Capteurs LiDAR : Guide complet de la technologie et des utilisations

Every modern autonomous car, every USGS flood map, every forestry carbon-stock estimate, and every warehouse AGV dodging pallets shares a common sensor class. LiDAR sensors turn pulses of laser light into geometrically exact pictures of the world — distance by distance, point by point. This technology has existed since the 1960s, was famously bounced off retroreflectors on the Moon by Apollo 11 in 1969, and has spent the last decade dropping from five-figure industrial prices to three-figure commodity units. What was once an instrument reserved for NASA and national surveys now sits inside iPhones and factory robots.

Here is an introduction to what the sensor really is, how a LiDAR sensor works physically, the four scanner families you will see on data sheets, the organizations that run on LiDAR today, and what to look at on a spec sheet before you make purchase.

What Is a LiDAR Sensor? And Where the Name Came From

A LiDAR sensor is an active remote-sensing instrument that emits pulses of laser light, listens for the reflection from a target, and calculates distance by timing the round trip. It stands for “Light Detection and Ranging” — defined in those exact words by NOAA’s Ocean Service, which describes lidar as “a remote sensing method that uses light in the form of a pulsed laser to measure ranges (variable distances) to the Earth.”

The concept predates the modern sensor by decades. On 21 July 1969, Apollo 11 crews set out the first array on the moon a laser-ranging e×periment, utilized by researchers even today. Apollo 14 and 15 deployed additional arrays.

Apollo 14 and 15 added more arrays; the Apollo 15 retroreflector, deployed in 1971, contains 300 quartz-glass corner-cube prisms, the largest of the Apollo reflectors. Ground stations on Earth bounce laser pulses off these reflectors and time the return, measuring the Earth–Moon distance to within millimetres. The experiment has run for over 55 years and is documented by NASA’s Jet Propulsion Laboratory.

Exact same time-of-flight physics measuring the 215, 000 miles from the Earth to the Moon drives the tiny LiDAR sensors of each factory robot.

How LiDAR Works: The Time-of-Flight Mechanism

On the physical end, all pulsed LiDAR does is the same thing over and over: fire a short laser pulse, await the reflection, measure the time, and divide. Formula is as simple as the physics get:

The numbers are small. Light propagates at approximately 30cm/ns so the 100 m measurement has a round-trip of approximately 667 ns. LiDAR pulses are typically 1-10 ns and the timing electronics that count the returned pulse require psec-scale accuracy to get to the centimetre. A LiDAR system design challenge is not the physics – even the shortest wavelengths used have nanoseconds of focus time – but enclosing this timing accuracy, plus a laser source, detector, beam-steering system into a hardware form factor that survives a car hood or factory ceiling. Back at the detector, each return is reflected light carrying a time-stamp distance message.

Each pulse fires and returns and adds one point to output. Today, hundreds of thousands – and not uncommonly, millions – of pulses per second are directed to scan across the field of view in order to generate a signature output of this system: a 3D point cloud, or for automated license plate monitoring, a 2D lidar map. The NSF NEON LiDAR tutorial walks through this geometry in detail for remote-sensing applications.

The Four Components of a LiDAR Sensor

If you open any LiDAR datasheet, four building blocks will appear in various combinations. NOAA names three – laser, scanner, and GPS receiver – and operational systems introduce an additional: the inertial measurement unit that tracks the sensor’s orientation between GPS fixes.

Indoor or short-range industrial units frequently make no mention of GNSS – an AGV in a warehouse will rely on odometry and landmark-based localisation rather than satellite fixes. Automotive unit packages combine the sensor with a dedicated processing unit which handles the high data-rate of millions-of-pulses-per-second and produces classified object lists rather than raw point clouds.

Types of LiDAR Sensors: 2D vs 3D, Mechanical vs Solid-State

Datasheets categorize lidar sensors along two axes: dimensionality – 2D or 3D – and scanner architecture – mechanical, MEMS, OPA, flash, FMCW. 2D lidar emit along a single plane – suitable for area-monitoring and AGV navigation. 3D lidar emit across a vertical plane as well, yielding the volumetric point cloud necessary for automotive autonomy or forestry canopy analysis.

Within the 3D subgroup, the scanner architecture is the factor for cost, ruggedness and configuration. A 2025 overview in Nature Light Science & Applications (s44310-025-00059-4) compares the various architectures and their relative advantages.

In the cost space, the landscape has changed dramatically. Yole Group’s 2024 automotive-LiDAR industry forecast said that the average selling price for Chinese automotive-lidar was US $450-500 in 2024, noticeably lower than 2022. Industry practitioners tracking the Chinese market document unit prices as low as US $138 -a reduction of 70 in two years for this segment. Global solid-state LiDAR industry reported revenue of around US $1.88 billion in 2024, and was expected to expand to the US $20-25 billion region in 2033 according to Straits Research.

