There are two categories of Vispect products: the Standard Products and the Solutions.

Standard Products are cost-effective devices massively manufactured and applied. A Standard Product system is neat and clean, usually consisting of one host and two/four cameras, to provide function as Front Collision Warning or Blind Spot Detection for vehicles and machines. Standard Products are made in our own factory.

Solutions refer to complex systems,consisting of several hosts, remote data center (if needed), and more sensors as camera, LiDAR, Gyro, and etc. Solutions are usually customized in small quantity but at higher costs. The parts of a Solution come from Vispect and other brands, while Vispect integrates the parts into a system.



Above: Subway cabin of TRT.
Above: Vispect system installed in TRT subway cabin.
Above: Subway driver fatigue detected by Vispect system.
Above: TRT subway station.
Above video: Vis system detected the subway driver’s semaphore when the train arrived a cross.

Vispect provides 40 sets of AI systems to TRT line-4. One system for one locomotive.

The functions include:

to monitor the train driver's fatigue status, distraction, and yawning.

to check the driver's semaphore before the train starts to move.

Abnormal conditions trigger alerts, visible and audible to remind the drivers, while reports with video footage are sent to the management center through subway intranet. This system has been certificated according to railway electronic OEM standards.

Above video: HMI of Vis system showed there was no obstacle on the rail in subway tunnel.
Above video: HMI of Vis system showed the abrasion of the over-head power line.

Vispect provides Route-Condition-Inspection System, to improve tunnel/route/wall/wire maintenance, for subways.

The system is installed on maintenance cart, to detect:

Abrasion of over-head power lines, at ±1.0mm

obstacles infringing train track

dropped screw/small stone on runway

ring power net overheat

The system consist of:


thread Laser

multi-plane LiDAR

high-speed matrix camera + laser lights

infrared camera + wireless temperature sensor


This system has been deployed on TRT line-6 on the line-sweep locomotive, for the purpose of route/track maintenance.

Above video: Vis AODS system was detecting obstacles on rail. The left is the original signals of the LiDAR, and on the right, People and box were separated by algorithm from the background and marked in green.

Automatic Obstacle Detection System

the brain of Driverless-Train or Train-Driving-Assistance-System, is to detect the obstacles that may threaten the train's approaching, including another train on the same track and track-side items infringing the train's borderlines.

Interface with ATP (Automatic Train Protection system) to brake (option)

Electronic map from ATP to adjust to slopes and crosses (option)

Multi-plane LiDARs

Stand-alone system. Can work without trackside facility like radars

To ensure its security and reliability, this system is being designed, tested, and certificated at SIL2 standards (under certificating)

Above video: Vis segmentation algorithm separated the rails from the background to detect the route of the subway.

Detection Distance (under standard environment):

other trains up to 200 meters

rail track up to 150 meters

box (30 x 30 cm) at 60 meters

people up to 80 meters

about img
Above: Palm fruits and leafstalks were detected by camera and LiDAR.


Vispect is an expert at industrial and agricultural robots for field applications.

In response to the various needs, we integrate different types of sensors, including visible light cameras, structured lights, infrared cameras, 3D cameras, LiDARs, etc., to form multi-level and multi-sensor fusion detection systems. The systems detect and identify various targets, accurately measure the distances of the targets, convert the coordinate system, and manipulate the mechanical arms and devices to act.

Project 1 Photovoltaic Panel Installation Robot (under development)

Large-scale photovoltaic power plants are mostly located in sunny Gobi or deserts. In the process of building the plants, solar panel installation is a challenge - the fitting between the mounting bracket and the solar panels is required to be highly accurate, but under the strong sunlight, the installers are easily fatigued, thus sometimes damages to the panels and even injuries to installers are caused.

Above: A typical large-scale photovoltaic power plant.

In order to solve this problem, Vispect has been cooperating with the Zhejiang University Robotics Institute to jointly develop a new robot. This robot can accurately detect the position of each component of the mounting bracket, and then automatically grab the solar panel, and fit the solar panel on the bracket with a detection accuracy of ±1 or 3 mm. Behind this installer robot is a escort robot that carries solar panels and autonomously follows the installed robot.

In the first stage of this project, semi-automatic working mode is adopted: a installer drives the robot to the working place, and the robot automatically fits the panels onto the mounting brackets one by one around the working place. The project includes othr functions as GPS-RTK Positioning System, 4G Communication System, Management Platform, Task Release Interface, Path Planning, Manual Takeover Interface, etc., and in the second stage of this project, 7X24 driver-less installation can be applied.


Above: Schematic diagram of the Photovoltaic Panel Installation Robot.

Project 2 Palm Fruit Picking Robot (under development)

Palm fruit is used to extract palm oil. Palm fruit picking is currently carried out by hand. Because palm orchards are often located in remote areas, it is difficult to recruit enough people. Picking robots can greatly solve the problem of manpower shortage. First, the worker drives the robot to the palm trees, and the robot automatically identifies, locates, and cuts the palm leaves to expose the palm fruits. The robot then identifies and locates the fruit stalk, and cuts the palm fruit from the stalk to complete the work.


Above: A palm fruit reflected and labeled by a LiDAR during an in-house test.


Above: A palm fruit identified by a camera during an in-house test.

Above: A Vispect system measured the height of parking lot, to avoid top-hitting of big cars.


Various needs of Infrastructural Management can be met by customized solutions based on Vispect technology.

Above: A Vispect system measured the heights of the obstacles crossing over the route of the vehicle.


Height Detection is a need of ambulances and fire trucks, as these vehicles sometime must enter small lanes. There are items, such as power lines crossing above street, that may be lower than the vehicles. So the vehicles need a system to detect if they are going to hit the power lines.

On the other hand, footbridges and overpasses are occasionally hit by high vehicles, so these infrastructures need to detect if the oncoming vehicles are higher than them.

We are using LiDAR installed on the bottom of the footbridge (or on the top of ambulance), to detect if the oncoming car (or the bottom of footbridge) is too high (or too low).

Above video: A Vispect system measured the heights of the obstacles crossing over the route of the vehicle.


On the other hand, footbridges and overpasses are occasionally hit by high vehicles, so these infrastructures need to detect if the oncoming vehicles are higher than them.

We are using LiDAR installed on the bottom of the footbridge (or on the top of ambulance), to detect if the oncoming car (or the bottom of footbridge) is too high (or too low).

Above: A custom-designed camera (left) and a Host (right) by Vispect, to meet strict EMC/EMI requirements of the client.


Vispect is proving other customized systems to our clients. An interesting example is here.

A prism camera, with two channels of image outputs, has a 5M pixel COMS and two light entrances each with a shutter as a reflection mirror. The shutters are crystal liquid, so they can be turned bright or dark by very small electric currents. While the shutters shifting between bright and dark, the CMOS senses the lights or not. When one shutter is bright, the other shutter is set as black. Thus at any moment, just only one entrance let lights arrive the CMOS. Of course, supplying power to the shutters and outputting image to the channels are strictly synchronized by just the same one controller. When the shifting frequency exceeds a certain value, you will see both channels have clear image outputs.

Above video: a custom-designed prism showed little latency when switching from the two lenses.