Intelligent Inspection Report Generation System

Abstract

A smart inspection report generation system includes a processing unit and a communication unit and a storage unit electrically connected to the processing unit. When the processing unit receives a verification request from a user terminal device, it performs a data preprocessing procedure to obtain processed image data, searches a historical case database based on the processed image data to obtain similar case data, inputs the processed image data into a multimodal model to generate a quality defect portion of an inspection report, and inputs the processed image data, the similar case data, and the quality defect portion into the multimodal model to generate a scene description portion, a quality risk description portion, and an image portion of the inspection report, and then provides the inspection report to the user terminal device.

Claims

1. A smart inspection report generation system, applicable to a user terminal device, comprising: a communication unit for providing network connectivity; a storage unit for storing a data preprocessing program, a quality specification checklist, a historical case database, a multimodal model trained using machine learning, and a quality setting prompt, wherein the quality specification checklist includes multiple check items corresponding to multiple specific objects, each check item being related to a quality specification checkpoint for the corresponding specific object, the historical case database containing multiple case records, each case record including image information corresponding to a specific object and text information related to the specific object; and a processing unit electrically connected to the communication unit and the storage unit, and communicating with the user terminal device through the communication unit; wherein... When the processing unit receives a verification request from the user terminal device containing image data to be verified, it performs a data preprocessing procedure on the image data to be verified to obtain processed image data. The processed image data is related to multiple image information corresponding to multiple target objects contained in the image data to be verified, a field where the target objects are located, scene text information related to the field, and a verification item classification result corresponding to the target objects. The verification item classification result includes one of the verification items. The processing unit searches the historical case database for the image information contained in the case data based on the processed image data to obtain the image information and text information corresponding to the special object that best matches the target objects, and takes the case data containing the image information corresponding to the special object as similar case data. The processing unit inputs the image information of the target object, the similar case data, and the inspection items included in the inspection item classification results of the target object into the multimodal model to generate a quality defect portion of an inspection report; the processing unit inputs the image information of the target object, the field, and the scene text information into the multimodal model to generate a scene description portion of the inspection report; the processing unit inputs the image information of the target object, the scene description portion, and the quality setting prompts into the multimodal model to generate a quality risk description portion; the processing unit inputs the image information of the target object and the quality defect portion into the multimodal model to generate an image portion of the inspection report, wherein the image portion is composed of at least one of the image information of the target object contained in the processed image data that does not meet the quality specification inspection criteria, outlined with a first color, and the inspection report is provided to the user terminal device.

2. The intelligent inspection report generation system as described in claim 1, wherein, The storage unit also stores a special object detection model that has been trained using machine learning. Each case data also includes a field where the special object is located and scene description information related to the field. In the data preprocessing procedure, the processing unit inputs the image data to be checked into the special object detection model to identify the image information corresponding to the target object and the field where the target object is located. The processing unit searches the historical case database for scene description information contained in the case data based on the field where the target object is located to obtain the scene description information that best matches the field where the target object is located as the scene text information related to the field. The processing unit inputs the image information corresponding to the target object and the quality specification checklist into the multimodal model to generate the check item that best matches the target object as the check item classification result corresponding to the target object.

3. The intelligent inspection report generation system as described in claim 2, wherein, The storage unit also stores a safety regulation database and an engineering safety setting prompt. The safety regulation database contains multiple safety regulations, each related to a legal provision for engineering safety. The processing unit inputs the image information of the target object, the scene description, and the engineering safety setting prompt into the multimodal model to generate an engineering safety risk description section of the inspection report. The processing unit inputs the image information of the target object and similar case data into the multimodal model to generate an engineering safety deficiency section of the inspection report. The processing unit inputs the text information contained in the similar case data, the engineering safety deficiency section, and the safety regulation database into the multimodal model to generate a violation section of the inspection report. The processing unit will also input the image information corresponding to the target objects and the engineering safety deficiencies into the multimodal model to generate the image portion of the inspection report. The image portion will also frame at least one of the image information corresponding to the target objects contained in the processed image data that does not comply with the safety regulations with a second color, which is different from the first color.

Images