Industrial vision inspection
System development for appearance defects, dimensional measurement, OCR identification, assembly error prevention and production line visual positioning.
JIVISION targets industrial manufacturing, intelligent equipment, robots, park security and special inspection scenarios, providing complete delivery from visual algorithm verification, camera light source selection, data training, embedded deployment to prototype development and on-site debugging.
The service content is organized in three directions: visual system development, project pre-production and upgrade, and embodied intelligent data, making it easy to check the corresponding capabilities and delivery boundaries according to the project stage.
Complete machine and instrument, peripheral control, device firmware, brand joint research and development and SDK adaptation.
Image processing, measurement, experimental analysis and equipment algorithm adaptation.
Light point positioning, light beam profile analysis, calibration and measurement application fields.
Data correction, heat map analysis, target recognition and multi-modal fusion.
Customized development of camera modules, vision boards, acquisition interfaces, embedded drivers and dedicated terminals.
System development for appearance defects, dimensional measurement, OCR identification, assembly error prevention and production line visual positioning.
Supports line laser, structured light, binocular/RGB-D, point cloud processing, 3D reconstruction and volume measurement scenarios.
Local inference deployment for RK3588, Jetson, industrial computers, smart cameras and embedded devices.
Covers grasping positioning, hand-eye calibration, pose estimation, visual servoing and robot data collection.
Supports multi-channel video access, cross-camera tracking, personnel positioning, regional intrusion and behavior recognition.
Covering band and device evaluation, simultaneous calibration, cross-spectral registration, fusion algorithms, edge deployment and SDK delivery.
Provide data collection, annotation, training, evaluation, error detection reflow and continuous optimization services.
Build the host computer, Web management background, inspection records, picture traceability, report export and equipment monitoring modules.
Provide on-site installation and debugging, acceptance data, version upgrades, exception troubleshooting and model iteration support.
Develop early project communication materials around requirements clarification, feasibility assessment, implementation boundaries and technical routes.
Carry out compatibility evaluation, secondary development and error detection optimization for existing cameras, light sources, algorithms and software systems.
Provide PoC verification, prototype development and pilot support for non-standard equipment, automated production lines and product teams.
Provide collection specifications, annotation rules, data cleaning, test set construction and model evaluation materials.
Develop collection specifications and organize traceable data assets around robot operation, environmental interaction, and multi-sensor synchronization.
Synchronously collect video, joint status, end pose, control instructions and task results.
Carry out synchronous collection of RGB/RGB-D, depth, IMU, force sensor, odometer and equipment status signals.
Build training and evaluation data sets around task definition, scenario configuration, failure samples, and label specifications.
Carry out simulation scenario generation, real data reflow and domain difference analysis to support training and strategy evaluation.
Organize visual capabilities with scene language to help customers quickly judge business matching.
PCB, semiconductor, 3C parts inspection.
Welding, final assembly, dimensional and assembly inspection.
Package identification, volume measurement, sorting and positioning.
Packaging, coding, foreign matter, compliance testing.
From image processing, deep learning, 3D vision, multi-camera systems to edge deployment, software development and data closed loop.
Detection, classification, segmentation, OCR, keypoints and anomaly detection.
Point cloud registration, fitting, segmentation, size calculation and 3D reconstruction.
Model quantization, NPU/GPU acceleration, low-latency local inference.
Imaging interface, board design, synchronous acquisition, driver adaptation and prototype verification.
The Case Center collects public application and project delivery cases, presenting project issues, technical routes, and delivery content. Customer names are not disclosed, and unconfirmed operating data is not used.
The case of a multi-channel camera personnel positioning platform uses 8-channel network cameras as a reference configuration, covering RTSP access, site map determination, personnel detection, cross-camera association, trajectory visualization, business interfaces and an architecture that expands according to the target scale.
Evaluation of edge vision solutions for fixed unloading areas, covering three types of events: vehicle unloading in non-designated areas, unauthorized personnel picking up goods, and people carrying goods out of the area, as well as 2 to 4-channel cameras, identity authorization and PoC verification boundaries.
JIVISION's visual data simulation case, based on the gantry crane data set and engineering scene requirements, generated 48 gantry crane scene enhancement samples, 40 four-category synthetic target samples, and 8 tunnel, port, and bridge case materials, covering long-tail working conditions such as night, rain and fog, backlight, occlusion, and compression.
JIVISION RK3588 industrial computer smoke and black smoke identification case covers RTSP video stream access, three-frame timing smoke detection, black smoke color discrimination, MCP2221A level linkage, false alarm suppression, on-site deployment and acceptance boundaries.
JIVISION RK3588 smart glasses offline front and side face detection case covers Android Face SDK, Core+Face AAR, Camera2 real-time input, direction output, delivery verification and customer testing boundaries.
A real project delivery case based on openEuler and Huawei Atlas 200I A2, covering 8-channel video, Ascend NPU, AI rules, video playback, and GB/T 28181, GA/T 1400 local protocol closed loop.
Continuously publish professional information and solution content around industry scenarios, technical routes and project issues.
From six aspects: product architecture, language resources, on-site conditions, data boundaries, acceptance documents and after-sales mechanism, the delivery capabilities that should be established before industrial visual inspection equipment enters overseas factories are explained.
Explain the method of decoupling industrial vision inspection software and computing hardware, and how to conduct computing power evaluation, platform migration and acceptance verification based on the complete processing link.
For electronic assembly and industrial components, the inspection task splitting, imaging design, algorithm combination, edge deployment, acceptance method and delivery boundary of multi-language visual inspection equipment are explained.
JIVISION Assembly Vision Inspection Project Case: In view of missed scratch detection, false alarms of structural textures and changes in workpiece posture, registration, surface masking, quality gating and edge-side operation optimization were completed.
For procurement, quality and engineering personnel planning to deploy visual inspection equipment in overseas factories, lists samples, cycle time, site, language, interface, data, acceptance and support materials.
Focusing on the selection of industrial cameras in production line visual inspection, this article analyzes key parameters such as resolution, frame rate, interface type, lens matching and environmental adaptability, and gives selection ideas and precautions based on common application scenarios. It is suitable for reference by technicians who are planning visual inspection projects.
JIVISION takes visual perception as its core, connects algorithms, hardware, data and on-site engineering delivery, and supports technology pre-research, solution verification, customized development, prototype delivery and continuous iteration.
Please fill in the detection object, on-site environment, speed accuracy requirements, and picture or video sample conditions.