LiDAR Applications Across Industries

The “LiDAR for autonomous vehicles” perspective ignores most of this field. Today, the sensor is part of a complex of at least seven application groups:

The largest coherent LiDAR dataset in existence is likely the USGS 3D Elevation Program (3DEP), which is systematically acquiring airborne LiDAR coverage of the entire United States at quality level QL1 or QL2 — roughly ±10 cm vertical accuracy at point densities of 8 points per square metre or better. Raw point clouds are hosted publicly on the AWS Registry of Open Data. Any infrastructure planner, hydrologist, or forester working in the US can draw directly from that dataset without flying their own survey.

LiDAR vs Radar vs Camera vs Photogrammetry

LiDAR rarely exists in isolation; on autonomous cars it calibrates with cameras and radar; on drones it replaces photogrammetry for dense vegetation sites; in industrial safety zones it replaces pressure mats and light curtains. Optimization depends on what the sensor perceives and how accurately.

In the real-world, most sensor stacks combine two or three within. E.g., an autonomous car might employ LiDAR for 3D geometry, radar for velocity in rain, and cameras for traffic-sign classification; a forestry drone may employ LiDAR to access below the canopy and the multispectral camera for tree species ID.

Key Specifications: Range, Accuracy, Point Density

When considering a LiDAR sensor, four metrics on the spec sheet significantly impact cost.

• ✔
  Gamme — maximum distance at which the sensor returns useful data on a specified target reflectivity (usually 10% Lambertian). Automotive spec sheets list both “maximum range” (high reflectivity) and “@ 10% reflectivity” ranges; the lower number is what matters for dark-car detection.
• ✔
  Accuracy — systematic error of an individual range measurement. USGS 3DEP QL1 specifies ±10 cm vertical RMSE for airborne deliverables; industrial 2D safety LiDAR achieves ±3 cm at short range.
• ✔
  Point density — points per square metre on the target surface. Airborne 3DEP-class surveys deliver 8+ points/m². Mobile mapping LiDAR reaches 500-1,000 points/m². Terrestrial static scanners achieve 10,000+ points/m² at close range.
• ✔
  Field of view & pulse rate — together these define how fast the sensor can cover an area. A 905 nm automotive LiDAR typically fires 600,000 to 2 million pulses per second across a 120° × 25° FOV.

How to Select a LiDAR Sensor

Criteria selection is a matter of choosing operational characteristics that fits the spec sheet. Several industrial and commercial procurement models can be described within.

1. Range — maximum consistent operating distance (in the same reflectivity as the objects you plan to observe e.g., dark pedestrians 10% Lambertian; metal pallets higher). Indicate the range at actual target reflectivity, not a generic ‘maximum range’.
2. Accuracy and precision — systematic error and repeatability. Survey grade require cm; perimeter safety-area monitoring needs dm. Select as needed.
3. Point density and frame rate — static object detection functions at lower density; obstacle avoidance require high PPs, high pulse frequency and low latency.
4. Operating environment– indoors/ outdoors, humidity, temperature, vibrations/shock. IP ratings, MTBF specs are relevant here. Automotive LiDAR operates in different thermal cycles than air-conditioned-airport-agv.
5. Budget and supply chain — unit pricing have dropped, but integration costs remain (installation, cabling, data processing). OEM/factory-direct sources in China deliver the best price per unit, while European and Japanese firms tend to be the most mature automotive-qualified suppliers, in terms of certifi

For industrial LiDAR sensor procurement specifically — 2D area scanners, 3D mechanical units, solid-state modules, and custom OEM variants — see our LiDAR sensor product line, which walks through product-family specifications and integration examples. CCH Shanghai Sensing brings 20+ years of industrial safety sensor manufacturing experience and provides OEM / factory-direct industrial sensors from China with free test samples on request.

Foire aux questions

Summary and Next Steps

LiDAR has moved from moon-bouncing experimental physics in 1969 to commodity automotive hardware in 2025. Sensor technology is conceptually simple — pulses of laser light, precise time measurement, and the speed of light — but the engineering behind a modern unit packs a laser source, scanner, photodetector, GNSS, and IMU into shapes that fit everywhere from phones to aircraft. Choosing one is less about the technology itself than about matching range, accuracy, point density, environment, and budget to the specific problem.

For industrial-safety and automation readers evaluating specific procurement options, our engineering team can help scope the match between application and sensor grade.