore the PS-processed image to an MP4 format file.
[0028] The other steps and parameters are the same as those in the first to third embodiments.
[0029] Specific embodiment 5: This embodiment differs from any one of specific embodiments 1 to 4 in that the method further includes step 4, which is specifically: Staff members use facial recognition to log in to the employee practice analysis system and select the project that requires training on the button page. When executing the selected practice project on site, the on-site camera is used to collect video data of the staff's practice process. After the video data is collected, it is stored in the corresponding recognition path, and the final test model is used to identify the video data extracted from the recognition path.
[0030] The other steps and parameters are the same as those in the first to fourth embodiments.
[0031] After the system test environment is prepared, the tester's face, work number, group and other information are recorded in the system for subsequent testing.
[0032] Specific embodiment 6: This embodiment differs from any one of specific embodiments 1 to 5 in that: confirming the recognition result of the test video data specifically includes confirming the recognition result of the operation status and confirming the operation time.
[0033] The other steps and parameters are the same as those in the first to fifth embodiments.
[0034] (1) Whether regression testing is required is evaluated based on the test results obtained based on the video action recognition accuracy and recognition time. If regression testing is required, the model parameters need to be modified and the model with modified parameters is pushed to the server. Moreover, the present invention verifies whether the recognition results of a single implementation project and the simultaneous recognition of two implementation projects are accurate, and obtains a test model in which the two implementation projects meet the simultaneous recognition requirements.
[0035] (2) If new test data is needed, add the new test data set to the existing data set, re-edit and photoshop it to generate a video format that can be recognized, re-execute the test process, generate test results and verify the recognition accuracy.
[0036] (3) The test is completed until all test cases pass, that is, the identification criteria are met.
[0037] Specific embodiment seven: This embodiment differs from any one of specific embodiments one to six in that: the recognition result confirmation of the operation state specifically includes: the recognition result confirmation of the operation state of each detection item point and the recognition result confirmation of the final state of each detection item point.
[0038] The other steps and parameters are the same as those in the first to sixth embodiments.
[0039] The inspection items include: (1) Coupler: protective red flag (protective red flags may be placed on both sides of a single end of the training vehicle), coupler tongue cotter pin, coupler tongue pin, coupler tongue, locking iron, locking pin, coupler tongue push iron; (2) Hose: protective red flag (protective red flags may be placed on both sides of a single end of the training vehicle), corner plug, whether the hose is connected to the corner plug, and the direction of the lower connector of the hose; (3) Type 120 control valve main valve and emergency valve: protective red flag (protective red flags may be placed on both sides of a single end of the training vehicle), cut-off valve, semi-automatic relief valve tie rod cotter pin, type 120 control valve mounting nut, emergency valve mounting nut; (4) Brake shoe: protective red flag (protective red flag may be placed on both sides of one end of the training vehicle), cut-off gate, relief rod clamp, brake adjuster detection, brake shoe pin pull ring (brake shoe No. 1 to No. 4), brake shoe pin (brake shoe No. 1 to No. 4).
[0040] Specific embodiment eight: This embodiment differs from specific embodiments one to seven in that: the operation time confirmation specifically includes: the operation time confirmation of each detection item point.
[0041] The other steps and parameters are the same as those in the first to seventh embodiments.
[0042] The platform compares and identifies the operation time of each action returned according to the operation sequence specified by each implementation project. If the operation time returned by the recognition is consistent with the operation time of the normal operation sequence, the operation time of the detection item meets the requirements. It is also necessary to verify the recognition results of one implementation project and the recognition results of two implementation projects separately.
[0043] Specific embodiment 9: This embodiment differs from specific embodiments 1 to 8 in that the recognition results of all test video data pass the definition: The operation status recognition results and operation time recognition results of all test cases are correct.
[0044] The other steps and parameters are the same as those in Specific Embodiments 1 to 8.
[0045] The above examples are merely illustrative of the calculation model and process of the present invention and are not intended to limit the embodiments of the present invention. Persons skilled in the art will readily appreciate that other variations or modifications based on the above description are possible. This list of embodiments is not exhaustive; however, any obvious variations or modifications derived from the technical solution of the present invention remain within the scope of protection of the present invention.
Claims
1. A testing method for a vehicle depot employee performance analysis system, characterized in that: The method specifically comprises the following steps: Step 1: Create a test video dataset including normal process video data and abnormal process video data, and store each test video data in the video dataset in the corresponding path; Step 2: The employee implementation analysis system uses the server to load the test model. The server extracts the test video data from the folders in each path and uses the test model to identify each piece of test video data until all the test video data are identified. The server then confirms the identification results of the test video data. Step 3: If the recognition results of all test video data are passed, the final test model is obtained; Otherwise, modify the parameters of the test model, and then use the test model with the modified parameters to return to step 2.
2. The testing method of the vehicle depot employee performance analysis system according to claim 1 is characterized in that: The video data specifically includes the red flag plugging and unplugging video, the operation video of replacing the 13-type coupler, the operation video of replacing the hose, the operation video of replacing the brake shoe and the operation video of replacing the 120 valve.
3. The testing method of the vehicle depot employee performance analysis system according to claim 2 is characterized in that: The specific process of step one is: Step 1: Use on-site cameras to collect video data of the correct operating procedures for each operation project, and also collect video data of flag-raising and flag-removing during the practical training of each operation project; Step 1 and 2: Determine all abnormal conditions of each operation item, and then generate video data of various abnormal operation conditions based on the video data of the correct operation process of each operation item; Step 13: Store the video data of the correct operation process and the abnormal operation process in the folder under the corresponding path.
4. The testing method of the vehicle depot employee performance analysis system according to claim 3 is characterized in that: The specific method for producing video data of various abnormal operation conditions based on the video data of the correct operation process of each operation project is as follows: Step 1: Edit the video data of the correct operation process of each operation project and edit it into MP4 format video; Step 2. Convert the edited MP4 video into a PNG image at the set frame rate. Step 3: Perform PS processing on the PNG format image according to various abnormal operation conditions; Step 4. Restore the PS-processed image to an MP4 format file.
5. The testing method of the vehicle depot employee performance analysis system according to claim 4 is characterized in that: The method further comprises step 4, which is specifically: Staff members use facial recognition to log in to the employee practice analysis system and select the project that requires training on the button page. When executing the selected practice project on site, the on-site camera is used to collect video data of the staff's practice process. After the video data is collected, it is stored in the corresponding recognition path, and the final test model is used to identify the video data extracted from the recognition path.
6. The testing method of the vehicle depot employee performance analysis system according to claim 5, characterized in that: The confirmation of the recognition result of the test video data specifically includes confirmation of the recognition result of the operation status and confirmation of the operation time.
7. The testing method of the vehicle depot employee performance analysis system according to claim 6 is characterized in that: The recognition result confirmation of the operation status specifically includes: recognition result confirmation of the operation status of each detection item point and recognition result confirmation of the final status of each detection item point.
8. The testing method of the vehicle depot employee performance analysis system according to claim 7 is characterized in that: The operation time confirmation specifically includes: the operation time confirmation of each detection point.
9. The testing method of the vehicle depot employee performance analysis system according to claim 8, characterized in that: The recognition results of all the test video data are defined as: The operation status recognition results and operation time recognition results of all test cases are correct.