On September 15, Hesai Technology unveiled its MT series LiDAR for high-precision 3D modeling at InterGEO 2026. One noteworthy change is the elevated priority given to spatial resolution. Although this LiDAR is primarily designed for high-precision modeling and high-accuracy 3D reconstruction rather than serving as the primary LiDAR for passenger vehicle autonomous driving, there is still much worth discussing.
Official data reveals that the MT series achieves a 70% reduction in spot area at a distance of 20 meters compared to the previous generation XT, with a ranging accuracy of 3.5 mm. This might lead many to wonder: since LiDAR is inherently used to acquire 3D spatial information and can already "see" objects, why can't it clearly resolve the details? In fact, for LiDAR detection, seeing an object and accurately reconstructing it are not the same thing.
01. What Is Still Missing After LiDAR Can "See"?
For LiDAR, acquiring the distance to a target only solves the problem of where the target is located. However, during actual 3D reconstruction, it is also necessary to know the object's edges, surface shape, distances between adjacent structures, and the relative positions of these structures in space. Especially in scenarios involving pipelines, building edges, equipment interfaces, and narrow gaps, if the spatial sampling is not fine enough, the resulting point cloud may capture the general outline of the target but lose many details.
This introduces an easily overlooked concept: spatial resolution. Simply put, spatial resolution focuses on the LiDAR's ability to distinguish closely spaced structures in space. Note that spatial resolution is not a unified metric that can be fully defined by a single parameter; rather, it is a comprehensive set of capabilities related to spot size, ranging accuracy, and sampling methods.
This is also why LiDAR performance cannot be judged solely by point rate. The point rate determines how many points can be acquired per unit of time, but more points do not necessarily mean more details; the ability to distinguish adjacent structures also depends on spot size, angular resolution, ranging accuracy, and sampling methods. If a single sampling covers a large spatial area, adjacent structures may still be blended together.
With spot size and angular resolution unchanged, increasing the number of points can improve sampling density, but it does not necessarily change the system's ability to distinguish two adjacent structures. Therefore, when the goal shifts from merely acquiring an object to accurately reconstructing it, the first problem to solve is actually whether adjacent fine structures can be distinguished in a single measurement.
02. Why Is a Smaller Spot Size the Starting Point for Improving Spatial Resolution?
Official data reveals that the MT series achieves a 70% reduction in spot area at a distance of 20 meters compared to the previous generation XT. It uses a smaller spot to reduce the area covered by a single laser pulse, thereby delineating object boundaries and structural details more finely and reducing mixed returns between adjacent targets. To put the benefits of a smaller spot simply: after the LiDAR emits a laser pulse, it obtains the target distance through the return signal. If a spot covers a relatively large spatial area, the probability of the same pulse hitting multiple adjacent structures increases.
When targets such as pipelines, building corners, and equipment edges are very close to each other, the returns from different structures are more likely to mix, and the boundaries in the final point cloud may become blurred. With a reduced spot size, the area covered by a single measurement decreases, and the spatial sampling scale of the system for fine structures also decreases accordingly, which is more conducive to resolving details. For high-precision 3D modeling, a smaller spot provides a better chance to distinguish structures that would otherwise be easily blended together. However, there is another key issue here: even if two adjacent structures can be sampled separately, if there is a large error in the distance measurement for each point, these points may still exhibit positional deviations in the distance/depth direction when placed back into 3D space.
03. From Distinguishing Structures to Accurate Positioning: Why Is Higher Ranging Accuracy Still Needed?
A smaller spot solves the problem of whether fine structures can be better distinguished, but resolving the structures is only the first step. If there is a large error in the distance measurement for each sampling point, these points, even if they come from different targets, may still exhibit positional deviations when placed back into 3D space. For scenarios requiring precise reconstruction of geometric structures, such as buildings and industrial facilities, this will further affect point cloud registration, stitching, and 3D reconstruction. Therefore, the MT series achieves a ranging accuracy of 3.5 mm. According to official data from Hesai Technology, this is a 30% improvement over the previous generation XT, and it is directly linked to the accurate reconstruction of point cloud surfaces, geometric profiles, and 3D models.
Here, it is necessary to distinguish the two technical points more clearly. A smaller spot solves the problem of whether structures can be resolved, while ranging accuracy solves the problem of whether the positions in the distance/depth direction can be accurately expressed after they are resolved.
Imagine a scenario where a factory building may contain a large number of pipelines, cables, equipment, and support structures simultaneously, and the spatial relationships between them are often very complex. The LiDAR must not only sample these structures individually but also reconstruct the distances, dimensions, and relative positions between them as accurately as possible; otherwise, the value of the final 3D model will be compromised. But even then, the problem is not completely solved. Even if the spatial position of a single point is accurate enough, if there are many unsampled areas during the scanning process, the final point cloud may still be incomplete.
04. From Accurate Positioning to Comprehensive Sampling: Why Is Continuous Densification Still Needed?
In its official introduction, Hesai Technology also mentioned infinite uniform dense scanning. This means that as the scanning continues, new sampling points are continuously added to previously uncovered areas, while making the point cloud distribution more uniform, reducing data gaps caused by local over-densification or local sparsity. It solves not the single-point measurement accuracy, but the completeness and uniformity of spatial sampling.
For example, when measuring in a dense forest, there is massive occlusion and narrow gaps between leaves, branches, and the ground. The system needs to obtain as much effective sampling as possible from these spatial relationships. According to the official introduction from Hesai Technology, the small spot of the MT series can reduce mixed returns and improve the detection capability of vegetation gaps, thereby acquiring data on the terrain beneath the canopy and multi-layered vegetation structures. In scenarios such as large buildings and industrial facilities, continuous scanning can constantly supplement previously under-covered areas, making the final point cloud more complete.
05. Concluding Remarks
Looking at it this way, the three technical points of the MT series can be connected. The small spot makes fine structures easier to distinguish; the high ranging accuracy allows the resolved structures to be positioned more accurately in the distance/depth direction; and the infinite uniform dense scanning further integrates these local pieces of information into more complete and uniform 3D spatial data. Revisiting the spatial resolution capability of LiDAR, it is not just a case of "the more points, the better"; it must simultaneously solve the three problems of structure resolution, geometric accuracy, and completeness of spatial sampling.
#LiDAR